No registration required! (Why?)

Evidence-Based Technical Analysis

Evidence-Based Technical Analysis

This book is about data mining to find trading rules and then verifying that the rules that were found are not co-incidences and are in fact valid rules that can be applied to current and future strategies. (I picked that up from the publisher's excerpt below.) Has anybody read this book and have any comments on it? (I have not yet read it?)

Evidence-Based Technical Analysis is a breakthrough book in that it rigorously applies the scientific method and recently developed statistical tests to determine the true effectiveness of trading strategies, rules or systems discovered by data mining. Traditional technical analysis – as currently practiced – is more like a faith-based folk art than a science, the author asserts. To move technical analysis forward, the author proposes a new type of technical analysis, which he calls: evidence-based technical analysis or EBTA. Unlike traditional technical analysis, EBTA is restricted to objective methods whose historical profitability can be quantified and then rigorously scrutinized. The author provides a new statistical methodology specifically designed for evaluating the performance of rules that are discovered by data mining, a process in which many rules are back-tested and the best performing rule(s) is selected. Experimental results presented in the book show that data mining is an effective approach for discovering useful rules. However, the historical performance of the best rule (s) is upwardly biased - a combined effect of randomness and data mining. Thus new statistical tests are needed to make reasonable inferences about the future profitability of rules discovered by data mining. Most importantly, in data mining case study the author evaluates over 6400 signaling rules applied to the S&P500 Index using these new tests. For technical analysts and traders, the book is a wake-up call to abandon subjective, interpretive methods and embrace an approach that is scientifically and statistically valid. For other traders, the rigorous testing of trading signals/rules may make their data mining efforts more productive and stimulate the development of new systems, signaling rules.
Originally posted by day trading

Originally posted by inventor

I assume this is a day trading set up?
wont it be simpler with 20 and 50 MA?

pt is talking about creating an MA fan. For that you need 3 or more moving averages I believe otherwise you can't create a fan.
Yes I believe that we are talking about day trading here.

The moving average fan is a trading pattern identified/developed by Gary (DD at #emini). He has a blog that covers the MA fan pattern in infinitely greater detail than we are in this thread. He uses slightly different moving average lengths than I did here, but the basic concept is the same.

Correct, this is a simple intra-day pattern of price and volume we see repeat almost every trading day. I felt it a good one to use for this discussion because it uses simple technical analysis to describe a fundamental concept in MP.
Originally posted by day trading

When you talk about a chart set-up like this do you typically only look at the Regular Trading Hours (RTH) data or do you look at the continuous session?
For time based charts I prefer RTH. For Tick charts I use all sessions.

Secondly, I notice that you talk about volume drying up. Have you used volume charts instead of time based charts to look at this sort of thing?
For this particular pattern you will not see it using a volume chart, since the volume dry up will be reflected in 1 or two bars, rather than 10 or 20.

One of the problems that I have with moving averages on a time based chart is that you are giving the same weighting to the closing price of each time period irrespective of how many traders were active in that time period.
True, It simply depends on what your trying to see. A volume based chart de-emphasizes the dead zones (inactive periods, like 3am or noon) from the chart. Doing this certainly has it's advantages on an all-sessions intra-day chart.

Another way of getting around my bias against time based charts, or rather of making me happier to look at a moving average based on a time based chart, would be to use the average of the high and low instead of the closing price as the input into the moving average formula.

My primary concern in setting up a chart and selecting or using a technical indicator on that chart is this... I am very selective / restrictive in what I look at when making trading desicions in real-time. I want to use things that I know from experience work most of the time. The 100 minute moving average is a good example of this idea. Price will stop there almost every time it is touched. I beleive this happens because it is a widely used and watched moving average length, specifically it is used by professional screen traders. Put the 20 ema on a basic 5 minute OHLC bar chart and watch how price reacts to it during the day. Gary uses the 144 minute moving average, the number 144 has some mystical significance to him personally, but if you plot it against the 100, the two lines are usually very close to each other. I personally prefer to de-emphasize the mystical from my trading plan, and prefer to stick to the simple basics of what works... bringing us back to the original point of this topic (EBTA)

I apologize for semi-threadjacking this but I value your opinion.

It's a good discussion, bringing out a lot of good ideas and questions.

Originally posted by pt_emini

...I want to use things that I know from experience work most of the time. The 100 minute moving average is a good example of this idea. Price will stop there almost every time it is touched. I believe this happens because it is a widely used and watched moving average length, specifically it is used by professional screen traders...

I was wondering if you reversed this logic and said "I wonder which moving average most people watch?" So instead of reading about and listening to what traders use you run a back test to see which moving average provides support/resistance most of the time. From the back test we find the one that works most often or consistently and come to the conclusion that the reason that it works is because it is being watched by the highest number of traders or at least it is being watched by the traders that control the most amount of money.

So by reversing this logic in any market (assuming that the support/resistance is self-fulfilling by the volume of traders) we can establish which moving average to use.

This is of course curve fitting but if everyone does it and follows it then it becomes stronger and stronger and the signals become better and better.

Now what happens is that we end up in a cycle. Traders will start front running the signals and so in order to get into the trade in a timely manner you need to shorten the length of the MA that you're using. This leads other traders to shorten the length and the cycle continues until it becomes too short and breaks and then it returns to the original "long" length.

So if we looked at this over a time period we see a gradually decreasing cycle of moving averages being the best fit over time. Say the cycle takes 70 days to complete and the moving average moves from 100 periods to 30 periods. Each day you reduce the MA that you use by 1 period a day (trading days not calendar days) and that way you stay just one step ahead of the crowd and making that little bit more.

This is of course pure theory and speculation and I have never tested this but I like the theory/idea - although it is far fetched!
Your front-running theory might explain why particular indicator settings tend to drift into and then back out of favor. I have always viewed time and price relationships 3-dimensionally, similar to a sphere (3-D) rather than a circle (2-D). Unfortunately standard charting products lock the user into this planar 2-D world.
From what you are saying pt_emini it looks like somebody needs to back test these moving averages at a level of "abstraction". What I would like to see is a 3-Dimensional chart showing the length of moving average that worked on each day and see how that changes over time.
elite: Sounds like a fun mathematical problem to work on but I think that you would need to find a skilled mathemetician and programmer to tackle that problem.
Sounds complicated to me.
I think that I am always expecting too much!!
If it was easy, then everyone would be doing it
This is true and trading has the illusion of being easy. You just need some cash and a trading system and off you go.
Not looking good for marriage if your win ratio is over 50%...