from:"Howard B"

 wisely and when should systems be switched off, reoptimized, etc.
 If anybody knows some books on this, Iâ#65533;#65533;d be very happy to
 read up on it so if someone knows an author, please post! So far, having
 many systems with different logics is the direction Iâ#65533;#65533;m
 going right now.
  
   Something thatâ#65533;#65533;s worth having a look inside is this
 one:

  
  
  
  
 http://www.amazon.de/Trading-Systems-Development-Portfolio-Optimisation/dp/1905641796
  
   again, itâ#65533;#65533;s scratching the topic only.
  
  
  
   The way I develop systems is always the same (If anyone
 doesnâ#65533;#65533;t agree with this approach, Iâ#65533;#65533;d like
 to know why)

  
  
  
   1 contract, share, etc.
  
   Commissions excluded
  
   3-5 (5max!) input parameters, such as indicators, maâ#65533;#65533;s:
 itâ#65533;#65533;ll result in a raw system with in-built exits and entries
 but no stops.

  
   Optimize the inputs for STABILITY, a profitable set of parameters with
 a stable surrounding region
  
   Plug-In Commissions
  
   Add filters
  
   Add stops/targets according to Sweeneyâ#65533;#65533;s MAE MFE
  
   Plug-in the position sizing algo, â#65533;#65533;the
 compounderâ#65533;#65533;

  
   Test out of sample
  
  
  
   Works fine for me.
  
  
  
   After this, I pretty much try to combine systems. Currently this is a
 bit troublesome in Ami, but I shall let you know when I have done enough
 research, unfortunately I believe itâ#65533;#65533;s gonna take me some
 months.
  
  
  
   What I believe to be very controversial is the topic
 â#65533;#65533;adding noiseâ#65533;#65533; to data. I
 donâ#65533;#65533;t think it reflects human emotion anymore. Same as
 dealing with a random sequence: a coin toss, e.g., can be displayed as a
 graph too, but itâ#65533;#65533;s still a normal distribution. So: When
 doing quant-trading and believing in it, you make one big assumption:
 Markets behave logically, sometimes. I found out that markets contain a
 large amount of noise, random behavior. But there are occasions when people
 get hit on the wrong spot and start buying or selling aggressively .
 Thatâ#65533;#65533;s the time when my systems jump in and make the kill -
 then Iâ#65533;#65533;m out. This behavior cannot be found in a random,
 probably noisy set of data.
  
   Monte Carlo pretty much changes the trade sequence only, right now I
 cannot really see why some people think itâ#65533;#65533;s so useful. It
 might give you a better estimate of a historic drawdown, but I achieve the
 same by multiplying my non-monte-carlo-drawdown with say 1.5â#65533;¦

  
  
  
  
  
   Greetings from Germany,
  
  
  
   Matthias
  
  
  
  
  
   From: amibroker@yahoogroups.com amibroker%40yahoogroups.com [mailto:
 amibroker@yahoogroups.com amibroker%40yahoogroups.com] On Behalf Of
 Howard B
   Sent: Montag, 26. Juli 2010 16:33
   To: amibroker@yahoogroups.com amibroker%40yahoogroups.com
   Subject: Re: [amibroker] Trading Systems, Position Sizing and Monte
 Carlo Analysis
  
  
  
  
  
   Hi Sohamdas --
  
   In my opinion, this is definitely a topic that deserves discussion in
 the AmiBroker forum.
  
   What position sizing should be used during backtests?
  
   If you will be evaluating each trade for its characteristics -- entry
 efficiency, exit efficiency, and so forth, then each trade should be the
 same size. For stocks and ETFs that means the same dollar amount. For
 futures that means the same number of contracts.
  
   If you will be comparing equity growth over a period of time to other
 alternatives, then you will want position sizing and / or compounding to
 some degree. For example, if you want to compare the results of a trading
 system to buy and hold, you will want to take the same size position at the
 beginning if the test period for each alternative, then compare equity
 smoothness, growth, drawdown, etc.
  
   If you are planning to use aggressive position sizing, there are
 several things to consider.
   1. I cannot state it too often -- your system must have a positive
 expectancy measured on strictly out-of-sample results. You absolutely cannot
 use in-sample results to estimate the likely future performance of a trading
 system in any event. And if aggressive position sizing is based on in-sample
 results, you will go bankrupt.
   2. Traders should have a business plan in place. They should know when
 to quit -- either when they have enough that they no longer need to trade,
 or when they have lost so much that they can no longer trade or realize that
 they should pick another profession.
   3. Aggressive position sizing depends on having:
   A. Positive expectancy.
   B. Understanding of risk. Both the risk that is acceptable for each
 trade from the account, and the risk associated with the trading system.
 Most trading systems have higher risk per trade than the account risk
 allows, so even taking a position that is everything you can afford to buy
 is aggressive

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

Greetings --

h3po wrote:
I can see one problem with modelling equity curve via Bootstrapmethod
described in the thread; the draws will be *independent** *(not correlated).
This will give less severe drawdowns in a simulation if the real situation
is that consequtive wins/losses are autocorrelated 

---

You can use statistical tests to determine the degree to which a time series
(either / both the raw data and the equity curve resulting from trading it)
is autocorrelated.  If there is a positive autocorrelation, gains (in either
price or equity) tend to follow gains and losses to follow losses.  If a
series is positively autocorrelated, trend following systems are more
profitable than mean reversion systems.  If there is a negative
autocorrelation, gains then to follow losses and losses to follow gains.  If
a series is negatively autocorrelated, mean reversion systems are more
profitable than trend following systems.

Thanks,
Howard


On Tue, Jul 27, 2010 at 5:21 AM, h3po h3p...@gmail.com wrote:



 I can see one problem with modelling equity curve via Bootstrapmethod
 described in the thread; the draws will be *independent** *(not
 correlated). This will give less severe drawdowns in a simulation if the
 real situation  is that consequtive wins/losses are autocorrelated

 Actully this gives idea for another interesting question: If one is able to
 model *stockprices* by means of a simple model, could this (monte carlo)
 simulated dataseries be valuable as additional data and be used as more OOS
 data or will it not give any new information in the testing process?
 (Asuming one is doing a WF test. If Yes, then this would be valuable for
 end-of-day trading system development as the amount of data here is usually
 limited). Has this been  discussed on this forum before?

 best regards


 --- In amibroker@yahoogroups.com, sohamdas soham...@... wrote:
 
  It was such a solid read,that I took a lot of time to process
 it...almost a day(kiddin, didnt know, it would attract so many high quality
 replies this soon)
 
  Many thanks, folks. Especially Dr. Bandy, Matthias,Lionel, Raymond and
 others..
 
  I am a programmer myself, and have a fair bit of mathematical background
 myself. In the past, I have developed a few trading systems(3 to be precise)
 and traded them myself.
 
  In essence, I found when delving with works of Vince, I found he uses the
 Kelly's Ratio extensively (till the pages I could crack, beyond which I lost
 my patience.)
 
  My experiments with Kelly's Ratio has yielded extremely mindshattering
 drawdowns, but yes, it does compound fairly quickly. So, my internet
 searches, with Fixed Fractional Ratio led to something called SubOptimal
 Ratio, in other words the Half-Kelly Ratio.
 
  But in some ways, I was not deeply convinced or sold on this. You see,
 in many ways, I am a believer in the axiom
 less-crowded-it-is-the-better-it-is. And this is not only in the way, of
 trading ideas, but also in methodologies.
 
  So my view is, out of an ensemble of 100 traders, 92 search for the Hail
 Mary Indicator/Entry/Setup and 6 of the rest search for the best possible
 exit.
 
  And I would say, you can have a worthless random entry, but a good exit,
 its possible you might have a better expectancy than with worthless exits.
 
  My belief, the pursuit of bettering entries and exits is surely a worthy
 one, but having a great rock solid position sizing and risk management
 built, can literally give the real edge to a trading system.
 
  My idea, of a really creative position sizing system is, the way Turtles
 built up their positions.
 
  A few of my own observations:
 
  1. When you read it, on paper, it seems a pretty tame pyramiding, but
 doing it in reality, will quickly tell you how aggressive it is.
 
  2. The position sizing is very apt for trend following. It creates a huge
 position and in such a pace as to take the maximum out of a trend(in my
 language: trend following is a slow system, not a bad system, slow
 system).
 
  3. You can't use the same position sizing of building in 1/2 ATRs in fast
 trading systems(most notably, mean reversion systems).[Again, note fast
 trading system is not necessarily a good system. It just is... fast.] The
 market moves too quickly, in case of a bo failure. You need to have profit
 stops at place as well.
 
  So coming back to the scenario, I was thinking, the main gold might lie
 in having really unique position sizing methodology, which includes position
 sizing,pyramiding(scaling in), scaling out(in my limited opinion, gives a
 better and smoother equity curve when dealing with fast systems), and risk
 management.
  --
  Which drives me to a very (I cant stress it enough) important pertinent
 and thought provoking point by Dr Bandy:
 
  Treat it like a business. Know when to quit.
 
  Doctor,I have got a piece of paper, which follows a series of steps I
 should take, if and when, god forbid a black swan strikes me.

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

Greetings --

Sohamdas wrote:
Dr Bandy, what you said is true, about the second alternative. The risk
conditions are violated, with further scale-ins,when we double up etc.

But consider, if with each scale-in, we also move the stop loss point.In
that scenario, the situation can morph into one, where the risk doesnt
increase linearly with position,it increases slower than the growth of
positions.

---

I agree that the risk calculations change as the price moves in the
direction that gives profit.  But in order to keep to the original risk
profile, the amount of the scale in is limited to the amount of the gain.
That is, before scaling in, recalculate the position size permitted and
scale in by that amount.  The amount of additional position size allowed is
usually much less than the original position was.

My argument against scaling in is with schemes that take a full position at
entry, then add an amount that is large relative to the initial position
(100%, 50%, 33%, or 25% are typical).  Unless the open profit supports the
new position size of 200%, 150%, etc, the risk is now higher than originally
permitted.  It is OK to do this if you realize that by doing so you are
using aggressive position sizing and have worked through the consequences.

The original position could be taken at a position size that anticipates all
of the scale in trades will be taken.  This assures that risk will remain
limited.  But for those trades when the conditions necessary to add a scale
in position do not occur, only a portion of the funds have been committed to
the trade and the profit is less than it would have been if a full position
was taken at the original signal.

I think it is better to treat each signal -- the original and each scale in
-- as a unique trading system.  Work through the design, testing, and
validation of each separately, then apply position sizing to each and to the
portfolio that results from trading all of them.  Be aware that all of these
systems will be trading the same data series in the same direction, and an
adverse price move will effect them all the same, increasing the risk rather
than diversifying it.

Think about it this way -- You have a profitable system with logic that
gives you good entry signals.  If you wait until the trade shows a profit to
add to the position, you forfeit the profit from the funds that were not
used at the initial signal.  If the scale in portion gives better results
than the original signal, then the entry logic needs to be revised.



The same situation occurs at scale out.  A commonly described technique is
to take partial profits and reset the stop when some condition is met.  In
almost all of the tests I have run using this technique, the performance of
the system over a period of many trades is poorer than if the entire
position is held until the final exit signal.

Think about it this way --  You have a profitable system.  If you reduce
your position size when trades have some profit and, on average, the trades
go on to be more profitable, you forfeit some of the potential profit.  If
you reduce your position size when trades have some profit and, on average,
the trades go on to become losers, your exit logic needs to be revised.

Thanks,
Howard

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

Greetings --

h3po wrote:
Actully this gives idea for another interesting question: If one is able to
model *stockprices* by means of a simple model, could this (monte carlo)
simulated dataseries be valuable as additional data and be used as more OOS
data or will it not give any new information in the testing process?
(Asuming one is doing a WF test. If Yes, then this would be valuable for
end-of-day trading system development as the amount of data here is usually
limited). Has this been  discussed on this forum before?

---

In my opinion, synthetic or simulated prices have only one purpose.  That is
to investigate the superposition of signals -- such as combining a trend
with one or more cycles with noise -- at the most basic level in order to
understand superposition and decomposition.

Synthetic price series have no value in testing trading systems.  The
algorithm portion of a trading system -- the model -- is the logic that is
attempting to distinguish the signal from the noise.  Recognizing the signal
portion and issuing buy and sell orders is what makes the system
profitable.  Everything that the model does not specifically recognize and
use to issue buy and sell orders is noise -- even if it is clearly
identifiable as signal by some other method.  The ratio of signal to noise
in financial data is very low.  Identifying the signal component among all
the noise is very difficult.

If I can generate synthetic price series that are adequate for system
development, then I do not need to generate synthetic price series.  If I am
able to synthesize the signal well enough that I can develop a trading
system based on it and trade that system profitable when using the actual
data, then the logic I used to generate the synthetic price series describes
the signal.  I simply code that logic into a real trading system and use
it.

Synthetic price series based on random values will be random.  No trading
system can reliably and repeatedly extract profit from random series.  The
second law of thermodynamics applies here.

My recommendation to avoid synthetic price series extends to reordering
observed prices.  Reordering, for example reordering the daily prices of
IBM, in an attempt to give additional data for system testing destroys the
time related components of the original series.  If there was information in
trends, patterns, seasonality, autocorrelation, intermarket correlation,
etc, none of that exists in the newly created series.  What could be left?

In any event, these questions are easy to test.  Develop a trading system on
synthetic or reordered price series.  Then test on out-of-sample actual
data.  If it is profitable, then the value is in the logic used to generate
the synthetic data.



It is valuable to add random noise to actual price series in order to test
how well a trading system differentiates between noise and signal.  But that
is the answer to a different question.

---

Thanks,
Howard

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-29 Thread Howard B

Greetings --

Mike's comment is correct --

As for the value of Monte Carlo (MC) vs. Walk Forward Analysis (WFA). My
view is that the two are not mutually exclusive, but rather are
complimentary. Your out of sample (OOS) results from WFA are exactly the
right kind of inputs for MC analysis.

---

I would make the statement even stronger.  The input to a Monte Carlo
Analysis _must_ come from strictly out-of-sample results.  One way to get
these is from the OOS runs of walk forward analysis.  Another is the trades
observed after the system being tested has been finalized.  Either paper
trades or actual trades count, provided the system has not been modified
during that period.

Thanks,
Howard

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-29 Thread Howard B

Greetings all --

Several people have mentioned Dr Van Tharp's book, The Definitive Guide to
Position Sizing.

I like Dr Tharp's work, I have corresponded with him about position sizing,
and he was kind enough to mention me in his book.  By all means, buy his
book and study it -- there is a lot to like in it.

But also read the threads on Aussie Stock Forums that discuss trading system
design, position sizing, and specifically Dr Tharp's book.  Here is the url
to the home page of ASF:
http://www.aussiestockforums.com

I have a few reservations about his book.  Dr Tharp begins by defining
System Quality Number (SQN).
1.  Then he states that he has registered SQN as a service mark.  I
seriously object to people trying to claim ownership of a well known idea.
2.  SQN is essentially a t-statistic, as Dr. Tharp admits on one page early
in the text.
3.  He discusses the characteristics of trading systems that have SQNs with
values from 1 to 7.

It is valuable to calculate the t-statistic of a trading system as a
metric.  The book would have been valuable if it simply stopped there and
discussed some realistic examples.

But much of the remainder of the book went down an unrealistic path.

Just for fun, look at a table of t-statistics.  Look up the probability of a
one-tailed t-statistic being greater than +2.0 for a test that has 20 or
more observations -- it is about 2.9%.  This assumes that the observations
come a normal distribution, which may or may not be the case for a trading
system, but similar statistics can be computed that are not dependent on
assumption about the distribution and the results are not different enough
to alter my argument.  The point is that the probability that a trading
system has results that exceed random by enough that the t-statistic is
greater than 2.0 is about 2.9%.  He continues on to suggest that trading
system developers search for systems that have scores of 3, 4, even 7, and
that is the part that concerns me.

I include a discussion of using the t-statistic as a metric in my advanced
workshops.  As part of that discussion, I hand out simulated closed trades
that come from various distributions with various t-statistics.  If you
provide an experienced trader with a system that trades frequently, limits
losses on losing trades, and has a t-statistic of 3.0 (truly out-of-sample),
he or she can own Manhattan in a couple of years.  If anyone knows of a
system that has a t-statistic (SQN) greater than 3.0 on 20 or more
out-of-sample trades, please call me so we can form a partnership and retire
very rich very soon.

