I guess you could say the market is in a constant state of flux at all time scales so there is unlikely to be any particular chart that is free of being in a constant state of change. One thing we do know is the prices and volumes, recorded and reported (expect for occasional glitches) are accurate values. This is not always the case. For example knowing the precise temperature of a piston head as an internal combustion engine is operating may not be measurable in real-time. In markets at least we do have accurate data points virtually all of the time. What one can do with these data is another matter entirely. David Aronson
Hi - may I ask what type of statistical test you apply to the back-test results you obtain? And what sort of test statistic do you use. As you known from I book I used the mean return because of its nice distributional properties. Something like a Sharpe Ratio does have this, although it does incorporate risk. Thanks David Aronson
Just to let you know for a future book or future research, some chart environments are more static than others. Some chart increments vary greatly the number of contracts or shares trades per bar (minute charts), some show a constant number of transactions but the contracts per transaction vary (tick charts) and finally there is volume bar charts, they show a specific & user defined number of contracts or shares traded per bar. This creates a fixed environment to view price movement. The only variable here is the occasional glitches from data providers but that is refreshed at the end of each day so one can see the sheer miniscule variance this creates. Statistics are a great tool in variable environments because one needs to look for the greatest percentages of potential success in environments that are always in a state of flux but fixed non-varying environments do not need this manipulated science applied to it. David would you classify yourself as a trader that does research, a researcher that trades or just a researcher looking to help traders?
I have been both researcher and trader. At the moment I trade commodities according to several trading systems whose performance has been uninspiring. The systems were acquired from a developer with the realization that they would need improvement. My efforts along these lines involve using data mining to derive signal filters. A signal filter can be seen as a model that supplements an underlying trading system by flagging signals that have a higher than normal probability of failure. I described this general approach in an article written in the Journal of the Market Technician's Association, Spring 1991 edition. If there is sufficient demand I will make a PDF file of the article and post it on the website www.evidencebasedta.com David Aronson
Perhaps I am off base here, but I do not think that the proper and diligent use of statistics alone is typically sufficient to arrive at a profitable trading method. While the use of statistics may help to validate a trading method, I don't think that anything can replace the need to first familiarize one's self with market action by following it for an extended period of time, both in real time and slowly scrolling through historical data. From my own experience, this is where the ideas and associations come from. For maximum effect, I don't think that this element can be ignored completely or circumvented by employing other people's observations. Of course, this is merely my own opinion. Although I studied statistics up to the intermediate level during my MBA studies over 20 years ago, I confess that I have never employed it rigorously first hand either in my previous banking career or in my own trading. Perhaps I should have, and maybe I would be doing better now or gotten profitable sooner. I don't know. But I do know that nothing could have replaced the first hand observation over an extended length of time. As Yogi Berra said, "You can observe a lot just by watching."
I would agree with you. Statistics comes it at the later stages of the scientific method. First, one must proposed a hypothesis which is born of a mysterious mixture of past experience, in this case with markets, inductive generalization, insight. However, this process is starting to change with data mining. Automated data mining is essentially performing the task of formulating many hypotheses and then testing them. David Aronson
To add to Thunderdog's comment on the need to "personalize" our systems, it is critical to be able to trade the systems we develop, regardless of the statistics, past performance. I know, for a fact, that I cannot trade Trend-following systems, whith their relative low percentage of wins and the need to let profits run for a long time. So, statistics, system testing, yes by all means , but within the framework of our personality and so the ability to trade them.
i use statistics all the time in my TA i trade index futures (scalping mostly) i keep track of every setups performance, positive expectancy, %age winning trades, etc. what i found really interesting in developing setups is how (relatively) small changes in target and stop settings can RADICALLY change results/expectancy, etc. for example, one setup i have uses a 10 pts stop and a 5 pt and 7 pt target (two contracts) even changing the stop (moved in) by 2 pts COMPLETELY changed the results of this setup, and even moving the target out by 2 pts did too! this goes to show, among other things, that TA is not just about "patterns" and some such stuff (i don'[t really use them anyways, but i digress). it is about developing setups based on PRICE (all TA is based on price, or a derivative of same) that have positive expectancy given the parameters used (when entered, contraindications, target, stop, etc.) a classic mantra of "let your winners ride and cut your losses short" may work great with SOME setups in trading (and i am talking trading here, not investing) but NOT with others. your targets and stops need to be logical for the structure of the setup you are trading, and the market you are trading i've found that many of my setups that work GREAT with YM, for instance, do not work at all with ES (negative expectancy). a number of theories why (price weighted vs. cap weighted? # of issues present, spread, etc.) but it was statistics that, among other things led me TO YM (vs. ES) simply because the statistics showed much better results using the same parameters. also, can look at adverse moves for example to more finely tune parameters. also, a big thing for me is to look at the stats for contraindicating factors. iow, i look at the losing trades for a setup and try to find commonalities . keeping out of bad trades more important than finding good ones imo
Just to clarify something for you, not all trend following systems have low percentages of wins or the need to let profits run for long periods. I agree some do and they taint the culture for the others. I agree as well that one must trade their own system/s, first and foremost. The biggest problem comes in when individuals can't or won't develop their own systems and can't successfully implement or understand the systems or methods of others. Each trader is unique and must trade a method that is comfortable to them but it takes screen time and research on the part of each one to make a success of whatever method they decide to go with. Blindly trusting the "next best thing" is fatal in this business without fully checking it out. Those that taut these as well help traders implode if they don't make it crystal clear that verification of information is crucial. Some new ideas are ground breaking but others are back breaking.
Interesting. But are you talking about actual statistical tests or simple bookkeeping comparisons? I certainly engage in the latter. Also interesting is that my favorite, and almost exclusive, trading vehicle is ES. I'm big on reliability and risking as little as possible per trade, despite the fact that some people have said that you compromise potential results with such restraints. Even so, that's where my comfort lies. But I don't really need to perform extensive and statistically valid testing to see that the way that I prefer to trade is best accommodated by ES. I can just informally scroll through charts covering different market conditions for the various index markets and comfortably arrive at this conclusion. Unlike Steve Martin, I guess that I'm just a mild and lazy guy.