I added a suggestion #1117 that would keep me in business with AB and likely generate more pro sales.
Dennis On Aug 1, 2007, at 11:39 PM, Dennis Brown wrote: > Jerry, > > Thanks for the suggestion, I will look into it again. > > I run eSignal for a data feed and it is supposed to be the most > reliable one. I keep tabs on other realtime trader's experiences, > and it is. However, I get bad ticks constantly in AB for high volume > issues like SPY. eSignal cleans up most of the bad ticks in a few > seconds, but about 10-20% of them are never fixed. Bad ticks are > easy enough to ignore when trading by eye on a chart, because the > brain can spot the absurdity of the tall spike that only happens for > an instant without a volume spike and ignore it. > > However, I have some quite complex indicator calculations that are > about the equivalent of 50 normal indicators. Some key calculations > that drive everything are based on the highs and lows of a 1 minute > bar. The bad ticks throw the calculations off by a factor of two. > It takes AB about 3 seconds to do my calculations for just 1000 > bars. If I can run the database 12 or more times faster than the 1 > minute bars, then I can run a simple algorithm that finds and removes > the bad ticks in realtime, then feed the clean 1 minute bars to the > rest of the calculations. It also lets me do things like running > clean volume bars off the same database. > > Also if AFL had static arrays, I could break up my calculations and > not recalculate everything all the time which kills the UI > responsiveness for drawing trend lines in realtime. > > I use the TimeFrameSet(60) after cleaning up the bad ticks which > turns everything into 1 minute bars. That works for everything > except the window still thinks it is in 5 second bars. All studies > that are hand drawn end up being relative to the 5 second bars (and > prices) that are implied, but not displayed, and not my 1 minute > bars. This means that the studies shift left one bar on the screen > every 5 seconds, while the price bars that I am plotting are shifting > left one bar every minute. This makes the studies march left 12 > times faster than the data I was trying to draw a trend line for. > > So TimeFrameSet() does not really take you to a new timeframe as far > as the chart pane is concerned, nor for drawing studies. > > My OHLC does not change unless I do a force backfill, which fixes > some bad tick data. If your provider is fixing bad ticks on the fly > with AB, it might be affecting your charts. > > I have been ignoring all the little gotchas while I was developing > the indicators to trade from. Now that my indicators are close to > done, I started worrying about the little gotchas that would ruin my > actual ability to execute successfully. I was counting on the little > things being cleaned up by now, or finding a work around over the > course of my one year development. AB is close to doing what I need, > but I don't know how to get from here to there. I have boasted about > how good AB was to a lot of realtime traders who are waiting for me > to prove it before they will use it. It looks like I will have to > eat my words now. > > Dennis > > > > On Aug 1, 2007, at 8:37 PM, Jerry Gress wrote: > >> Dennis, >> >> You might want to change to Ninja Trader Charting 6.0. Has neat >> order entry >> that pops up on chart etc. Just one problem my first day of trial >> (yesterday) I had a low over 2 points diff then Amibroker(and >> exchange). >> They have not gotten back to me after sending pictures of diff. >> >> Otherwise you might want to get a different data supplier. You >> state your 5 >> second drawing lines move, does your h,l,o,c also move? (See >> previous post, >> mine does). >> >> Jerry Gress >> Stockton, Ca. >> >> >> -----Original Message----- >> From: [email protected] [mailto:[EMAIL PROTECTED] >> On Behalf >> Of Dennis Brown >> Sent: Wednesday, August 01, 2007 8:07 AM >> To: [email protected] >> Subject: Re: [amibroker] Plotting different timeframes --help needed >> >> Ok, this approach will not work. Even though my graphs look right >> and the OHLC arrays are right, because I have to stay in the 5 second >> timeframe, studies are useless. All the lines I draw are internally >> based on the underlying 5 second timeframe and move every 5 seconds >> relative to the chart prices. Totally useless! >> >> SAD COMMENTARY: >> >> I continually try to solve problems related to my real time trading >> and have ask for help understanding how to apply AFL to the problems >> I am trying to solve. Most of the replies I get are from myself! I >> know this board is not shy to help when someone knows the answers, so >> I am left to draw an unfortunate (for me) conclusion: >> >> My selection of AB for my type of real time discretionary trading of >> one stock at a time was a mistake and I have wasted a year trying to >> make it work for me (and it is so close!). The majority of users >> seem to be using the rotational, portfolio, EOD, and swing trading >> methods (or are too inexperienced to know what they want) and that is >> where the support and development seems to be headed (I have used >> those methods before and wrote my own simple programs to do that nine >> years ago). >> >> Things that would make it possible to speed up real time analysis >> like support for static arrays have been put on the back burner due >> to lack of interest. Cleaning up the hand drawn study glitches and >> inconveniences seem to not be a priority because they have not been >> addressed. >> >> It really does hurt me to say it, because I really love the basic >> architecture of AFL and AB as a tool, and the community here, and I >> know that TJ is doing his best to satisfy the majority of his >> customer's needs, but if I am the odd man out and my trading needs >> are not the direction AB is headed, I have to face up to it and look >> for an alternative solution. :(( >> >> Dennis >> >> >> On Jul 31, 2007, at 5:02 PM, Dennis Brown wrote: >> >>> Hello, >>> >>> I am totally frustrated with bad ticks getting through on my 1 >>> minute >>> bars and messing up my indicators. I decided to try cleaning >>> them up >>> on my own by running a 5 second database (instead of 1 minute) and >>> cleaning up the bad ticks with a Median(c,2) statement --I would >>> have >>> used ticks, but I can only get a couple of hours worth of ticks in >>> 100,000 bars of SPY and I need several days of 24 hour data. >>> >>> I filtered the 5 second data, assigned it to OHLC, and then used the >>> TimeFrameSet(60) to create clean OHLC arrays. >>> I have to leave the overall timeframe at 5 seconds all the time for >>> the filter to work, but I want to display and work in 1 minute bars. >>> >>> When I plot it out, the bars are 1 minute bars, but the time >>> scale at >>> the bottom of the chart is set for 5 seconds per bar. >>> >>> I want to know how I can also get the time scale to display in 1 >>> minute bars? >>> >>> Or is there a completely different approach that is better? >>> >>> Thanks, >>> Dennis >>