To recap -- buy and study Dr Tharp's book, but be aware of the unreasonable
assumptions he makes and the impossible task he sets.

Thanks for listening,
Howard

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-27 Thread Howard B

Greetings --

MK wrote:

I believe that is called Bootstrapping. In Monte Carlo you have to first
guess the distribution of trades - more assumptions.

---

This response is about assumptions about the distribution (not about
bootstrapping).

It is not necessary, or even advisable, or even good practice, to make
assumptions about the distribution of the trades.  It is important that the
trades from which the Monte Carlo Analysis draws its data are representative
of the trades the system under study produces.

There are techniques for determining the theoretical distribution that is
closest to the observed data, but you do not need to use them unless you are
planning a mathematical analysis.  Even then, many of the theoretical
distributions do not lend themselves to closed form solutions.

If you are going to use Monte Carlo Analysis, begin with as many closed
trades as you can generate (always out-of-sample only).  About 100 is a
reasonable minimum number.  There is probably no need to go beyond 1000.
Using a spreadsheet, sort the trades from worst loss to best gain, then form
a Cumulative Distribution (CDF) from the data.  It will look like a smooth
s shaped curve going from the lower left to the upper right.  The
horizontal axis has the worst trade as its leftmost value and the best as it
rightmost value.  The vertical axis has 0 as its lowermost value and 1.00 as
its uppermost value.  Look it statistics texts or do an Internet search for
more details.  [The CDF is the integral of the probability density function
(pdf).  The way the normal distribution graph is usually displayed is a
pdf.]  Once you have the CDF, you can select a new trade whenever you need
it by using a uniform random number generator.  Generate a random number,
look up the corresponding value of the trade using the CDF.  Essentially you
generate a uniform random number, go up the vertical axis to that value, go
across until you bump into the CDF, go down to the horizontal axis, and read
off the value for that trade.  Repeat as often as necessary.  This gives
sampling with replacement.  You can generate as many trades as you want --
not necessarily the same number you used to build the CDF.

If you prefer sampling without replacement, build a two dimensional array.
Column 1 has the trades.  Column 2 has values generated from a uniform
random number.  Sort the two columns on column 2.  Keep an index into the
table.  Every time you need a new trade, pick the one at the index, then
increment the index.  For a repeated run, use a fresh set of random numbers
in column 2, resort, and start selecting values from the beginning again.

Thanks for listening,
Howard

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-27 Thread Howard B

 this is a very specialized topic already. There not many books found on
 this, thatâ€™s why I spend so much time thinking about it: Nothing to be
 found means a lot to be made and understanding it in great detail might give
 me/you a larger edge. Especially a scientific approach that is not based on
 gut-feel, is missing. This is what I am currently working on: Portfolio
 Management in terms of Trading Systems. So to say, have a trading systems
 farm, probably 3-5 systems on one underlying to have a very smooth equity
 curve. This includes diversification across a) methodology (Mean Reversion,
 Trendfollowing, Breakouts,etc.) and b) timeframe (day-trading, swing
 trading, position trading). Well, I have found 2-3 books on this topic, all
 of them sort of scratching the surface, mentioning the topic, but thatâ€™s
 it. So how can systems be combined wisely and when should systems be
 switched off, reoptimized, etc. If anybody knows some books on this, Iâ€™d
 be very happy to read up on it so if someone knows an author, please post!
 So far, having many systems with different logics is the direction Iâ€™m
 going right now.
 
  Something thatâ€™s worth having a look inside is this one:
 
 
 
 
 http://www.amazon.de/Trading-Systems-Development-Portfolio-Optimisation/dp/1905641796
 
  again, itâ€™s scratching the topic only.
 
 
 
  The way I develop systems is always the same (If anyone doesnâ€™t agree
 with this approach, Iâ€™d like to know why)
 
 
 
  1 contract, share, etc.
 
  Commissions excluded
 
  3-5 (5max!) input parameters, such as indicators, maâ€™s: itâ€™ll result
 in a raw system with in-built exits and entries but no stops.
 
  Optimize the inputs for STABILITY, a profitable set of parameters with a
 stable surrounding region
 
  Plug-In Commissions
 
  Add filters
 
  Add stops/targets according to Sweeneyâ€™s MAE MFE
 
  Plug-in the position sizing algo, â€œthe compounderâ€
 
  Test out of sample
 
 
 
  Works fine for me.
 
 
 
  After this, I pretty much try to combine systems. Currently this is a bit
 troublesome in Ami, but I shall let you know when I have done enough
 research, unfortunately I believe itâ€™s gonna take me some months.
 
 
 
  What I believe to be very controversial is the topic â€œadding noiseâ€ to
 data. I donâ€™t think it reflects human emotion anymore. Same as dealing
 with a random sequence: a coin toss, e.g., can be displayed as a graph too,
 but itâ€™s still a normal distribution. So: When doing quant-trading and
 believing in it, you make one big assumption: Markets behave logically,
 sometimes. I found out that markets contain a large amount of noise, random
 behavior. But there are occasions when people get hit on the wrong spot and
 start buying or selling aggressively . Thatâ€™s the time when my systems
 jump in and make the kill - then Iâ€™m out. This behavior cannot be found in
 a random, probably noisy set of data.
 
  Monte Carlo pretty much changes the trade sequence only, right now I
 cannot really see why some people think itâ€™s so useful. It might give you
 a better estimate of a historic drawdown, but I achieve the same by
 multiplying my non-monte-carlo-drawdown with say 1.5â€¦
 
 
 
 
 
  Greetings from Germany,
 
 
 
  Matthias
 
 
 
 
 
  From: amibroker@yahoogroups.com amibroker%40yahoogroups.com [mailto:
 amibroker@yahoogroups.com amibroker%40yahoogroups.com] On Behalf Of
 Howard B
  Sent: Montag, 26. Juli 2010 16:33
  To: amibroker@yahoogroups.com amibroker%40yahoogroups.com
  Subject: Re: [amibroker] Trading Systems, Position Sizing and Monte Carlo
 Analysis
 
 
 
 
 
  Hi Sohamdas --
 
  In my opinion, this is definitely a topic that deserves discussion in the
 AmiBroker forum.
 
  What position sizing should be used during backtests?
 
  If you will be evaluating each trade for its characteristics -- entry
 efficiency, exit efficiency, and so forth, then each trade should be the
 same size. For stocks and ETFs that means the same dollar amount. For
 futures that means the same number of contracts.
 
  If you will be comparing equity growth over a period of time to other
 alternatives, then you will want position sizing and / or compounding to
 some degree. For example, if you want to compare the results of a trading
 system to buy and hold, you will want to take the same size position at the
 beginning if the test period for each alternative, then compare equity
 smoothness, growth, drawdown, etc.
 
  If you are planning to use aggressive position sizing, there are several
 things to consider.
  1. I cannot state it too often -- your system must have a positive
 expectancy measured on strictly out-of-sample results. You absolutely cannot
 use in-sample results to estimate the likely future performance of a trading
 system in any event. And if aggressive position sizing is based on in-sample
 results, you will go bankrupt.
  2. Traders should have a business plan in place. They should know when to
 quit -- either when they have enough

Re: [amibroker] Re: Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-27 Thread Howard B

Greetings --

MK wrote:

What is the benefit of scaling in and out vs. treating them as separate
systems? They basically are separate system.

-

I agree that entries made from secondary signals, or whatever you would call
the signals that trigger the scale in, are a separate system.  Test them
separately.

In my experience and research, neither scaling in nor scaling out often
improve a system.  This is based on an understanding that risk is fully
assessed before the initial portion of the trade is entered, and that the
full amount of the trade as determined by the risk calculation is not taken
until the final scale in has been triggered.  That is, if three scale ins
are perm itted, then the initial position must be taken with only 25% of the
amount permitted by the risk calculation.  If the alternative is to take a
full position at the first signal, and then add as the trade is in progress
through scale in trades, the risk guidelines will have been violated.

Thanks for listening,
Howard

Re: [amibroker] Trading Systems, Position Sizing and Monte Carlo Analysis

2010-07-26 Thread Howard B

Hi Sohamdas --

In my opinion, this is definitely a topic that deserves discussion in the
AmiBroker forum.

What position sizing should be used during backtests?

If you will be evaluating each trade for its characteristics -- entry
efficiency, exit efficiency, and so forth, then each trade should be the
same size.  For stocks and ETFs that means the same dollar amount.  For
futures that means the same number of contracts.

If you will be comparing equity growth over a period of time to other
alternatives, then you will want position sizing and / or compounding to
some degree.  For example, if you want to compare the results of a trading
system to buy and hold, you will want to take the same size position at the
beginning if the test period for each alternative, then compare equity
smoothness, growth, drawdown, etc.

If you are planning to use aggressive position sizing, there are several
things to consider.
1.  I cannot state it too often -- your system must have a positive
expectancy measured on strictly out-of-sample results.  You absolutely
cannot use in-sample results to estimate the likely future performance of a
trading system in any event.  And if aggressive position sizing is based on
in-sample results, you will go bankrupt.
2.  Traders should have a business plan in place.  They should know when to
quit -- either when they have enough that they no longer need to trade, or
when they have lost so much that they can no longer trade or realize that
they should pick another profession.
3.  Aggressive position sizing depends on having:
A.  Positive expectancy.
B.  Understanding of risk.  Both the risk that is acceptable for each trade
from the account, and the risk associated with the trading system.  Most
trading systems have higher risk per trade than the account risk allows, so
even taking a position that is everything you can afford to buy is
aggressive.
C.  The ability to use leverage.  Brokers allow use of margin, and some ETFs
have leverage.  By using these, it is possible to get 12 to 20 times
leverage trading stocks and ETFs in an ordinary brokerage account.
D.  Frequent trading, because that provides frequent compounding.  For most
trading systems, the final equity of an account is a multiple of the initial
equity that can be computed from:
terminal_equity = (1 + expectancy) ^ number_of_trades
where expectancy is the average percentage gain per trade.

There are two general schemes for aggressive position sizing.  As you dig
into the math, you will see that they are closely related.
The first is fixed fraction, popularized by Ralph Vince.
The second is fixed ratio, popularized by Ryan Jones.
Both men have written books and papers describing their methods, and you can
do an Internet search on each phrase and get a lot of information.

The essence of both methods is to increase position size when the system is
operating profitably.  In gambling terminology, you are betting the run of
the table.  When winning, increase; when losing, pull back.  Both of these
are betting schemes called anti-martingale.

Ralph Vince has also popularized the notion of optimal f -- that fixed
fraction that should be bet on each play to maximize the terminal equity.
The fraction of the account used for each play is determined by the largest
anticipated drawdown or trade loss.  He, and everyone else who is working
with real money, shows that the fraction bet on each play Must be less than
optimal f if the account is to remain solvent.  In fact, the fraction must
be much less than optimal f if the account is avoid large drawdowns.
Trading at optimal f essentially guarantees drawdowns in the 80 percent
range.

Ryan Jones essentially creates two sub-accounts.  One is the original stake,
say $100,000.  The other is the profits from trading.  Ryan waits until
there are some profits, then uses a high percentage of the profits for each
trade.

The two methods converge mathematically.  Some traders prefer to begin using
one method, then switch to the other as profits accumulate.

Before you consider using an aggressive position sizing scheme, and buying
all the stock you can afford is aggressive, please read my other comments in
both this forum and Aussie Stock Forums related to trading system
development and position sizing.  And read both Vince's and Ryan's books,
and other material you can find on the Internet.

The input to the simulators that model either fixed fraction or fixed ratio
need is a list of closed trade results.  These are individual trades, each
the same size.

Monte Carlo Analysis is used to rearrange the sequence of trades many
times.  The position sizing rules are applied to each sequence and the
equity curve computed and drawn.  Typically many sequences are used -- 1000
or more -- each of many trades -- 100 or more.  After all 1000 runs, all of
the equity curves are draw on a single chart and statistics computed that
will allow you to estimate the final equity and probability of both going
broke and

Re: [amibroker] Can I make buy signal based on date

2010-07-21 Thread Howard B

Hi Timur --

Buydate = DateNum()==1100716;

Buy = Buydate;

Thanks,
Howard


On Wed, Jul 21, 2010 at 4:08 AM, TimurLangit i4...@yahoo.com wrote:



 Help me please.

 How is afl code to create buy signal based on specific date? Like 'buy=1 if
 date==July 16,2010'?

 Thank you so much,

 Timur

Re: [amibroker] Position Size vs. Max. Sys % Drawdown

2010-07-21 Thread Howard B

Hi Ray --

The answer to your question is a couple of chapters in my next book,
Advanced AmiBroker.  While waiting for me to finish (no definite date for
release), you might read:
1.  any of Ralph Vince's books.  All are worth reading.  His most recent is
The Leverage Space Trading Model.
2.  Van Tharp's Trade Your Way to Financial Freedom.  Get the second
edition.  In the first edition his explanation of expectancy is wrong.  In
the second, it is more nearly correct, but still non-standard.  If you read
his Definitive Guide to Position Sizing, be aware that all of his examples
are artificial and unrealistic.  There is a thread on Aussie Stocks Forum
discussing that book with many of my comments.
3.  Nassim Taleb's Fooled by Randomness.
4.  David Aronson's Evidence Based Technical Analysis.

Several authors correctly point out the importance of position sizing in the
long-term profitability of a trading system.  The very important point to
keep in mind is that estimates of trading results used to determine the
position sizing must come from unbiased out-of-sample data.  Using in-sample
data will seriously underestimate the risk and lead to significant losses or
bankruptcy.

There are two components to the risk analysis, and they must be kept in
balance.
One is the risk a trader is willing to take in his or her account on any
single trade.  The recommended maximum risk is often in the 1 or 2% range.
If a person has a (notional) $100,000 trading account, 1% risk is $1000.
This means that the trader should hold the maximum loss of any single trade
to $1000.
The other is the risk associated with the trading system.  Measure the
(out-of-sample) mean and standard deviation of trades, or mean and maximum
adverse excursion of trades.  The standard deviation or MAE is (almost)
always significantly greater than the mean.  Per trade risk is, or at least
can be, determined as some multiple of standard deviation or MAE.  If a
trading system has an expectancy of 0.5% with a standard deviation of 1.5%
(not an untypical ratio), and maximum risk is computed as 2 times standard
deviation, the risk associated with this system is 3%.

Assume the system gives a signal to buy some issue -- say an ETF that trades
at $50.00 per share.  The maximum risk is 3% or $1.50 per share.  If the
entire $100,000 is allocated, 2000 shares could be purchased.  If the trade
goes bad, $3000 could be lost.  Since the $3000 risk associated with the
trading system is greater than the $1000 risk associated with the account,
the position size is too big.  The maximum position size that can be taken
is the ratio of the account risk to the trading system risk -- 1:3.  To keep
the two risk components in balance, the trader can take a position of 1/3 of
$100,000, or 666 shares.  Even without using leverage or margin, a trader
who takes a position with all of the funds in his or her account is
seriously at risk.  In this example, the trader allocating all the funds is
trading at 3 times leverage in a cash, un-leveraged, account

All trading systems that have a non-zero probability of unlimited loss on
any single trade have a non-zero probability of going bankrupt.  The
probability of that system going bankrupt increases to certainty as trading
continues.

Aggressive position sizing is necessary if a trader hopes to accumulate
serious wealth.  Ralph Vince developed / popularized optimal f, an
algorithm to calculate the fraction of a trading account that should be
bet on any single trade in order to maximize terminal wealth.  Systems
that trade at optimal f are expected to have drawdowns in the 80% range on
their way to the final wealth.  Trading at optimal f, or anywhere near it,
will create drawdowns that most traders cannot tolerate.  In fact, when
applying statistical methods to determine whether a trading system is broken
or not, most systems would appear to be broken before they recovered.  As
Vince points out, trading at a risk level higher than optimal f assures
bankruptcy.  And, as he points out in Leverage Space, if a multisystem is
being traded (either multiple systems or multiple tradables), if any one of
the systems is traded at a level higher than optimal f, the multisystem is
assured of bankruptcy.

The topic is a little more complex than explained here, but this should be a
reasonable start.

