Progster, > I would add that a fix would be trivial - simply supply one additional > bit (or field) of data which indicates the HL precedence!
7000+ people in this forum - all trading (or working towards that) in different ways. No chance to get providers to do any pre-conditioning of data to suit one group of users. > IMO. "Adjusted" data in truth means "actually invalid for > all analysis other than the exact scenario "adjusted for". Definitely the data type required is specific to the 'project' at hand. Data 'quality' is often discussed, by traders, whereas suitability of the data for the task and whether the way in which the data is manipulated is the most relevant way to do it are seldom discussed (except in futures trading where the pros and cons of creating continuous contracts are well known) e.g. if option trading was being analyzed, with reference to the underlying (say for comparing implied volatility to actual volatility == indirectly the cost of the option i.e. the premium) then there is no point using adjusted data if the options are not also 'adjusted' in a corresponding way. > The correct approach, IMO, would be to supply the historical actual > prices as traded on each day, with additional fields indicating >splits > or other relevant actions. It would then be the proper job of the > application software to display "continuous" charts if requested by > the user. Some data providers actually do this (at the UKB I have tried to highlight the differences between data providers, where adjustment is concerned - so far I have only covered the ASX but I will do the US markets at some point). Some traders are keeping their own splits/div records etc and do their own adjustments (I can only watch in awe). I agree a good data provider would supply raw data, events calendar, holdday/market closed/trading suspended and the means to optionally adjust/pad data. How many good providers are out there? IMO a large part of trading involves understanding, acquiring and managing our data. Very little is said about this in the trading books (millions of pages on complex indicators though). That is why I am pleased to see it discussed in the AB 'forums' and also why I have made it a topic at the UKB (even though I am not the best person for the job I regard it as so central to trading that I took some baby steps on the subject over there). Personally I think we are on very shaky ground once we start to analyze data 'quality'. I am not a data guru so all I can do at the UKB is provide some insight into the key components of data management so that newcomers to trading are equipped to go further on their own. Lucky for me my temperament and style allows me to not be too concerned about it. As I have said before, I have sometimes used Yahoo data (US only) at the initial stages of some projects, and then gone on to successful completions, even though I know the data isn't the best. That is just me and I'm not concerned whether others do or don't. brian_z --- In [email protected], "progster01" <[EMAIL PROTECTED]> wrote: > > >5) imo, the OHLC bar structure is obsolete. Imagine the design > >advantage if you knew the HL precedence. Of course, in the majority > of >cases, you can be sneaky and calculate HL precedence by assuming > that >the price follows path of least resistance from the Open to the > >Close... or subscribe to real-time data. > > Well said, and I agree with this. > > I would add that a fix would be trivial - simply supply one additional > bit (or field) of data which indicates the HL precedence! IMO, data > vendors (starting with the exchanges) are negligently cavalier in > failing to provide this simple, factual bit of information about what > actually happened (when the only purpose of the data to begin with is > to provide a record of _what actually happened_ !). > > While I'm on this particular soapbox, the same thing applies to > splits, IMO. "Adjusted" data in truth means "actually invalid for > all analysis other than the exact scenario "adjusted for". > > The correct approach, IMO, would be to supply the historical actual > prices as traded on each day, with additional fields indicating splits > or other relevant actions. It would then be the proper job of the > application software to display "continuous" charts if requested by > the user, according to the various optional ways that can/could be > done. IOW, "adjustment" needs to be a user-specified action, > according to the algorithm appropriate for the intended purpose(s). > > For many purposes, no adjustment at all would be required or desired, > because indicators (whose discontinuities might alarm!) would not be > at the heart of system. Instead, the strategies would simply be aware > of share count changes at split points and update the shares held (and > share price) for position open across that point. This would create > calculations and results which would be exactly accurate historically, > in all respects, margin-wise, percentage-wise, capital-use-wise, etc. etc. > > User software would be more complex, and users would have more to > think about - but that would all be much preferable to everything one > does simply being silently wrong. (Unless, of course, you are > counting on that to provide a continuous supply of sheep to be shorn!) >