Thanks for listening,
Howard


On Tue, Jul 20, 2010 at 1:55 PM, raymondpconnolly 
raymondpconno...@gmail.com wrote:



 Hi Howard,

 When deciding on position size vs. Max. Sys % Drawdown is it advisable to
 use the minimum (minimum negative value) or the mean of Max. Sys % Drawdown?
 Does use of the control chart methodology you outline in the ATAA
 presentation alleviate the risk of a minimum Max. Sys % Drawdown scenario so
 that more aggressive position sizing can be taken based on mean rather than
 the minimum of Max. Sys % Drawdown ?

 My system seems be able to recover even when minimum Max. Sys % DD = -33%
 which based on OOS results occurs

Re: [amibroker] calculation of the t-stat of expectancy for the OOS periods of WF Simulation

2010-07-17 Thread Howard B

Hi Ray --

The t-test is used to test whether two samples come from the same
distribution.  The example you may be referring to asks whether the
expectancy of the out-of-sample results could have come from a random or
unprofitable system.  The null hypothesis (the one we want to say is very
unlikely) is the OOS results came from a random system.  Calculate the mean
and standard deviation of the metric you are using -- expectancy, in this
case -- for the OOS results.  Calculate the mean and standard deviation for
the other, random, distribution.  The random mean of expectancy is 0.0.  In
this case, you can assume that the random standard deviation is the same as
the OOS standard deviation.  Use the count of data points from the OOS
results, and assume the random results have the same count.  Compute the
t-statistic.  Look in the precalculated table to see how likely it is that
the OOS results came from the same distribution as the random results.

You could make a similar test to see whether the IS and OOS results came
from the same distribution.  That test would require using all of the
trades, IS and OOS, not just the OOS ones.  It might be interesting, but
that is not the test you probably want.

I hope this helps,
Thanks,
Howard

On Sat, Jul 17, 2010 at 7:09 PM, raymondpconnolly 
raymondpconno...@gmail.com wrote:



 Hi All,

 I'm looking for input on the calculation of the t-statistic of expectancy
 for the OOS periods of the walkforward simulation for the purposes of
 validation and control as put forward by Dr. Howard Bandy in his ATAA
 presention of October 2009. Many thanks to Howard for making available this
 excellent information.

 At first glance I thought this was a pretty straightforward matter ...
 until I got down to the nuts and bolts of the calculation. So let me explain
 what I've done and the questions I'm having:

 Assumptions:
 -t-statistic of expectancy (for the null hypothesis Ho: expectancy = 0 ) =
 ((mean)/(standard deviation))*sqrt(N-1) (1)
 - use data for each OOS period generated by Amibroker (no CBT generated
 data)

 Calculation Possibilities:
 1.0 calculate the mean and standard deviation using the expectancy values
 for each OOS period
 2.0 calculate the weighted mean and weighted standard deviation of
 expectancy where the weight for each OOS period is
 wt = (number of trades for the OOS period / total trades for all OOS
 periods) and where the weighted standard deviation
 is defined here:
 http://www.itl.nist.gov/div898/software/dataplot/refman2/ch2/weightsd.pdf
 3.0 N = number of OOS periods
 4.0 N = total trades for all OOS periods

 Observations and Questions:
 A) What is the correct combination of inputs from the possibilities above
 or from possibilities I have failed to recognize that yields the valid
 t-statistic ?
 B) comparing 1.0 and 2.0 above yeilds slightly different means in my case (
 where the standard deviation in the average number of trades per OOS period
 is small). 2.0 above also yeilds the same result as calculating expectancy
 by the alternative method:
 Expectancy ($) = (TotalProfit - TotalLoss) / NumberOfTrades = NetProfit /
 NumberOfTrades over all OOS periods so my suspicion is that 2.0 above is the
 way to go
 C) with respect to 3.0 and 4.0 above I'm at a loss. I can see using the
 number of OOS periods as there are that many observations of expectancy
 however I can also see using total trades for all OOS periods since the
 results of all trades ultimately generate the expetancies over all OOS
 periods.

 Thanks for your input.

 Ray

Re: [amibroker] Re: Random Backtesting

2010-07-15 Thread Howard B

Hi Joe --

The afl program that you posted has some problems other than the use of
random() to set PositionScore.  The code below was copied from the AmiBroker
5.30 User's Guide, then modified to use MTRandom() to set PositionScore.

The afl compiler is the final arbiter of what works and how.

Thanks,
Howard


//
//PositionScoreExample.afl
//
//Taken from the AmiBroker 5.30 User's Guide
//entry under Portfolio-level backtesting
//
//Modified to use a random PositionScore

Portfolio-level back testing

/*

Portfolio-level back testing

You can use new PositionScore variable to decide which trades should be
entered if there are more entry signals on different securities than maximum
allowable number of open positions or available funds. In such case
AmiBroker will use the absolute value of PositionScore variable to decide
which trades are preferred. See the code below. It implements simple MA
crossover system, but with additional flavour of preferring entering trades
on symbols that have low RSI value. If more buy signals occur than available
cash/max. positions then the stock with lower RSI will be preferred. You can
watch selection process if you backtest with Detailed log report mode
turned on.

*/


SetOption(InitialEquity, 2 );
SetTradeDelays(1,1,1,1);
RoundLotSize = 1;

posqty = Optimize(PosQty, 4, 1, 20, 1 );
SetOption(MaxOpenPositions, posqty);

// desired position size is 100% portfolio equity
// divided by PosQty positions

PositionSize = -100/posqty;

// The system is very simple...
// MA parameters could be optimized too...
p1 = 10;
p2 = 22;
// simple MA crossover
Short=Cross( MA(C,p1) , MA(C,p2) );
Buy=Cross( MA(C,p2) , MA(C,p1) );
// always in the market
Sell=Short;
Cover=Buy;

// now additional score
// that is used to rank equities
// when there are more ENTRY signals that available
// positions/cash
//PositionScore = 100-RSI(); // prefer stocks that have low RSI;
PositionScore = mtRandom()-0.50;


///

On Thu, Jul 15, 2010 at 12:29 AM, j0etr4der j0etr4...@yahoo.com wrote:



 Howard, the documentation states, The score (PositionScore) for all
 securities is calculated first. Then all scores are sorted according to
 absolute value of PositionScore. Then top N are choosen to be traded. Has
 that changed?

 Joe


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Howard B
 howardba...@... wrote:
 
  Hi Mikey --
 
  Positive values of PositionScore are used to rank issues to enter long
  positions. Negative values are used to rank issues to enter short
  positions. Try transforming the random numbers so that they run from
 -0.50
  to +0.50. That is, try:
 
  PositionScore = random() - 0.50;
 
  MTRandom() is better than random(), so you may want to use it instead.
 You
  can read about it in the AmiBroker reference guide.
 
  Thanks,
  Howard
 
 
 
  On Wed, Jul 14, 2010 at 7:11 AM, mikk12345 mikk12...@... wrote:
 
  
  
   Hi, This is what i want to do. I have 10 stocks in my ticker list. I
 want
   to randomly select 5 long and 5 short each worth 100,000 and hold for
 50
   days in the backtester.
  
   Ive tried the following:-
  
   Buy= 1;
   Sell= 0;
   Short=1;
   Cover=0;
   ApplyStop(stopTypeNBar,stopModeBars,50,1);
  
   PositionScore=Random();
  
   PositionSize=10;
  
   Ive put 1,000,000 in the equity box.
  
   This only returns long only.
  
   Best Regards,
  
   Mikey.

Re: [amibroker] Random Backtesting

2010-07-14 Thread Howard B

Hi Mikey --

Positive values of PositionScore are used to rank issues to enter long
positions.  Negative values are used to rank issues to enter short
positions.  Try transforming the random numbers so that they run from -0.50
to +0.50.  That is, try:

PositionScore = random() - 0.50;

MTRandom() is better than random(), so you may want to use it instead.  You
can read about it in the AmiBroker reference guide.

Thanks,
Howard



On Wed, Jul 14, 2010 at 7:11 AM, mikk12345 mikk12...@yahoo.com wrote:



 Hi, This is what i want to do. I have 10 stocks in my ticker list. I want
 to randomly select 5 long and 5 short each worth 100,000 and hold for 50
 days in the backtester.

 Ive tried the following:-

 Buy= 1;
 Sell= 0;
 Short=1;
 Cover=0;
 ApplyStop(stopTypeNBar,stopModeBars,50,1);

 PositionScore=Random();

 PositionSize=10;

 Ive put 1,000,000 in the equity box.

 This only returns long only.

 Best Regards,

 Mikey.

Re: [amibroker] Optimization issues

2010-07-11 Thread Howard B

Hi AJ --

Are the settings the same?

Are the issues being tested the same?

Are the individual values that give good results in the backtest included
within the ranges of the optimization variables?



Thanks,
Howard


On Sat, Jul 10, 2010 at 12:04 PM, AJ axjur...@gmail.com wrote:



 Good morning,
 I'm working on a small project where I have 4 variables. If I do a backtest
 with certain values I get the correct answer (in my particular case 60% gain
 10%drawdown). However, when I do an optimization with these variables (the
 values that I used to get the right answer are within the values that I
 usein the optimization) the only result I get is 0 calmar, and the gains are
 5%. Is there any reason why I would gt the right answer doing backtesting
 and completely diffrent results when optimizing? did I forget to set
 something? Thanks for your help,
 Augusto

Re: [amibroker] Re: Random Entries and Exits?

2010-07-07 Thread Howard B

Greetings --

This is the code being referred to from my book, Quantitative Trading
Systems.

Thanks,
Howard


//EnterAtRandom.afl
//
//Entry a position at the close of a random bar.
//
//A random entry for use as a benchmark.
//
//Expect this entry to mirror the buy and hold.
//It cannot be profitable taking long positions in
//a declining market, or taking short positions
//in a rising market.
//
//This code just waits a fixed number of bars to exit.
//It can be used to test any of the exit systems.
//
//For market orders, enter and exit Market On Close
//with no delay
SetTradeDelays(0,0,0,0);
BuyPrice = C;
SellPrice = C;

//Frequency is the number of entries per year.
Frequency = Param(Entries per Year,12,1,100,1);

//Repeatable is a switch.
//True (1): the sequence of random numbers will be repeated.
//False (0): each sequence is random.
Repeatable = Param(Repeatable,0,0,1,1);

//Seed is the number used to start the random sequence
//when repeatable sequences are desired.
Seed = Param(Seed,13331,1,9,1);

//Generate a fraction, uniformly distributed
//between 0.0 AND 0.9.
NextRandom = IIf(Repeatable,Random(Seed),Random());

Buy = IIf(NextRandomFrequency/252,1,0);

//The code for the exit being tested starts here.
//
//HoldBars is the number of bars to wait for exit.
HoldBars = Param(HoldBars,3,1,100,1);

Sell = BarsSince(Buy)=HoldBars;
//The code for the exit being tested ends here.

//Remove extra Buy and Sell signals.
Buy = ExRem(Buy,Sell);
Sell = ExRem(Sell,Buy);
//Figure 7.5 Enter at Random

/// end of afl 

On Wed, Jul 7, 2010 at 1:39 AM, reinsley reins...@yahoo.fr wrote:




 Hi,

 I don't know the last author.

 Best regards

 _SECTION_BEGIN(ct1);
 /*
Coin Toss model , it has nothing to do with Technical OR Fundamental
 Analysis

It has nothing to do with entry or exit rules.
Here One books profits and losses at a predefined percent.


This Model was Inspired by many authors
---
a)Victor Sperandeo   :- Methods of Wall Street Master
b)Nasim Taleb :-   Fooled by Randomness
c)Ed Thorp:-   #Beat the Market   #Beat the Dealer
d)William Poundstone   :-   Fortune's Formula

thanks for AB codes and concepts by
a)Dimitris Tsokakis
b)Herman van den Bergen
c)Anthony Faragasso
d)Graham Kavanagh
and countless good souls.

Hearty thanks to

Tomasz Janeczko for AB Platform.


Disclaimer : any error in code or concept here belongs to me.
This Amibroker Script is for educational purposes only. I cannot
 guarantee
 it's accuracy nor take any responsibility for your use of it.  Use at your
 own
 risk.


 */
 SetOption(InitialEquity, 5000);
 SetOption(CommissionMode, 2); //2 - $ per trade
 SetOption(CommissionAmount, 1.35); // commission in BPS per leg
 SetOption(MarginRequirement, 100);// no margin
 SetOption(UsePrevBarEquityForPosSizing, 1);
 SetOption(AllowSameBarExit,1);   // Sq off Daily
 SetOption(PriceBoundChecking,0);
 SetTradeDelays( 0, 0, 0, 0 );
 *RoundLotSize*=1; //Change this
 according to your needs [i did it for Nifty].

 posSizeX = Optimize(posSizeX, 1, 1, 10, 1);

 winloss = Optimize(winloss, 50, 10, 50, 2);// Assumed
 probability of winning, higher the better

 winPercent = Optimize(winPercent, 1, 0.25, 10, 0.25);// higher
 the better
 lossPercent = winPercent/2;   
 //Optimize(lossPercent,
 0.5, 0.25, 10, 0.25);// lower the better

 //iterations = Optimize(iterations,1,1,20,1);// to test
 randomness at different levels, not used in any equation.

 HitRatio = mtRandomA()  (winloss/100);  //hit ratio


 *PositionSize* = -(posSizeX);   //exposure of
 capital

 *Buy* = Day() != Ref(Day(), -1 );   // trade daily
 *Sell* = *Buy*; // Sq off
 Daily

 *Short*=*Cover*=0;

 *BuyPrice*=100;  // to make it generic from EURUSD to Hangseng to Bovespa

 //if you win you get winPercent, if you loose you loose lossPercent
 *SellPrice*=IIf(HitRatio,*BuyPrice**(1+(winPercent/100)),*BuyPrice**(1
 -(lossPercent/100))  );

 Plot(Equity(),Equity :  , 1,1);
 RequestTimedRefresh(1);

 SetChartBkGradientFill( ParamColor(BgTop, *colorWhite*),ParamColor(
 BgBottom,*colorViolet*));
 _SECTION_END();




 Le 07/07/2010 01:17, spacebass5000 a écrit :



 wow, how did I miss this... i have read and reread this book many times.
 thanks for the reminder!!!

 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com,
 notanaiqgenius notanaiqgen...@... notanaiqgen...@... wrote:
 
  If you happen to have Howard Bandy's book, Quantitative Trading Systems,
 you can see page 91-92 for an example

Re: [amibroker] Re: Random Entries and Exits?

2010-07-07 Thread Howard B

Greetings --

Tomasz has added the Mersene Twister algorithm for generation of random
numbers since my book Quantitative Trading Systems was printed.

See the AmiBroker help files for mtrandom for details.

mtrandom is a better algorithm than the one distributed with most language
packages, including C++ and Excel.

Also, download the pdf file from the link Addendum and Errata from this
page:
http://www.quantitativetradingsystems.com/book.html
which has an example of using mtrandom.

Thanks,
Howard


On Wed, Jul 7, 2010 at 4:34 AM, Howard B howardba...@gmail.com wrote:

 Greetings --

 This is the code being referred to from my book, Quantitative Trading
 Systems.

 Thanks,
 Howard


 //EnterAtRandom.afl
 //
 //Entry a position at the close of a random bar.
 //
 //A random entry for use as a benchmark.
 //
 //Expect this entry to mirror the buy and hold.
 //It cannot be profitable taking long positions in
 //a declining market, or taking short positions
 //in a rising market.
 //
 //This code just waits a fixed number of bars to exit.
 //It can be used to test any of the exit systems.
 //
 //For market orders, enter and exit Market On Close
 //with no delay

 SetTradeDelays(0,0,0,0);
 BuyPrice = C;
 SellPrice = C;

 //Frequency is the number of entries per year.
 Frequency = Param(Entries per Year,12,1,100,1);

 //Repeatable is a switch.
 //True (1): the sequence of random numbers will be repeated.
 //False (0): each sequence is random.
 Repeatable = Param(Repeatable,0,0,1,1);

 //Seed is the number used to start the random sequence
 //when repeatable sequences are desired.
 Seed = Param(Seed,13331,1,9,1);

 //Generate a fraction, uniformly distributed
 //between 0.0 AND 0.9.
 NextRandom = IIf(Repeatable,Random(Seed),Random());

 Buy = IIf(NextRandomFrequency/252,1,0);

 //The code for the exit being tested starts here.
 //
 //HoldBars is the number of bars to wait for exit.
 HoldBars = Param(HoldBars,3,1,100,1);

 Sell = BarsSince(Buy)=HoldBars;
 //The code for the exit being tested ends here.

 //Remove extra Buy and Sell signals.

 Buy = ExRem(Buy,Sell);
 Sell = ExRem(Sell,Buy);
 //Figure 7.5 Enter at Random

 /// end of afl 


 On Wed, Jul 7, 2010 at 1:39 AM, reinsley reins...@yahoo.fr wrote:




 Hi,

 I don't know the last author.

 Best regards

 _SECTION_BEGIN(ct1);
 /*
Coin Toss model , it has nothing to do with Technical OR Fundamental
 Analysis

It has nothing to do with entry or exit rules.
Here One books profits and losses at a predefined percent.


This Model was Inspired by many authors
---
a)Victor Sperandeo   :- Methods of Wall Street Master
b)Nasim Taleb :-   Fooled by Randomness
c)Ed Thorp:-   #Beat the Market   #Beat the Dealer
d)William Poundstone   :-   Fortune's Formula

thanks for AB codes and concepts by
a)Dimitris Tsokakis
b)Herman van den Bergen
c)Anthony Faragasso
d)Graham Kavanagh
and countless good souls.

Hearty thanks to

Tomasz Janeczko for AB Platform.


Disclaimer : any error in code or concept here belongs to me.
This Amibroker Script is for educational purposes only. I cannot
 guarantee
 it's accuracy nor take any responsibility for your use of it.  Use at your
 own
 risk.


 */
 SetOption(InitialEquity, 5000);
 SetOption(CommissionMode, 2); //2 - $ per trade
 SetOption(CommissionAmount, 1.35); // commission in BPS per leg
 SetOption(MarginRequirement, 100);// no margin
 SetOption(UsePrevBarEquityForPosSizing, 1);
 SetOption(AllowSameBarExit,1);   // Sq off
 Daily
 SetOption(PriceBoundChecking,0);
 SetTradeDelays( 0, 0, 0, 0 );
 *RoundLotSize*=1; //Change this
 according to your needs [i did it for Nifty].

 posSizeX = Optimize(posSizeX, 1, 1, 10, 1);

 winloss = Optimize(winloss, 50, 10, 50, 2);// Assumed
 probability of winning, higher the better

 winPercent = Optimize(winPercent, 1, 0.25, 10, 0.25);// higher
 the better
 lossPercent = winPercent/2;   
 //Optimize(lossPercent,
 0.5, 0.25, 10, 0.25);// lower the better

 //iterations = Optimize(iterations,1,1,20,1);// to test
 randomness at different levels, not used in any equation.

 HitRatio = mtRandomA()  (winloss/100);  //hit ratio


 *PositionSize* = -(posSizeX);   //exposure of
 capital

 *Buy* = Day() != Ref(Day(), -1 );   // trade
 daily
 *Sell* = *Buy*; // Sq off
 Daily

 *Short*=*Cover*=0;

 *BuyPrice*=100;  // to make it generic from EURUSD to Hangseng to Bovespa

 //if you win you get winPercent, if you loose you loose lossPercent
 *SellPrice*=IIf(HitRatio,*BuyPrice**(1

Re: [amibroker] How to reset the Backtester Results window?

2010-07-07 Thread Howard B

Hi Michael --

Right-click in the Results window.  This brings up a context menu.
Left-click Clear Results List.

Or did I mis-understand the question?

Thanks,
Howard

On Wed, Jul 7, 2010 at 1:27 PM, michaels_musings michaels_musi...@yahoo.com
 wrote:



 I did something to mess up the display of the backtest output. After trying
 a variety of things to reset it, including trying all three Result list
 shows: (Trade list, Detailed log, Summary) and completely uninstalling all
 copies of AB and reinstalling, it is still showing the Entry
 signals(score): oddity (below) instead of the trade list.

 Since it's propagated through a full uninstall, is there something in the
 registry that might be causing this? Someone know of a config setting I
 messed up?

 Thanks,
 Michael

 Backtester Results window

 Use to Show:
 Ticker,Net Profit,Net % Profit,Exposure %,CAR,RAR,...

 Now Shows:
 Date Information
 11/19/2009 5:12:00 AM
 Entry signals(score):
 Exit signals:
 0 Open Positions: , Equity: 10, Cash: 10
 11/19/2009 5:13:00 AM
 Entry signals(score):
 Exit signals:
 0 Open Positions: , Equity: 10, Cash: 10
 11/19/2009 5:14:00 AM
 Entry signals(score):
 Exit signals:
 0 Open Positions: , Equity: 10, Cash: 10

 continues on for 500,000 rows

Re: [amibroker] Intrady Time interval problem

2010-07-04 Thread Howard B

Greetings --

On the Tools menu  Preferences  Intraday tab

You can control whether the time associated with a bar is the start of the
bar or the end of the bar.

Thanks,
Howard

On Thu, Jul 1, 2010 at 1:04 AM, edakad1 edak...@yahoo.com wrote:



 Hi,

 I am facing a problem with intraday charts. Suppose the day session ends at
 3.30 pm. I am using 5 min charts. Then the last bar should end at 3.30 pm.
 Instead I see the last bar close at 3:29:59 and there is another bar with a
 few ticks. It shows time 3.30.

 Plese help to resolve this issue.

 Thanks and regards

Re: [amibroker] Re: Percentile Rank of P/S Ratio

2010-06-18 Thread Howard B

Hi Abbie --

I wrote a paper, Use of Fundamental Data in Active Investing, that
discusses some the issues that arise when using fundamental data.
You can download it for free from this site:
http://www.blueowlpress.com/activities.html

Thanks,
Howard

On Fri, Jun 18, 2010 at 1:44 AM, Paolo pcavat...@gmail.com wrote:



 Abbie,

 as far as I know yahoo doesn't provide historical fundamental data which
 you need for proper historical backtesting.

 Regarding the for cycle it looks fine.


 Paolo

 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com,
 sendfreetest sendfreet...@... wrote:
 
 
  Thank you for getting back to me with some information on
  how to tackle the Percentile Rank of P/S Ratio code.
 
  I was thinking of using Yahoo! historical data but
  your comments make me think their data may not be
  accurate enough? Any thoughts on good data?
 
  I will think over and try to implement your idea
  about using PositionScore.
 
  What I was working on is a for a cycle to calculate
  the percentile. (Most of the code I found in back
  posts of this group. )
 
  Right now the code seems to work in an exploration.
  I've pasted it below. (As I am new to coding it may
  not be the prettiest.)
 
  I'll continue to work on the code to try to make it
  useful for actually testing the idea.
 
  ---
 
  Var = C / GetFnData(SalesPerShare);
 
 
  /* PERCENTILE RANK */
  /* set watchlist , and indicator twice */
 
  //Adapted From Count tickers in watchlist function - Anthony Faragasso
 
  listNum=1; // DEFINE WATCHLIST
 
  function CountTickersInWatchList( Listnum )
  {
  // retrive comma-separated list of symbols in watch list
  list = GetCategorySymbols( categoryWatchlist, listnum );
  for( i = 0; ( sym = StrExtract( list, i ) ) != ; i++ )
  {
  if( i == 0 ) i = 0;
  else i = i ;
  }
  return i;
  }
  // Adapted from Ranking Ticker List - Nenapacwanfr
  list = GetCategorySymbols( categoryWatchlist, listnum );
  for( i = 0; ( sym = StrExtract( list, i ) ) != ; i++ )
  {
  SetForeign(sym);
  VarSet(MyInd+i, (C / GetFnData(SalesPerShare))); //INDICATOR - DEFINE
  Rank =1;
 
  for( j = 0; ( item = StrExtract( list, j ) ) != ; j++ )
  {
  SetForeign(item);
  VarSet(ThisInd+j, (C / GetFnData(SalesPerShare))); //INDICATOR -
 DEFINE
  Rank=Rank + IIf( VarGet(MyInd+i)  VarGet(ThisInd+j),1,0);
  RestorePriceArrays();
  }
  VarSet(Rank+i,Rank);
  }
  Count=0;
  BreakLoop = False;
  for( k = 0; NOT(BreakLoop) ( ticker = StrExtract( list, k ) ) != ;
 k++ )
  {
  if(Name()==ticker)
  Breakloop=True;
  else
  Count=Count+1;
  }
  Rank = VarGet(Rank+ Count);
  PercentRank = (1 - rank/CountTickersInWatchList(Listnum))*100;
 
  Filter = PercentRank  17 AND PercentRank  42;
 
  AddColumn(RANK,Rank);
  AddColumn(CountTickersInWatchList(Listnum),Total);
  AddColumn(Close,Close);
  AddColumn(PercentRank,Percentile Rank);
  AddColumn(C / GetFnData(SalesPerShare), P/S Ratio);
 
 
  ---
 
  Thanks again for your help,
 
  Abbie
 
 
  --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Paolo
 pcavatore@ wrote:
  
   First of all you need P/S data (or Sales per share and calculate the
 ratio) with enough historical range and clean data. A rare combination
 especially when talking about fundamentals.
  
   Secondly you can use Rotational backtesting with PositionScore being
 equal to the inverse of the P/S (the higher the better).
  
   Otherwise I would recommend using a for cycle calculating the
 percentile for each stock in your universe and use it to both rank your
 stocks for proper buy/sell rules and plot also the ranking as an indicator.
 Be careful 'cause for large number of stocks (500) it tends to be very
 slow.
  
   I hope it helps,
  
   Paolo
  
   --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com,
 sendfreetest sendfreetest@ wrote:
   
Hi,
   
I really need coding assistance.
   
I am attempting to code into AmiBroker Charles Kirkpatrick's
trading methodology for testing.
   
Kirkpatrick creates a percentile rank of stocks' price-to-sales
ratio and uses it as a filter. He purchases only stocks
with a percentile rank between the 17th and 42nd level.
   
When a stocks price-to-sales percentile rank goes above 67
he sells.
   
With AmiBroker's PERCENTILE function it seems to evaluate
only one market's array to determine the percentile.
   
Is their an AmiBroker funtion to evaluate say 100+
stocks P/S Ratio and give each a percentile rank.
   
Thanks in advance for any help.
   
Abbie

Re: [amibroker] Re: How to set the amount in ApplyStop(stopTypeTrailing, stopModePoint,amount,1,1,1)

2010-06-18 Thread Howard B

Hi Jolly --

Something like this:?

TrailingStopPercentage = 2.5; // percentage points
ApplyStop(StopTypeTrailing,StopModePercent,TrailingStopPercentage,1);

TrailingStopPercentage is a variable that you control completely -- you name
it, you assign it a value, the value can be an expression including
calculations using other variables.  It can be used with Optimize.

See the user's manual entry for ApplyStop.

Thanks,
Howard

On Fri, Jun 18, 2010 at 6:34 AM, jollypollyanna jollypolly...@yahoo.comwrote:




 --- In amibroker@yahoogroups.com, Joe Bolton j0etr4...@... wrote:
 
  Greetings,
 
 Following the work-around for the trailing stop is the code I started for
 the trailing Applystop, but of course the code for the amount is wrong.
 What is there is the closest I could come to getting the right amount and
 it's very wrong.
 
  Thank you for any help you can provide.
 
 
  Best regards,
 
  Joe

 ==

 Reading through this thread showed the issue had not been resolved in the
 thread . That being the *setting of amount outside the ApplyStop
 function*.

 I am thinking that amount cannot be initiated outside the ApplyStop
 function. If this is the case could someone please confirm this? If amount
 can be initiated outside of ApplyStop function, could a simple example be
 provided please.
 Thank you.

Re: [amibroker] read individual values

2010-06-16 Thread Howard B

Hi Erich --

I still do not understand.  Can you post (or send me privately) an afl that
illustrates the condition?  Include the settings that you are using.

Thanks,
Howard

On Tue, Jun 15, 2010 at 10:44 AM, Erich Matos Rodrigues 
erich...@interjato.com.br wrote:



  Hi Prof. Howard,







 In this way they never agree and the difference is great. However, when I
 plot my trade system in range of bars chart, I look indicator value and it
 is same value that I choose in condition to buy or sell. Then, I think that
 in range of bar chart and in backtest with range of bar, I have same value.
 Perfect.



 But How to know value of CCI (14), for example,  every time TS (in range of
 bar) to buy or sell? If I use explorer or function below, CCi (14) or other
 indicator, give results in range of DATE.  Course will be different…. Then I
 think that need change in function.



 This value in example, I have in my range of bar chart,   when I add CCI
 indicator in range of bar chart, but I´d like a spreadsheet with this
 indicator value when buy or sell.



 Thank you for your response,





 Erich







 Hi Erich --

 I am still a little confused.

 You are permitted to have different values for parameters for an indicator
 that is displayed in a chart and for an indicator that it being tested in a
 backtest. They Can agree and they May agree, but they are not Required to
 agree. Is this the confusion?

 Thanks,
 Howard

 On Tue, Jun 15, 2010 at 4:37 AM, erichrod1 
 erich...@interjato.com.brerichrod%40interjato.com.br
 wrote:

 
 
  Hi,
 
  In fact it is little confuse and I didn't explain very well.
 
  I use only range of BARS, in chart and backtest. To improve my TS I´d
 like
  to know, for example, CCI value (in range of bars) when my TS buy or sell
  (in range of bars). This CCI value will be same that I see in range of
 bars
  chart.
 
  If I use explorer to do this, I choose range in bars (like I define in
  preferences, intraday, in Custom N-tick chart settings), but indicator
  results (value) are in range of dates... Amibroker buy or sell based in
 CCI
  value create from range of bars, right, but give me value from range of
  dates...
 
  After that, I tried to use backtest and a custom function to generate
  indicator value, but didn't work, trade system works very well in range
 of
  bars, but indicator value is always in range of date, different of value
 in
  chart, that is in range of bars also.
 
  I don't know if is possible.
 
  Thank you for your response,
 
 
  Erich Rodrigues
 
  --- In amibroker@yahoogroups.com amibroker%40yahoogroups.comamibroker%
 40yahoogroups.com, Howard B
  howardba...@... wrote:
  
   Hi Erich --
  
   Thanks for the kind words about my books.
  
   I am afraid I do not understand your question.
  
   Usually the backtest is made over a range of dates or bars. Can you
  express
   the rangebar range you want tested in dates or bars?
  
   You can export the results of a backtest or explore, including
 indicator
   values and Buy or Sell signals using AA's File  Explore menus. Does
 that
   do what you need?
  
   Thanks,
   Howard

Re: [amibroker] Re: doubt with stops

2010-06-16 Thread Howard B

Hi Gonzaga --

Check the users guide for ApplyStop.

Your afl can have more than one exit, and it is common for a trading system
to include both exits described by logic (Sell = xxx;) and by ApplyStop --
perhaps more than one of each.  If you have several ApplyStop statements
(profit target, trailing stop, maximum holding period), AmiBroker keeps them
straight.  If you have several logic exits, assign each its own variable
name (SellCond1 = xyz;   SellCond2 = wxy;), then combine them into one Sell
statement (Sell = SellCond1 OR SellCond2;).

The precedence and order in which stops and logic exits are applied and
depends on settings.  If you have very complicated exits, you might want to
write looping code to handle them.  This gives you complete control,
assurance that you understand how they are being applied, and the ability to
debug them using Plot and AddColumn statements.

Thanks,
Howard


On Tue, Jun 15, 2010 at 4:42 PM, Gonzaga gonzag...@yahoo.es wrote:



 I've just realized that my problem my be related with the sell signal.
 For I use a profit stop, but also a sell condition; and both are different.

 So, how does combine sell conditions with stops?

 Could I have problems if I mix both?


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Gonzaga
 gonzag...@... wrote:
 
  Hi. I have an idiot doubt. Sometimes AFL gets me crazy :-)
 
  I have applied a very simple stop profit:
 
  //---
  SetTradeDelays(1,1,1,1); // buy or sell after signal
  //SetOption(ActivateStopsImmediately,False); -- I don't know the use
 of this
  stoplevelp=1;
  ApplyStop(stopTypeProfit, stopModePercent, stoplevelp, 1);//1: exit
 intraday on same bar
  //---
 
  So I though the profit stop sell the stock even in the day of buy, just
 1% over buy price.
 
  But I see, it sells 1,3% over buyprice. or more..
  In settings I have established a commission table of 0.005 per share, so
 I suppose that's not related.
 
  And, also, how does it work the axtivatestopsimmedately?
 
  Is there any good help file to study all these..?
  Thanks a lot..

Re: [amibroker] read individual values

2010-06-14 Thread Howard B

Hi Erich --

Thanks for the kind words about my books.

I am afraid I do not understand your question.

Usually the backtest is made over a range of dates or bars.  Can you express
the rangebar range  you want tested in dates or bars?

You can export the results of a backtest or explore, including indicator
values and Buy or Sell signals using AA's File  Explore menus.  Does that
do what you need?

Thanks,
Howard


On Mon, Jun 14, 2010 at 8:17 AM, erichrod1 erich...@interjato.com.brwrote:



 Hi Dr. Howard,

 I have your book and love it. It help me improve my backtests.

 Talking about how to read individual values in the portfolio-phase of the
 backtest, I have a question: How to use this custom objective function to
 take value IF TIMEFRAME IS RANGEBAR? I use this function in my tradesystem,
 that buy and sell in range bar, but function results are timed...

 Is it possible change this function to do this? Basically, I need export
 indicator value when my backtest says to buy or to sell.

 TIA,

 Erich Rodrigues

 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Howard B
 howardba...@... wrote:
 
  Hi RZ --
 
  My book Quantitative Trading Systems discusses objective functions in
  general, the importance of choosing one early in the system design
 process,
  the metrics built-in to AmiBroker, the process of designing
 multi-component
  custom objective functions, the custom backtest code necessary to
 implement
  custom objective functions, and the use of objective functions in
  optimization and walk forward testing.
 
  The objective function incorporates the subjective aspects of trading
  systems. Proper choice of the objective function guarantees that trading
  systems that are highly ranked by the search process and chosen by the
 walk
  forward process will be comfortable to trade. It would be difficult to
  overemphasize the importance that proper selection of objective function
  plays in trading system development.
 
  http://www.quantitativetradingsystems.com/
 
  Thanks,
  Howard

Re: [amibroker] Self Adaptive Bollinger

2010-06-13 Thread Howard B

Hi G --

If it does not violate any copyright or non-disclosure, post the
TradeStation code and someone will translate it and post the equivalent
AmiBroker code.  If that would be in violation, then this is not the
appropriate forum.

Thanks,
Howard


On Sun, Jun 13, 2010 at 11:36 AM, googool123123 bfall...@yahoo.com wrote:



 I have Trade Station code for self adaptive Bollinger Bands.

 I was wondering if any one would be interested in translating it into
 Amibroker Language

 Thanks in Advance

Re: [amibroker] Divergence Indicator

2010-06-12 Thread Howard B

Hi Vishvesh --

Please reply to an open thread rather than beginning a new one.

In order to program an indicator for divergence, it must first be described
as a formula.  Description of the visual appearance is not adequate.  Can
you describe divergence as a formula?

Thanks,
Howard


On Fri, Jun 11, 2010 at 11:43 AM, Vishvesh vishvesh.chau...@gmail.comwrote:


  [Attachment(s) #129281c24215e039_TopText from Vishvesh included below]

 When i say divergence it mean prices(low/high/close) have made newer
 reading but the RSI or MACD failed to do so. I have attached chart for ur
 perusal.

 Regards
 Vishvesh



 On Fri, Jun 11, 2010 at 10:12 PM, Howard B howardba...@gmail.com wrote:



 Hi Vishvesh --

 If you can define divergence in an unambiguous way, it can be programmed.
 What is your definition?

 Thanks,
 Howard


 On Fri, Jun 11, 2010 at 10:11 AM, Vishvesh vishvesh.chau...@gmail.comwrote:



 Hello Friends,

 I was wondering is there AFL available which highlights the price bar
 when there is divergence in momentum (RSI,MACD) etc.

 Any help highly appreciated,

 Thanks  Regards
 Vishvesh

Re: [amibroker] Limiting Drawdown in a Backtest

2010-06-12 Thread Howard B

Hi RZ --

My book Quantitative Trading Systems discusses objective functions in
general, the importance of choosing one early in the system design process,
the metrics built-in to AmiBroker, the process of designing multi-component
custom objective functions, the custom backtest code necessary to implement
custom objective functions, and the use of objective functions in
optimization and walk forward testing.

The objective function incorporates the subjective aspects of trading
systems.  Proper choice of the objective function guarantees that trading
systems that are highly ranked by the search process and chosen by the walk
forward process will be comfortable to trade.  It would be difficult to
overemphasize the importance that proper selection of objective function
plays in trading system development.

http://www.quantitativetradingsystems.com/

Thanks,
Howard

On Fri, Jun 11, 2010 at 9:00 PM, i cs ics4...@yahoo.com.au wrote:



 Thank you Howard, re_Rowland and Mike

 Was trying to avoid the CBT but it looks like I gotta dive in.


 BTW Howard, which of your books would best cover this level of experience?

 Thanks for your help again

 RZ


 --
 *From:* Howard B howardba...@gmail.com
 *To:* amibroker@yahoogroups.com
 *Sent:* Sat, 12 June, 2010 2:39:54 AM
 *Subject:* Re: [amibroker] Limiting Drawdown in a Backtest



 Hi RZ --

 You can do this by creating a custom objective function.

 It could as simple as assigning the custom function a value of 0 whenever
 the parameter set results show maximum system drawdown exceeds 30%.

 Try this:

  / / / //

 //LimitDrawdown. afl
 //
 //Use CustomBacktester to limit
 //drawdown in trading systems.
 //
 //Howard Bandy
 //June 2010
 //
 OptimizerSetEngine( cmae );

 //The custom backtest code
 //to penalize results that do not
 //meet your specifications

 DesiredDDLimit = 30;

 SetCustomBacktestPr oc(  );

 if ( Status( action ) == actionPortfolio )
 {
 bo = GetBacktesterObject ();
 bo.backtest( );
 st = bo.getperformancest ats( 0 );

 MaxSysDD = abs( st.getvalue( MaxSystemDrawdownPe rcent ) );
 NetProfPct = st.getvalue( NetProfitPercent );
 ObFn = ( MaxSysDD  DesiredDDLimit ) * NetProfPct;
 bo.addcustommetric( FilteredNetProfit, ObFn );
 }


 //The Buy and Sell signals

 MA1Len = Optimize( MA1Len, 5, 1, 100, 1 );

 MA2Len = Optimize( MA2Len, 20, 1, 100, 1 );

 MA1 = EMA( C, MA1Len );

 MA2 = EMA( C, MA2Len );

 Buy = Cross( MA1, MA2 );

 Sell = Cross( MA2, MA1 );



 e = Equity();

 Plot( e, E, colorGreen, styleLine );
  // end  

  / / / /

 Thanks,
 Howard


 On Fri, Jun 11, 2010 at 6:37 AM, ics4mer ics4...@yahoo. 
 com.auics4...@yahoo.com.au
  wrote:



 Hi,

 Doing a bit of backtesting lately, and I was wondering how to limit the
 drawdown while using another metric for the objective function.

 Lets say that I want to optimize for Net Profit, but I won't accept
 parameter sets that result in a drawdown of more than 30%

 I know it can be done (AB does anything).

 Could somebody give me a pointer please?
 I can't find explicit references to it in the archive, and I don't think
 I'm ready for some of the more intricate parts of the CBT.

 TIA

 RZ

Re: [amibroker] Divergence Indicator

2010-06-11 Thread Howard B

Hi Vishvesh --

If you can define divergence in an unambiguous way, it can be programmed.
What is your definition?

Thanks,
Howard

On Fri, Jun 11, 2010 at 10:11 AM, Vishvesh vishvesh.chau...@gmail.comwrote:



 Hello Friends,

 I was wondering is there AFL available which highlights the price bar when
 there is divergence in momentum (RSI,MACD) etc.

 Any help highly appreciated,

 Thanks  Regards
 Vishvesh

Re: [amibroker] Simple Code Difficulty

2010-06-08 Thread Howard B

Here is the Buy part with plot statements to help visualize what is
happening

-

//Buy20DayHigh.afl
//
//Buy when ema(c,5)  ema(c,20)
//and issue reached new 20 day high
//
//Howard Bandy
//June 2010
//
Cond1 = EMA( C, 5 )  EMA( C, 20 );
Cond2 = H  Ref( HHV( H, 20 ), -1 );
Buy = Cond1  Cond2;

Sell = 0;

shapes = Buy * shapeUpArrow + Sell * shapeDownArrow;
shapecolor = IIf( Buy, colorGreen, colorRed );
PlotShapes( shapes, shapecolor );

Plot ( Close, C, colorBlack, styleLine );
Plot ( Cond1, Cond1, colorGreen, styleLine | styleOwnScale, 0, 1.5 );
Plot( Cond2, Cond2, colorBlue, styleLine | styleOwnScale, 0, 2 );



Thanks,
Howard


On Fri, Jun 4, 2010 at 1:40 PM, abramway abram...@yahoo.com wrote:



 I'd just like to be able to get signals for

 Buy: When EMA(5)EMA(20) AND Stock @ Fresh 20-Day Highs.
 Exit: When EMA(20)EMA(5) OR Stock @ Fresh 10-Day Lows (haven't
 incorporated the second part yet)

 This is what my code reads now and its giving me the wrong signals. It
 gives me BUY signals before a new 20-Day high.

 Buy = EMA( Close , 5 )  EMA( Close , 20 )
 AND High  Ref( HHV( High , 20 ) , -20 );

 Sell = EMA( Close , 20 )  EMA( Close , 5 );

 Short = 0;

 Cover = 0;

 Thanks for the help

Re: [amibroker] Request For An AFL To Scan Open=High or Open=Low In Any Time Frame

2010-06-08 Thread Howard B

See if this helps

--

//OpensAtHigh.afl
//
//Identify that the opening price
//on the first bar of the day
//is the high of that bar.
//
//Howard Bandy
//June 2010
//

dn = Day();
FirstBar = dn != Ref( dn, -1 );
OpenAtHigh = FirstBar  ( O == H );


Plot( C, C, colorBlack, styleCandle );
shapes = shapeUpArrow * OpenAtHigh;
shapecolor = OpenAtHigh * colorGreen;
PlotShapes( shapes, shapecolor );

--

Thanks,
Howard


On Wed, Jun 2, 2010 at 11:51 PM, FLOYD LEWIS fj...@yahoo.com wrote:



 Dear Seniors

 Greetings to you all

 Am new to Amibroker and equally new to the field of AFL writing.

 Does anyone on this group by any chance have an afl which can be run on any
 time frame including real time to explore and scan stocks which meet the
 condition of open=high or open=low for the first bar only.

 Definition of first bar - It could be any time frame in RT i.e. 3 min or
 hourly etc as per the selection but it should be the bar of the opening bell
 which should be scanned for the above condition and not the subsequent bars.

 Would sincerely appreciate if any one from this group could share the same.

 I hope i have been able to explain the requirement clearly.

 Thanks in advance.
 God Bless
 Regards

Re: [amibroker] Deleting database

2010-06-08 Thread Howard B

Hi K --

Each database is a subdirectory to C:\Program Files\AmiBroker.  For example,
the small database that installs when AmiBroker is first installed is named
Data.  It is in C:\Program Files\AmiBroker\Data.  Use Windows Explorer and
you can see the folders where each ticker is stored according to its first
character -- IBM is in \Data\I, 3M is in \Data\M, etc.

If you really want to delete the entire database, use Windows Explorer to
delete the subdirectory that holds the database you want to remove.

If you plan to reconstruct the database with the same name but new data, and
you want to be able to use the layouts and watchlists, copy the \Layout and
\Watchlist subdirectories to a safe place before deleting the database.

Thanks,
Howard


On Thu, Jun 3, 2010 at 3:57 PM, homosapeins85 ktha...@gmail.com wrote:



 How do I delete a database? Alternatively, is there a way to delete all the
 symbols from a database other than doing it manually for each symbol?

 Thanks,
 K

Re: [amibroker] Volatility bands

2010-06-06 Thread Howard B

Hi Mithal, and all --

Not to discourage your efforts, but no one will be able to develop, test,
and validate a trading system, then trade it with confidence, without having
skills in both programming and statistics.

Please, take the time to learn AmiBroker's afl.

Thanks,
Howard

On Sat, Jun 5, 2010 at 12:46 PM, Infinity Home Loans infynh...@yahoo.comwrote:



 Hi Mubashar,

 I am not at all comfortable with writing a formula, as I do not have any
 coding background.

 I have tried using BB but they do not serve the purpose as I had expected.

 Can you please try and provide some support in building this formula as per
 the tradestation code mentioned earlier?

 Regards
 Mithil

 --- On *Sat, 5/6/10, Mubashar Virk mvir...@gmail.com* wrote:


 From: Mubashar Virk mvir...@gmail.com

 Subject: Re: [amibroker] Volatility bands
 To: amibroker@yahoogroups.com
 Date: Saturday, 5 June, 2010, 11:15 AM



 It should be easy provided you have custom atr formula.

 On 6/4/2010 12:47 PM, infynhome wrote:



 I am referring to the book Technical analysis for the trading
 professional by constance brown.

 She had mentioned about Volatility band formula written in tradestation
 language.

 I need this formula to be converted into afl language.

 Can somebody help me with this.

 Input: Coefdwn[2.1] ,Coefup[2. 3];
 Plot1[Average[ [RSI[Close, 14]],6]]+ [Coefup*[ Average[TrueRang eCustom[[
 RSI[Close, 14]],[RSI[ close,14] ]],14]]], Plot1];
 Plot2[Average[ [RSI[Close, 14]],6]]- [Coefup*[ Average[TrueRang eCustom[[
 RSI[Close, 14]],[RSI[ close,14] ]],14]]], Plot2];

 Plot3[[RSI[Close, 14]], plot 3];

 IF CheckAlert Then Begin
 IF Plot1 Crosses Above Plot2 or Plot1 Crosses Below Plot2
 or Plot1 Crosses Above Plot3 or Plot1 Crosses Below Plot3
 or Plot2 Crosses Above Plot3 or Plot2 Crosses Below Plot3
 Then Alert = TRUE;
 End;

 thanks
 Mithil

Re: [amibroker] Loop for MA calculation

2010-06-03 Thread Howard B

As requested --

//CalculateMovingAverageLooping.afl
//
//Calculate a simple moving average using looping code
//
//Howard Bandy
//June 2010
//

//In Formula Editor, click Apply Indicator.
//Note that all three moving averages are the same


//Plot the price series
Plot( C, C, colorBlack, styleLine );

MALen = 10;

//Assume the afl statement to be replicated is:

MAAFL = MA( C, MALen );

//Plot the builtin simple moving average
Plot( MAAFL, MAAFL, colorBlue, styleLine );

//A couple of ways to compute the simple moving
//average using looping code follow.

//This example is coded for clarity, rather than for efficiency
//
//Initialize the bars for which there are insufficient
//bars for a full moving average to be the closing price
//for each of those bars.

for ( i = 0; i  MALen; i++ )
{
MALoop1[i] = C[i];
}

//TempSum holds the sum of the values being averaged
//TempSum is a scalar.
for ( i = MALen; i  BarCount; i++ )
{
TempSum = 0;

for ( j = 0; j  MALen; j++ )
{
TempSum = TempSum + C[i-j];
}

MALoop1[i] = TempSum / MALen;
}

Plot( MALoop1, MALoop1, colorGreen, styleLine );


//This example is a little more efficient
//

for ( i = 0; i  MALen; i++ )
{
MALoop2[i] = C[i];
}

TempSum = 0;

for ( j = 0; j  MALen; j++ )
{
TempSum = TempSum + C[MALen-1-j];
}

for ( i = MALen; i  BarCount; i++ )
{
//Drop off the oldest value and add the newest
TempSum = TempSum - C[i-MALen] + C[i];
MALoop2[i] = TempSum / MALen;
}

Plot( MALoop2, MALoop2, colorRed, styleLine );


/// end

Re: [amibroker] State of Advanced Amibroker

2010-05-16 Thread Howard B

Hi Frank --

Thanks for your interest in Advanced AmiBroker.

I am still working on writing and editing the book.  Time to work on it is
in competition with my own trading and several other of my projects.  I am
making progress on the book, but have no definite release date.  I will post
notices when it goes to the printer and is available for pre-publication
orders.

Thanks,
Howard

On Sat, May 15, 2010 at 7:43 AM, frankphd_us frankphd...@yahoo.com wrote:



 Will the book be out in the foreseeable future - or is it postponed again?
 I couldn`t find any new announcements on the publisher`s site.

 Thanks

 frankphd_us

Re: [amibroker] Re: SP GICS clasifications

2010-04-23 Thread Howard B

Hi Ara --

According to information from Norgate Premium Data, they will will have GICS
data real soon now.  Richard Dale, who often posts here, is their contact.

Thanks,
Howard


On Fri, Apr 23, 2010 at 12:54 AM, Paolo pcavat...@gmail.com wrote:




 http://www.amibroker.org/userkb/2008/03/25/ascii-import-standard-and-poors-global-1200/

 Let me know if it helps,

 Paolo


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Ara
 Kaloustian a...@... wrote:
 
  Does anyone know where I can find a list of stocks with their GICS codes
 so I can import them into AB?
 
  Thanks
 
  Ara

Re: [amibroker] Re: Taking 2 Trades in 1 Bar

2010-04-17 Thread Howard B

Greetings --

You may need to use bars with finer resolution.

Any assumptions of the order of any prices other than Open and Close are
unreliable.  Conservative trading system design assumes that prices happen
in the order that is least favorable to the trader.

For any single bar, you know four prices -- Open, High, Low, Close -- and
the time at which two of those took place -- Open, Close.  You have no
information about the timing, or even the relative timing, of the High, the
Low, or any other price.

The possibilities for a single unambiguous trade within a single bar are:
Buy (enter) at the Open, Sell (exit) at a limit or stop price, provided that
price is within the Low to High range.
Buy at Open, Sell at the Close.
Buy at a limit or stop price within the Low to High range, Sell at the
Close.

Note that you cannot Buy at one intrabar price and Sell at another intrabar
price, because you have no information about which happened first.

The possibilities for two unambiguous trades within a single bar are:
Buy at the Open, reverse to Short at a price within the Low to High Range,
Cover at the Close.


Thanks,
Howard


On Fri, Apr 16, 2010 at 3:18 AM, ang_60 ima_c...@yahoo.it wrote:





 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Pmxgs
 pm...@... wrote:
 
  Hello,
 
  I also had the same issue and I found that if you choose
  SetBacktestMode( backtestRegularRawMulti ) AB allows you to do what you
 want, without using Custom backtester.
  By doing this, as the previous user said you can never be sure of the
 order of the trades, but if you are willing to accept this assumption it'll
 do what you want.
 

 Hello pmxgs,

 I tried your suggestion on the type of system commented here (that is,
 acting on the breakout of the previous bar's range), around this code(these
 are just trials, so I don't care about the chronological order of the
 trades):

 
 SetBacktestMode(backtestRegularRawMulti);
 Buy = Cover= H  Ref(H, -1);
 Sell= Short = L  Ref(L, -1);
 BuyPrice=CoverPrice = Max(O, Ref(H, -1));
 SellPrice = ShortPrice = Min (O, Ref(L, -1));
 

 but I still get just 1 trade per bar.

 Are you using some other settings?

 If it is not too much pain, could you give just a rough example of an AFL
 code taking 2 trades per bar?

 Greetings,

 Angelo.

Re: [amibroker] How to Change the Sort Order

2010-04-11 Thread Howard B

Hi Alan --

In order to sort them into the order you want them in, you will need to
assign each a number that will be used as the sort key.  I assume you can
define the key and do that.

Assign that key to a variable.
Use the AddColumn to list the variable.
Run the Exploration.
Click on the header for the column that has your sortkey.

Thanks,
Howard




On Sat, Apr 10, 2010 at 1:45 AM, Alan a...@thenorthams.us wrote:



 In the first column of the results of my exploration is a list of Ticker
 Symbols. These symbols are listed alphabetically in ascending order. I
 would like to have these tickers listed in a certain specified order
 which would not be an alphabetical ascending or descending order but in
 an order of my own choosing. Is there a way to do this?

Re: [amibroker] Question about PositinScore in backtesting

2010-04-06 Thread Howard B

Hi Tony --

There is a trading system methodology known as rotational trading where a
watchlist is evaluated every bar and those issues that have the highest
PositionScore are selected.  To enable rotational trading, the afl needs
this statement:
EnableRotationalTrading();

When using rotational trading, Buy, Sell, Short, and Cover statements are
ignored.  Only PositionScore matters.  Trading takes place at the Open or
Close depending on the settings of BuyPrice, SellPrice, ShortPrice, and
CoverPrice.  Trading takes place with the delay between signal and execution
according to the settings of SetTradeDelays(n,n,n,n);

Rotational trading was very popular in the 1990s for switching among mutual
funds, and continues to have some following.  Rotational trading is a subset
of general trading, and is limited by the inability to use Buy and Sell
signal, Stops, Timed exits, and other techniques.

Review the AmiBroker User's Guide for details.

-

If the afl does not have rotational trading enabled, then PositionScore is
used to break ties only when new entries are being considered.  In the
situation you describe, the ten issues selected on day one from more than
ten candidates are those that have the highest PositionScore on that day.
On day two there are three available slots.  The three that will be chosen
are the three that have the highest PositionScore from among those
candidates that have Buy signals that day.  Issues that have already been
purchased do not have their PositionScore evaluated and will not be sold
simply because of a low PositionScore.  After they have been sold due to a
Sell signal, or execution of a stop, they then rejoin the pool of candidates
and will have their PositionScore calculated in anticipation of competing
for entry at the next Buy signal.

Thanks,
Howard





On Sat, Apr 3, 2010 at 8:27 PM, Tony M pqj...@yahoo.com wrote:



 My formula generates a lot of signals every day, and my max position is set
 to 10. So I use a PositionScore to rank the stocks that have a buy signal
 and buy at the open the next day. On the first day of backtesting, the top
 ten stocks ranked by PositionScore are picked and bought. This is exactly
 what I want it to happen. At the closing of first day, three stocks meet the
 exit criteria and get sold. On the second day backtesting, there are total
 seven old positions and three new stocks are bought. However the three new
 stocks get selected are not the highest ranking stocks in the PositionScore
 list. The three stocks are ranked No. 3rd, 18th and 25th in the
 PositionScore list. This is very puzzling to me. My formula is very simple
 and is like this:
 BUY = condition1 and condition2 and condition3;
 PosQty = 10;
 SetOption(MaxOpenPositions, PosQty );
 PositionSize = -100/PosQty;
 SetTradeDelays(1,0,0,0);
 BuyPrice = Open;
 Sell = sell_condition;
 SellPrice = Close;
 PositionScore=100-RSI(14);

 My question is why PositionScore works perfectly on the first day of
 backtesting, but not the days thereafter? I have been working on this
 question for quite a few days, but I just couldn't find an answer. Can
 anyone here help me on this?

 Thanks,
 Tony

Re: [amibroker] Re: Exponential Moving Average

2010-03-27 Thread Howard B

Greetings --

There may be two questions in this thread.

Question 1 -- What is the relationship between the smoothing constant and
the lag.

Given a Simple Moving Average with a lookback of SMALength (say, 10 bars),
what is the smoothing constant for an Exponential Moving Average that
performs approximately the same smoothing as the SMA?

The name given to the smoothing constant is alpha.

An EMA is defined by a recursive relationship, here based on Closing price,
C:
E[0] = C[0];
E[i] = alpha*C[i] + (1-alpha)*E[i-1];

i is the index that starts at the oldest data point and continues through
the most recent data point.

alpha is the smoothing constant that defines the EMA.
alpha is approximately 2/(SMALength+1)

A SMA using a lookback of 10 bars is roughly equivalent to an EMA using an
alpha of 2/(10+1) = 2/11 = 0.1818.

When a formula using an Exponential Moving Average is specified, it is
defined by its lookback period, EMALength.  For example:
EMALength = 10;
Sm = EMA(C,EMALength);
Within the compiler, ELALength is used to compute alpha, and the series Sm
is computed using the recursive relationship.

You can compute your own EMA using this afl code:
EMALength = 10;
alpha = 2/(EMALength+1);
CustomEMA[0] = C[0];
for (i=1; iBarCount; i++)
{
  CustomEMA[i] = alpha*C[i] + (1-alpha)*CustomEMA[i-1];
}
Plot(CustomEMA,CustomEMA,colorblue,styleline);



Question 2 -- When optimizing a system based on the crossover of two moving
averages, how to keep the length of the second period greater than the
length of the first.

FastLength = Optimize(FastLength,5,1,10,1);
SlowIncrement = Optimize(SlowIncrement,1,1,20,1);
SlowLength = FastLength + SlowIncrement;
FastMA = EMA(C,FastLength);
SlowMA = EMA(C,SlowLength);
Buy = Cross(FastMA,SlowMA);
Sell = Cross(SlowMA,FastMA);

--

Thanks,
Howard

On Fri, Mar 26, 2010 at 9:31 PM, Mike sfclimb...@yahoo.com wrote:



 Keith,

 He was asking for the smoothing factor, not the period. The smoothing
 factor is a value between 0 and 1, as described in the document pointed to
 by his link.

 It is a good question. After some brief searching, I could not find the
 answer. Though, I suppose you could fire up a chart using a short period and
 do the math, as per the comments of DEMA.

 http://www.amibroker.com/guide/afl/afl_view.php?id=42

 Mike


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Keith
 McCombs kmcco...@... wrote:
 
  It's settable by you. Right click in the chart that is displaying ema.
  Select edit formula, and find EMA(n) -- n is the period and you can
  set it to whatever you want.
 
  On 3/26/2010 18:48, mpdx58 wrote:
  
   Does anyone know what the smoothing factor is that Amibroker uses when
   calculating the exponential moving average?
  
  
 http://en.wikipedia.org/wiki/Exponential_smoothing#The_exponential_moving_average
   
 http://en.wikipedia.org/wiki/Exponential_smoothing#The_exponential_moving_average

Re: [amibroker] rotational trading buying time

2010-03-20 Thread Howard B

Hi Donald --

Does something like this work for you?

TradingDay = 15;
dn = Day();
newMonth = (dn=TradingDay)  (Ref(dn,-1)TradingDay);

Thanks,
Howard


On Sat, Mar 20, 2010 at 7:36 AM, donald_brown_48367 
donald_brown_48...@yahoo.com wrote:



 I have a rotational system that rotates at the beginnning of each month
 using this code;

 m = Month();
 newMonth = m != Ref( m, -1);
 PositionScore = IIf(PositionScore  0, 0, PositionScore); // Long only
 PositionScore = IIf(newMonth, PositionScore, scoreNoRotate);

 Is there an easy way of changing the rotation to any particular day of the
 month, say the 15th? (Or nearest trading day if this falls on the weekend).

Re: [amibroker] AmiBroker as an intraday trading platform

2010-03-09 Thread Howard B

Yes

On Mon, Mar 8, 2010 at 3:38 AM, victorio91306 vabb...@socal.rr.com wrote:



 Is the AmiBroker software a viable platform to day trade The ESMINI.

Re: [amibroker] Amibroker Perspective

2010-03-08 Thread Howard B

Hi Jchi --

Herman is right.

When profits look too good to be true, the first thing I check for is a
future leak -- using data in the trading system code that is available to
the backtester that would not be available to the trader.  Depending on the
code used, checking this might be as easy as using the Code Check and
Profile tool available from the Formula Editor's Tools menu.  Included files
(#include) are checked, but dlls are not checked by this tool.

There are several ways you can adjust for commission and slippage.  One is
using the Backtester Settings tab from the Automatic Analysis dialog box.
Another is add a slippage adjustment that you define to the BuyPrice and
SellPrice variables.

Thanks,
Howard

On Mon, Mar 8, 2010 at 8:31 AM, jchi18 inbox@gmail.com wrote:



 What has been people's experience with Amibroker in terms of accuracy?
 Initially, I found it very difficult to design a system that makes a decent
 profit. However, I have recently designed a system that would have netted me
 (through the backtester years 2000-2010) almost $1 billion dollars had I not
 put an positionsize restrictions on it. Now that seems obviously
 unrealistic. Now I made my positionsize a little more realistic (a scale of
 10-30% of current equity). And I am still coming up with $29 million over
 the past ten years. I am wondering what I may be doing incorrectly.

 Here are some questions:

 1. How accurate is amibroker with an EOD trend following design where I buy
 or short on the open after the trigger?
 2. What sort of things do I have to be aware of to make the system more
 realistic?
 3. How do you account for slippage in your EOD trading system?
 4. On a different note, what is the most profit that you were able to
 backtest?

 I just want to understand and be realistic with what I have designed. I am
 a little naive because I have already started dreaming about my new found
 wealth. I would appreciate any thoughts.

Re: [amibroker] Adding Colums to Backtest

2010-03-08 Thread Howard B

Hi Jchi --

I recommend that you do this in two steps.

One -- write the trading system, including Buy and Sell statements.  Use the
Automatic Analysis Backtest button to run the backtest and get trading
results.
Two -- add AddColumn statements to the code.  (See the reference materials
for examples)  Use the Explore button to get a listing in the Results window
with a column for everything you want listed and a row for every bar you
want that list to include.

Thanks,
Howard


On Sat, Feb 27, 2010 at 8:16 AM, jchi18 inbox@gmail.com wrote:



 Is there a way to add columns to the backtest results that show:

 1. The MACD at the time of trigger
 2. RSI at the time of trigger

 Thanks.

Re: [amibroker] Few Newbie Questions

2010-03-08 Thread Howard B

Hi Jchi --

High quality, low cost end-of-day data is available from several vendors.
Two I recommend (I subscribe to them both) are:
Quotes Plus -- http://qp2.com/joomla/index.php
Norgate Premium Data -- http://www.premiumdata.net/

Read the reference materials related to BuyPrice, SellPrice, and
TradeDelays.  They work in combination.
The pair of statements
BuyPrice = Open;
SetTradeDelays(0,x,x,x);
uses the price of the open of the bar being processed.

The pair of statements
BuyPrice = Open;
SetTradeDelays(1,x,x,x);
uses the price of the open of the bar following the bar being processed.

Thanks,
Howard




On Fri, Feb 26, 2010 at 3:58 PM, jchi18 inbox@gmail.com wrote:



 1. When is the earliest I can download EOD data from Yahoo? i.e., when will
 today's data be available for download?

 2. In the trades tab when performing automatic analysis, what does buy
 price mean if I set it to open? i.e., I perform a scan of my trading system
 and it says that a buy signal was found for 2/26/10. When it is set to buy
 at the open price, will amibroker get the open price of 2/26 or 2/27?

 3. What is the best but reasonably priced way of getting accurate EOD data
 into amibroker? I understand that Yahoo is inaccurate.

 Feel free to contact me at inbox@gmail.com inbox.chi%40gmail.com.
 THanks.

Re: [amibroker] Re: Walk Forward IS/OOS Period Optimization?

Hi SpaceBass --

I would like to read your paper.  Let us know how to get a copy.

Thanks,
Howard


On Mon, Feb 22, 2010 at 9:39 PM, spacebass5000 spacebass5...@yahoo.comwrote:



 Awesome, lots to ponder. Thanks a lot for the input everyone!

 I'm actually writing a paper that looks at various Quant Trading models and
 need to hash out my OOS/IS periods. You all have been a big help.


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com,
 spacebass5000 spacebass5...@... wrote:
 
  I was wondering if there was a way to optimize the In-Sample and Step
 time periods within AB. If not, can someone point me in the direction of a
 good resource on this topic?

Re: [amibroker] AddToComposite - using PremiumData

Hi Pat --

I regularly use AddToComposite with data from Norgate Premium Data without
any difficulty.  I suspect the problem is with the way ATC is being used,
not with the data.  Post the entire code so we can look at it.  (Or make up
a small program that fails to work correctly to illustrate your problem)

Thanks,
Howard


On Sun, Feb 21, 2010 at 7:35 PM, patsgreatdeals patsgreatde...@yahoo.comwrote:



 Using AddToComposite along with PremiumData.net I get no results once the
 composite is built using Scan. I get the ~Composite calculated, but there is
 nothing in it. I used AddToComp a while back with no issues and used
 AMIQuote at that time.

 Now that I switched to premium data I am not getting any results.

 I am not saving the PremiumData in the native AMIBroker database, is this
 why I am not getting any results?

 Has anyone experienced this issue?

Re: [amibroker] Scale in and exit signal on the same bar

Hi Pedro --

For any single bar, you know the values of four data points -- Open, High,
Low, and Close.  You know the time of two data points -- Open and Close.
There is no way to determine the sequence in which any prices other than
Open and Close occur.

You can, within one bar do any of these things:
1.  Buy the Open.  Then exit at a stop, at a limit, or at the Close.
2.  Buy sometime within the bar at a stop or at a limit.  Then sell at the
Close.
3.  Buy at the Open.  Then sell at the Close.

These have no scaling.  You are certain of enough information to make one
entry and one exit. You cannot rely on assumptions of prices that would both
buy and sell within a single bar -- that is after the open but before the
close.

It sounds like you need bars with finer resolution, then take the actions
you can be sure of without ambiguity in each bar.

Thanks,
Howard

On Tue, Feb 23, 2010 at 6:55 AM, Pmxgs pm...@vizzavi.pt wrote:



 Hello,

 in the system I'm coding I have several scale in signals and it's working
 the way I expect, except at some bars which have both a scale in signal and
 the sell signal (profit target).
 I ran the backtest in the detailed log mode and from my understanding AB
 first closes the position (based on the sell signal) and then opens a new
 one (based on that scale in signal), which is not what I want.
 I want to scale in first and after that close the position.

 I should note that I'm not trading at the close nor at the open. I scale in
 at several prices inside each bar, and the scale in price in below the sell
 price (in the case of a long position).

 Is it possible to change the default behaviour of AB, so that it scales in
 first and only then, exit the position? (in these cases where we have both
 signals on the same bar)

 thanks,
 Pedro

Re: [amibroker] Re: Mutual funds are outperforming ETFs big time, but they sh...

Hi Kevin --

There are several difference between ETNs and ETFs.

ETNs make up less than 1% of the ETF/ETN products.  A few of them have
reasonable volume, but most do not.  There can be tracking error (and
arbitrage opportunity for major players) with ETNs.  ETNs are not equities,
or baskets of equities, they are swaps.  As you point out, there is greater
counterparty risk with ETNs.

Thanks,
Howard

On Tue, Feb 23, 2010 at 10:57 AM, kevin...@aol.com wrote:



  How about ETN's, shouldn't ETN's not have this issue?. of course they
 have other issues, like whether the underlying Bank stays in business

 Kevin

Re: [amibroker] Walk Forward IS/OOS Period Optimization?

2010-02-20 Thread Howard B

Hi Joris, and all --

If the typical time a trade is held, and trading frequency is fairly
constant, both of which I recommend, then constant time periods will have
similar number of trades.

Some of us carry on a lot of our lives using time to schedule -- how many
dollars per year we earn, how many calories per day we eat, how many miles
per week we run.  So it does make sense to compare trading system
performance over equal length time periods -- how many dollars per year the
system makes.

The logistics are considerably more complicated to run walk forward testing
using number of trades rather than number of bars.

Be careful when evaluating out-of-sample results.  If the system is modified
based on out-of-sample results, the data used to produce those OOS results
is no longer out-of-sample, but is now part of the in-sample data.
Subsequent test results, even when the previously OOS data was not
explicitly included in the in-sample data mining, no longer give an unbiased
estimate of future performance.  Rather, the estimate of future performance
is overestimated and of risk is underestimated.

Thanks for listening,
Howard

Re: [amibroker] Re: Pyramiding / ScaleIn Questions

Greetings all --

An earlier post in this thread:
I also read Connor's book and found many of his strategies worked, probably
because they were simple. I did not find that scaling in, as is being
discussed in this thread, worked better than just buying a full position.
Anyone else come up with the same conclusions?
Larry

I have done quite a lot of research into scaling in, scaling out, placing
stops, moving stops to break even, and so forth.  I have found very few
situations where scaling in or scaling out was a better implementation
technique than full entry and full exit.

Thanks,
Howard

Re: [amibroker] Walk Forward IS/OOS Period Optimization?

Hi SpaceBass --

The only way to determine the correct length for the in-sample period is by
running experiments.  The length needs to be long enough for the model to
synchronize with the data and learn to recognize the signal.  But not so
long that the signal has changed significantly, making it hard to identify.
And not so short that there is not enough signal to learn, resulting in a
system that has synced to the noise.

In a few words -- the length of the in-sample period should be as short as
is practical and effective.

Once the length of the in-sample period has been determined, the length of
the out-of-sample period is easy.  It is the length of time that the system
remains profitable.

There is no general relationship between the anything and the length of the
in-sample period.  There is no relationship between the length of the
in-sample period and the length of the out-of-sample period.

There is some controversy in the modeling and simulation field about whether
the in-sample length is a legitimate variable in an optimization.  While it
can be dangerous to system validity to run an optimization on in-sample
length, some trials using different lengths are necessary.

Thanks for listening,
Howard


On Thu, Feb 18, 2010 at 7:23 PM, spacebass5000 spacebass5...@yahoo.comwrote:



 I was wondering if there was a way to optimize the In-Sample and Step time
 periods within AB. If not, can someone point me in the direction of a good
 resource on this topic?

Re: [amibroker] Walk Forward IS/OOS Period Optimization?

Hi Herman, and all --

The issue determining the length of the in-sample period is not the number
of trades, but the number of data points it takes for the model (afl) and
the data (OHLC) to become synchronized.  Imagine a model that is looking for
cyclic behavior in the data.  If the cycle completes in one month it takes
many fewer data points to sync than if the cycle completes in one year.

As for out-of-sample, the statistics can be run on any number of data
points.  The data points can be daily equity changes or closed trades --
whatever time granularity the trader will be making position adjustments.
Providing the relationship between the mean of the metric being studied, its
standard deviation, and the number of data points is reasonable, control
charts can used to indicate when the system is operating well and when it is
beginning to fail.

Assume the metric is expectancy.  If the mean expectancy is positive and the
ratio of the mean to the standard deviation is high (high would be a ratio
of 0.5, extremely high would be 1.0, normal would be 0.2), it takes fewer
than 20 trades to have very high confidence that the system is performing
well.  Since systems go out of sync -- either periodically or forever --
once there are enough OOS data points to give the confidence that the
developer requires, there is no benefit in waiting for further
confirmation.  Trade now, while the system is in sync.

Thanks for listening,
Howard


On Fri, Feb 19, 2010 at 6:51 AM, Herman vandenBergen psy...@magma.cawrote:



 Hello Howard,

 I cannot help but wonder why the OOS periods in optimizations aren't based
 on the number of trades, instead of some arbitrary time period. Could it be
 that this is just another example of how TA methodology is based on
 convenience, i.e., a reluctance to redesign tools and/or a reluctance to
 change :-)

 I am no statistician, but I would guess that if one wanted some specific
 OOS performance the first thing to do would be to figure out how many trades
 would be needed to do the calculations. If 100 trades gives you the
 confidence why would you run a OOS period that gives you 200 trades?

 Best regards,
 herman


 Friday, February 19, 2010, 8:18:45 AM, you wrote:





 Hi SpaceBass --

 The only way to determine the correct length for the in-sample period is by
 running experiments.  The length needs to be long enough for the model to
 synchronize with the data and learn to recognize the signal.  But not so
 long that the signal has changed significantly, making it hard to identify.
  And not so short that there is not enough signal to learn, resulting in a
 system that has synced to the noise.

 In a few words -- the length of the in-sample period should be as short as
 is practical and effective.

 Once the length of the in-sample period has been determined, the length of
 the out-of-sample period is easy.  It is the length of time that the system
 remains profitable.

 There is no general relationship between the anything and the length of the
 in-sample period.  There is no relationship between the length of the
 in-sample period and the length of the out-of-sample period.

 There is some controversy in the modeling and simulation field about
 whether the in-sample length is a legitimate variable in an optimization.
  While it can be dangerous to system validity to run an optimization on
 in-sample length, some trials using different lengths are necessary.

 Thanks for listening,
 Howard


 On Thu, Feb 18, 2010 at 7:23 PM, spacebass5000 spacebass5...@yahoo.com
  wrote:

 I was wondering if there was a way to optimize the In-Sample and Step time
 periods within AB. If not, can someone point me in the direction of a good
 resource on this topic?

Re: [amibroker] Re: Pyramiding / ScaleIn Questions

Hi Richard, and all --

I categorize exits in several general ways.

1.  The exit from the logic of the trading system.  This is usually the best
exit.

Exits that can be good include:
2.  Exit caused by a trailing stop that follows the price movement in the
direction of the trade -- the stop portion of the Parabolic Stop And Reverse
or the Chandelier stop as examples.  This is often an excellent exit for a
trend-following system.  It requires a few bars in the trade for the exit
level to catch up to the price, so this is best used with trades that hold
at least a few bars.  It works with daily or intra-day bars.
3.  Exit caused by a profit target being met.  This is often an excellent
exit for a system that holds only a few bars and is never looking for the
big win.  In order to be a valid exit, the profit target must be the exit a
substantial portion of the exits -- cherry picking is not allowed.
4.  Exit caused by a time limit running down.  This could be a stop strictly
based on time, as in holding a maximum number of bars.  Or it could be an
inactivity exit, as in exit when there has not been favorable price movement
within some number of bars.

The poorest exit:
5.  Exit because a maximum loss limit has been reached.  This is the exit of
last resort.  It is easy to test yourself.  Beginning with any trading
system, add a maximum loss stop using the ApplyStop function.  Set the limit
so far away that it is never hit and run a backtest.  Gradually bring the
stop closer.  Note the system performance as the loss point is closer to the
entry and the maximum loss stop is hit a greater percentage of the time.
System performance usually degrades.

Moving an exit to break even falls into this last category.  It is always an
attractive sounding exit -- giving you a free trade, and all -- but seldom
performs as well as expected.  If you plan to use it, test it for yourself.

By all means --- if you have a trading system where these generalities do
not hold, do not be bound by them.  I am not suggesting that one solution
fits all situations, or that these exits will perform in the order I have
listed for your system.  But do perform your own tests so that you
understand how your exits work with your systems.

Thanks for listening,
Howard

Re: [amibroker] Re: Can Amibroker handle small decimal numbers?

2010-02-17 Thread Howard B

Hi Zozu --

K-ratio is one of the metrics that is computed by AmiBroker and reported.
If you are creating a custom objective function, you can include a term
related to k-ratio.  In that term, you can refer to the values that are
typically reported.  For example, require that k-ratio be = 0.05.

All values stored in AmiBroker variables are stored as floating point
numbers and are in IEEE single precision (32 bit) floating point notation.
The form of the number is .xxxeyy, where xxx is about 7 decimal
digits of significance and yy is an exponent.  Together they allow numbers
of either sign with absolute values between about 10^-38 and 10^38 to be
stored.

Some of the difficulty using very small numbers comes when they are added to
larger numbers.  Only the 7 most significant digits are stored -- all the
information in digits from 8 on is lost.

Thanks,
Howard

On Wed, Feb 17, 2010 at 3:00 PM, zozuzoza zoz...@gmail.com wrote:



 Hi Howard

 I meant kratio for example that have typical values of 0.01. If I use it as
 a fitness function, do I need to multiply by 100?

 Or if I optimise a parameter that have 0.01 step?

 Br,
 Zozu


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Howard B
 howardba...@... wrote:
 
  Hi Zozu --
 
  How small are the numbers and how are you using them?
 
  Multiplying by 100 or 1000 is often useful when working with indicator
  values so that they scale reasonably when plotted.
 
  Multiplying by 100 or 1000 may be useful when combining a variable with a
  small value with a variable with a large value so that the significant
  digits of the smaller value are retained.
 
  Adding a constant and / or multiplying by a constant may be necessary to
  create a number that has a positive integer part if that value will be
 added
  to a composite and stored in the volume field.
 
  Thanks,
  Howard
 
 
  On Tue, Feb 16, 2010 at 5:23 AM, droskill drosk...@... wrote:
 
  
  
   I haven't read the book - but for me it hasn't been a problem - I
 routinely
   use small numbers as you describe them
  
  
   --- In amibroker@yahoogroups.com amibroker%40yahoogroups.comamibroker%
 40yahoogroups.com, zozuzoza

   zozuka@ wrote:
   
Hi,
I am reading in Howard's book that 'Amibroker handles the negative
 values
   without and problem, but the small numbers loose all the significance
 beyond
   the decimal point.' Furthermore Howard recommend to multiply the
 numbers
   below 1 by 100 or 1000.
   
Is it really the case?

Re: [amibroker] Re: Can Amibroker handle small decimal numbers?

2010-02-16 Thread Howard B

Hi Zozu --

How small are the numbers and how are you using them?

Multiplying by 100 or 1000 is often useful when working with indicator
values so that they scale reasonably when plotted.

Multiplying by 100 or 1000 may be useful when combining a variable with a
small value with a variable with a large value so that the significant
digits of the smaller value are retained.

Adding a constant and / or multiplying by a constant may be necessary to
create a number that has a positive integer part if that value will be added
to a composite and stored in the volume field.

Thanks,
Howard


On Tue, Feb 16, 2010 at 5:23 AM, droskill drosk...@yahoo.com wrote:



 I haven't read the book - but for me it hasn't been a problem - I routinely
 use small numbers as you describe them


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, zozuzoza
 zoz...@... wrote:
 
  Hi,
  I am reading in Howard's book that 'Amibroker handles the negative values
 without and problem, but the small numbers loose all the significance beyond
 the decimal point.' Furthermore Howard recommend to multiply the numbers
 below 1 by 100 or 1000.
 
  Is it really the case?

Re: [amibroker] Re: statistivcs definition

2010-02-12 Thread Howard B

Industrial Strength

Is that a good thing or a bad thing?

Thanks,
Howard


On Tue, Feb 9, 2010 at 9:00 PM, reefbreak_sd reefbreak...@yahoo.com wrote:



 I am no expert in backtesting, you need to address those questions to
 Howard Bandy on this forum or buy his EXCELLENT book Quantitative Trading
 Systems where he creates and backtests a variety of trading sysstems along
 with commentary. I understand Howard is coming out with an advanced version
 very soon. Also he has available a recording of a lecture he gave in
 Australia detailing the development and backtesting of a trading system. The
 cost is a trivial $1.95, but beware the lecture is industrial strength.

 You can limit the dollar size of each position and limit the number of
 positions you can hold, thereby removing the total dollars in the account as
 a determinant of profit.

 Also note that there is several different levels of running the backtester.
 Allowing a very high level of control over what it does - This comes at a
 cost of complexity of programming.

 You can add custom written calculations to the backtest report.

 Reef

 PS
 I personally feel that backtesting is a hall of mirrors where you can
 easily fool yourself into thinking that you have a profitable system when
 you don't. Or conversely - needlessly discard a workable system because of
 poor backtest results. Be VERY skeptical of any results either good or bad.


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Ted Byers
 r.ted.by...@... wrote:
 
  Thanks. That helped a lot.
 
  However, I am not especially happy with the following:
 
  System test report window

 
  correctly_annualized_perc_return = 100% * ( (final_value/initial_value) ^
 (
  365 / days_in_test ) - 1 )
 
  where x^y means rising x to the power of y
 
 
  The problem I find lays in knowing precisely what initial_value is
 supposed
  to be. Yes, the formula is correct, and makes sense, if the initial value
  is fully utilized in the investment (such as the initial and final value
 of
  a house one has purchased). However, in the system I am working on, at no
  point does the total value of all open positions exceed $100,000, and yet
 I
  have been instructed to set, as a default for backtesting purposes,
 initial
  cash at 1,000,000 (the trader we're working with doesn't want to deal
 with
  any limits posed by the available cash). But that means I can arbitrarily
  improve the rate of return by a factor of ten if I use an initial value
 of
  100,000 or arbitrarily make it worse by using an initial value of
  2,000,000. Obviously, I can't use an initial value less than 100,000 or
 the
  trades the system recommends would not be possible, but apart from that
  constraint, that I can arbitrarily alter initial value without affecting
  total return seems to make this figure just so much meaningless BS. At no
  time, when using this code, is position size computed from the available
  cash.
 
  How, then, can I define things in such a way that ensures that the rate
 of
  return actually has a meaningful value? So far, it seems I can only
 compute
  a meaningful value for the rate of return for a specific trade (where the
  initial value is obviously the cost of buying the shares, for a long
  position, or the value received from the sale of shares, for a short
  position), but not for a collection of trades distributed over a period
 of
  several years. Is there a standard way for handling this? One that is
  demonstrably valid?
 
 
  Thanks
 
 
  Ted
 
 
  On Mon, Feb 8, 2010 at 4:11 PM, reefbreak_sd reefbreak...@... wrote:
 
  
  
   In AB, click on HELP - SEARCH
   type in BackTester Report in the search window
   Under the Select Topic that comes back, click on System Test Report
   Window and you will see a definition of each term. Many of these are
   industry standard metrics of performance.
  
   You can go to www.investopedia.com and type in for example risk
 adjusted
   return and get a more detailed explaination of most of the report
 items.
  
   Reef
  
  
   --- In amibroker@yahoogroups.com amibroker%40yahoogroups.comamibroker%
 40yahoogroups.com, Ted Byers

   r.ted.byers@ wrote:
   
Hi All,
   
   
I am digging into how AmiBroker computes the various statistics it
   reports
on a given back test.
   
For example, I see the following reported:
   
Exposure %
Net Risk Adjusted Return %
Annual Return %
Risk Adjusted Return %
   
Where will I find the details of exactly how AmiBroker computes these
   from a
given suite of trades? Now, I know what a risk adjusted return is,
 but in
my previous work, it is based on 100% of the cash being invested, and
 it
carries a specific definition of risk. In the system I am working
 with
now, it is rare for 100% of the available cash to be invested, and a
   really
really long lived trade lasts only a couple weeks (though 9 out of 10
   trades
were profitable in my last back test using my

Re: [amibroker] Monte Carlo Analysis in AMIBROKER?

2010-01-25 Thread Howard B

Hi Joseph --

One straightforward way to study the effect of survivorship bias is to set
up a watchlist for each period.  Enter the tickers of those issues that are
components of that index at the beginning of the period.  On the AA Settings
dialog box, select a watchlist and set the Range to the dates that watchlist
would have been current.  Click Backtest or Explore.

You can make a single run with the date range set to a longer period, but
you still need some way of identifying which issues to use for which date
range.  You could use a watchlist for each period, then have the afl check
the date being processed and verify that the ticker being processed is a
member of the watchlist that goes with that period.  Coding that in the afl
is possible but tedious to code and hard to maintain.

The first way will probably give some distortion at the date boundaries for
those trades that are open across the boundary when the watchlist changes.
Since you hold no position overnight, you will not be affected by this.  The
second way avoids the boundary problem, but at considerable added program
complexity.

The longer typical positions are held and the further back in time the tests
are run, the more survivorship bias becomes an issue.  The key to avoiding
survivorship bias is to work with a list of tickers that you would have
chosen if you were running the tests on the date the test date range begins.

Thanks,
Howard


On Sun, Jan 24, 2010 at 8:27 PM, Joseph Occhipinti 
joseph_occhipi...@yahoo.com wrote:



 Thank you Howard

 Greatly appreciated

 Just a question regarding the following comment of yours:

 *Use the lists of stocks that were in an index at the start of each year
 and run tests one year at a time, with lists reconstructed at the beginning
 of each year.*
 *
 *
 Instead of back testing 'All Quatotaions', I select 'From' (selecting one
 year at a time), will this address the issue of survivorship bias? As an
 example:


 as at 01/01/2008
 a
 b
 c
 d
 e

 as at 01/01/2009
 a
 b
 c
 d

 as at 25/01/2010
 a
 b
 c


 If I were to run tests one year at a time - say i chose 01/01/2008 to
 01/01/2009 - does that mean that AB will only pick up stocks a, b and c
 given they are the only ones left in the index as at todays date?

 Or will AB pick the results of a,b,c,d and e for that year? And then simply
 omit 'e' for the following year's test?

 (When a stocks get delisted does that mean it is ommitted from any
 historical testing done?)

 MY SYSTEM:
 In short, my system is an intraday one - im in by the open and out by the
 close. So i hold no stocks overnight. So i am not exposed to trading halts
 or overnight gaps etc.
 (not sure if this makes a difference)


 Thanks again for your time Howard

Re: [amibroker] Re: Monte Carlo Analysis in AMIBROKER?

2010-01-25 Thread Howard B

Greetings --

Norgate Premium Data, a subscription service with end-of-day and snapshot
data, maintains historical data for delisted stocks on US and Australian
markets.  They are in the process of preparing historical lists of
components of indexes, including changes and change dates.

Thanks,
Howard


On Mon, Jan 25, 2010 at 10:53 AM, Mike sfclimb...@yahoo.com wrote:



  If I were to run tests one year at a time - say i chose 01/01/2008 to
 01/01/2009 - does that mean that AB will only pick up stocks a, b and c
 given they are the only ones left in the index as at todays date?

 Most likely your data provider will have deleted the remaining symbols from
 your database such that, yes, only {a, b, c} will be used at any date. Some
 providers will have moved the symbol to a special 'delisted' area, but they
 might no longer be marked as included in the index. So, again, only {a, b,
 c} will be used.

 Mike


 --- In amibroker@yahoogroups.com amibroker%40yahoogroups.com, Joseph
 Occhipinti joseph_occhipi...@... wrote:
 
  Thank you Howard
 
  Greatly appreciated
 
  Just a question regarding the following comment of yours:
 
  Use the lists of stocks that were in an index at the start of each year
 and run tests one year at a time, with lists reconstructed at the beginning
 of each year.
 
  Instead of back testing 'All Quatotaions', I select 'From' (selecting one
 year at a time), will this address the issue of survivorship bias? As an
 example:
 
 
  as at 01/01/2008
  a
  b
  c
  d
  e
 
  as at 01/01/2009
  a
  b
  c
  d
 
  as at 25/01/2010
  a
  b
  c
 
 
  If I were to run tests one year at a time - say i chose 01/01/2008 to
 01/01/2009 - does that mean that AB will only pick up stocks a, b and c
 given they are the only ones left in the index as at todays date?
 
  Or will AB pick the results of a,b,c,d and e for that year? And then
 simply omit 'e' for the following year's test?
 
  (When a stocks get delisted does that mean it is ommitted from any
 historical testing done?)
 
  MY SYSTEM:
  In short, my system is an intraday one - im in by the open and out by the
 close. So i hold no stocks overnight. So i am not exposed to trading halts
 or overnight gaps etc.
  (not sure if this makes a difference)
 
 
  Thanks again for your time Howard
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  
  From: Howard B howardba...@...

  To: amibroker@yahoogroups.com amibroker%40yahoogroups.com
  Sent: Mon, 25 January, 2010 4:12:08 AM
  Subject: Re: [amibroker] Monte Carlo Analysis in AMIBROKER?
 
 
  Hi Joseph --
 
  There are many uses of Monte Carlo in the fields of econometrics and
  financial analysis and modeling. But the three described below are the
  most applicable to trading systems development. Some are easy to
  implement in AmiBroker, others are more difficult. Some are useful,
  others are not useful or poor practice.
 
  1. Use Monte Carlo techniques to study the robustness of a trading
  system to small changes in the data. Small, random amounts of noise can
  be added to the open, high, low, close, and volume to see if the
  trading system is sensitive to noise in the data. This is easily done
  and useful. There is a more detailed explanation, including code, in my
  book, Quantitative Trading Systems.
 
  2. Use Monte Carlo techniques to study the robustness of a trading
  system to small changes in values of parameters. When an optimization
  is performed, the value of an objective function is calculated for
  every set of parameter values tested. The best set of parameters is the
  set that give the highest value of the objective function. If we
  consider a two dimensional optimization, say the lengths of two moving
  averages, then we can imagine and visualize the objective function as a
  surface above (or below) the plane defined by the two variables. If the
  highest value of the objective function is an isolated peak, then the
  system is sensitive to changes in the relationship between the model
  and the data being modeled, and even small changes in the
  characteristics of the data will cause a shift in the position of the
  optimal solution. That is, the system is not robust relative to changes
  in the values of the parameters. If, on the other hand, the highest
  value of the objective function is a broad plateau, then the system is
  relatively insensitive to changes in the relationship between the model
  and the data and small changes in the characteristics of the data will
  not result in significant changes in the profitability of the system.
  That is, the system is robust relative to changes in the values of the
  parameters.
 
  Monte Carlo techniques can be used to study the sensitivity of the
  system by adding random noise to the values of the parameters, testing
  solutions near the optimal solution. There are many subtle issues that
  arise when performing this type of study, making general solutions very
  difficult. Specific solutions are easy

Re: [amibroker] Monte Carlo Analysis in AMIBROKER?

Hi Joseph --

There are many uses of Monte Carlo in the fields of econometrics and
financial analysis and modeling. But the three described below are the most
applicable to trading systems development. Some are easy to implement in
AmiBroker, others are more difficult. Some are useful, others are not useful
or poor practice.

1. Use Monte Carlo techniques to study the robustness of a trading system to
small changes in the data. Small, random amounts of noise can be added to
the open, high, low, close, and volume to see if the trading system is
sensitive to noise in the data. This is easily done and useful. There is a
more detailed explanation, including code, in my book, Quantitative Trading
Systems.

2. Use Monte Carlo techniques to study the robustness of a trading system to
small changes in values of parameters. When an optimization is performed,
the value of an objective function is calculated for every set of parameter
values tested. The best set of parameters is the set that give the highest
value of the objective function. If we consider a two dimensional
optimization, say the lengths of two moving averages, then we can imagine
and visualize the objective function as a surface above (or below) the plane
defined by the two variables. If the highest value of the objective function
is an isolated peak, then the system is sensitive to changes in the
relationship between the model and the data being modeled, and even small
changes in the characteristics of the data will cause a shift in the
position of the optimal solution. That is, the system is not robust relative
to changes in the values of the parameters. If, on the other hand, the
highest value of the objective function is a broad plateau, then the system
is relatively insensitive to changes in the relationship between the model
and the data and small changes in the characteristics of the data will not
result in significant changes in the profitability of the system. That is,
the system is robust relative to changes in the values of the parameters.

Monte Carlo techniques can be used to study the sensitivity of the system by
adding random noise to the values of the parameters, testing solutions near
the optimal solution. There are many subtle issues that arise when
performing this type of study, making general solutions very difficult.
Specific solutions are easy to code by running a second set of optimizations
that look at the solution space near the previously selected optimum.
Additionally, some of the optimization methods included with current
releases of AmiBroker (such as the non-exhaustive method known as cmae --
Covariance Matrix Adaptation Evolutionary Strategy) have a robustness
component that is used with no need for additional coding by the trading
system developer.

3. Monte Carlo techniques can be used to study the risk profile of a
sequence of trades.

Your question prompts me to ask how the tests you are running are defined.
If the universe of stocks being tested is comprised of the 3000 stocks that
are the current members of the Russell 3000 index, and the test period is
the past ten years, then there is a considerable survivorship bias in the
test runs. That is, the 3000 companies that are in the index now have
survived the past ten years, but those companies that disappeared during
that period are not included in the tests. That bias strongly affects the
test results. In some of my research, I have compared two studies:
1. Use the list of stocks currently in an index.
2. Use the lists of stocks that were in an index at the start of each year
and run tests one year at a time, with lists reconstructed at the beginning
of each year.
The results of the first study are always significantly better than the
results of the second study. Ignoring the survivorship bias will cause the
trading system developer to significantly over-estimate the likelihood that
the system will be profitable in the future.

Norgate Premium Data (http://www.premiumdata.net/) is an excellent source of
end-of-day data for the US and Australian markets, including data for issues
that have been delisted. They are in the process of developing historical
lists of components of major indexes which will be very valuable for study
of the effects of survivorship.

Your question also raises a related issue about how trades are selected.
Some developers run a general test using a large universe of possible issues
to trade, which results in a number of potential positions to enter that is
greater than the funds available to take those positions. They then consider
using Monte Carlo techniques to analyze what might happen if different
combinations of issues are purchased. This is an inappropriate use of Monte
Carlo analysis and is poor trading system development practice. I do not
know of a single trader or trading company who runs a test or report,
generates a list of signals, sees that it has more signals than he or she
has money, and rolls dice to determine which of the

Re: [amibroker] Monte Carlo Analysis in AMIBROKER?

Hi Ton --

Yes, Equity Monaco is a NeoTicker product. It accepts a list of closed
trades from a text file as input. You can get that by exporting results
from the AA report, cleaning it up with a spreadsheet if necessary, and
importing it into Equity Monaco. You definitely can use Equity Monaco with
AmiBroker.

My next book is Advanced AmiBroker. I am taking a short break from working
on it this morning to read postings to the forum. Writing is going well,
but some of my other projects have had big increases in activity and are
competing heavily for time, so the book is not yet ready for the editors or
the printer. No promises, and no pre-orders yet, but the availability date
now looks to be April or May 2010.

Thanks,
Howard

On Sun, Jan 24, 2010 at 10:44 AM, Ton Sieverding
ton.sieverd...@scarlet.bewrote:

Hi Howard,

*Returning to the question of reordering trades to study the risk
associated with the trading system. Use of Monte Carlo analysis in this area
is very valuable. It is best done using a program that accepts a list of
closed trades and performs the risk analysis. Equity Monaco, available free
(**http://www.tickquest.com/product/equitymonaco.html*http://www.tickquest.com/product/equitymonaco.html
*), is a good one to start with. And Market Systems Analyzer (**
http://adaptrade.com/* http://adaptrade.com/*) has more capability and a
trial version.*

My experience with Equity Monaco stops at the first page of the manual.
Isn't this a NeoTicker product ? Or at least for NeoTicker. Can I use it
with AmiBroker ?

oh yes, something else. And you know what ... your next book. When ?

Regards, Ton.

Re: [amibroker] Monte Carlo Analysis in AMIBROKER?

Hi Markus --

I also own Tharp's books.  He does use a proprietary package called Know
Your System for his analysis.

There are many ways to use Monte Carlo techniques.  The different algorithms
are more about generating random numbers and defining the distributions from
which the random values are drawn and the way they are used.

Tomasz has implemented the Mersenne Twister algorithm in AmiBroker.
Mersenne twister is much better (higher in metrics that describe randomness,
such as having a long period and not having patterns within the series) than
the random number generators that come with C++, Excel, etc.  If you use
Excel for Monte Carlo analysis, be sure to get a Mersenne Twister Excel
addin before you start.  Here is one source:
http://www.financial-risk-manager.com/risks/analytics/random/rand.html#mt19937.
There are others.

The distributions can be very problem specific, but all start out using
uniform distributions and build from there to normal, and so forth.

Back to Tharp.  I have some criticisms of Tharp's Definitive Guide to
Position Sizing.  There are two in particular.

1.  He arbitrarily and inappropriately sets an upper limit on the number of
data points to be used in calculating his System Quality Number.  (His SQN
is essentially a t statistic, which he acknowledges in one sentence early in
the book, but ignores from then on.)
2.  He is completely unrealistic about what level of SQN trading system
developers should be able to achieve.  As a quick and dirty measure, a t
statistic of about 2.0 suggests significance at about the 0.05 level for N
of about 20 data points.  A trading system that uses expectancy as its
metric, computes the t statistic on actual trades or truly out-of-sample
results (in-sample results have no value as estimators of future
profitability), and has a t statistic of 2.5 to 3.0 will result in
extraordinary profitability.  Tharp talks about achieving scores of 6 or 7.
Give any one of us a system that has a score of 6 and we can buy Manhattan
in about a year starting from $10,000.

So, read Tharp.  His books do have value.  But be aware that there are many
aspects of trading system development, testing, and validation that he
simply does not understand -- and some of his writing are seriously
misleading.  I have made many posts on Aussie Stock Forums (
http://www.aussiestockforums.com/) on this topic, and also on trading system
development.  Search using my name.

I have corresponded with Van Tharp about some of these issues and he
acknowledges my points.  He even gives me credit in DGPS.  One of the
attendees of my workshops in Australia a few months ago is a personal friend
of Vans and agrees with my assessment.

Thanks,
Howard



On Sun, Jan 24, 2010 at 10:42 AM, Markus Witzler funny...@web.de wrote:



 Hello Howard,

 just stepping in here since MC analysis may be an issue keeping me busy in
 near future (though I haven´t yet developed much expertise in it).

 Are there different algorithms of MC simulators?

 I wonder what to look for when considerung a product, since different
 qualities in algorithms may turn out different levels of quality (i.e.
 what scenarios to expect in the future).

 Van Tharp for instance claimes to have Chris Anderson developed a software
 package called Know your system with an MC simulator in it - he uses it
 for studies for his book on money mangement which I own.

 I´m looking forward to buying your book since MC is covered there to as you
 said below...

 Thanks

 Markus

Re: [amibroker] Monte Carlo Analysis in AMIBROKER?