Take the 2-minute tour ×
Quantitative Finance Stack Exchange is a question and answer site for finance professionals and academics. It's 100% free, no registration required.

I am aware of: Encyclopedia of Chart Patterns, Encyclopedia of Technical Analysis.

Question

I'm wondering if there's something similar, but in the form of:

"Encyclopedia of Statistical Backtesting to see if you're being fooled by randomness" and "Encyclopedia of techniques for generating random forex patterns."

share|improve this question
    
you probably just need to understand what you are doing, every statistical test is different and you need a good understanding of the underlying assumptions, judging by your open ended questions you might be better off reading some statistics or probability books –  pyCthon Sep 28 '12 at 18:32
    
I honestly think (and i know that obviously many people like this question) that this question is out of scope (refering to the faq-section "questions i should not ask"). In my time as mathematician i have come across so many statistical tests (and i am not even a statistician) which are always at least somehow tied to their application. I mean there are literally thousands of them and imho it would be useless effort to pot them together in an encyclopedia. Why? If you know some statsitics you can develop a test yourself or look into the specifics if needed. A cookbook does more harm than good –  vanguard2k Oct 2 '12 at 6:28
add comment

6 Answers

I believe this is very difficult to do because of the different nature of statistical tests. Some of them are used to test the assumption of normality, some of them allow you to compare the volatility of different samples, some of them allow you to determine the suitability of a specific model.

Essentially you will find the basic ones on any good statistical book with a hypothesis testing chapter. For the more advanced one, you will need an introduction to the topic first which require ... a book by itself.

share|improve this answer
add comment

There is "A Review of Backtesting and Backtesting Procedures" by Sean D. Campbell.

It is related to VaR backtesting, but the article discusses the general properties of unconditional coverage and independence, giving the statistical power properties of different backtests.

I do hope it helps.

share|improve this answer
add comment

If you run enough tests, you are guaranteed to find something that "works" in a backtest. The problem is.....Does it work in real time trading? If not, then you were "fooled by the backtest". If it does work, the next question is.....For how long? Markets evolve 100% of the time. So, when the thing that "works" eventually dies out, was it really there or were you just lucky?

The bottom line is.....Is it reasonable to assume that what you're looking for is predictable? If not, then why waste your time. If so, then what advantage do you have over others in exploiting that predictability?

share|improve this answer
add comment

You probably would do well to understand the question you are asking. Read the seminal text "Subset Selection In Regression" on the subject.

An encyclopedia of statistical tests applied without understanding would be as damaging to your wealth as both the Encyclopedias you mention above - if I read into your question correctly.

share|improve this answer
    
(+1) for funny last paragraph. –  user2921 Sep 29 '12 at 8:44
    
@DangerMouse "Subset Selection in Regression" hm... as far as a major online distributor is concerned the book is not the cheapest of its kind. Do you happen to have a book review for this one available? –  vanguard2k Oct 2 '12 at 8:58
    
@vanguard2k Not sure I understand your question. The ones on the .com site of a major online distributor are pretty accurate. It is a fairly expensive book, but at least that stops everyone reading it. I came accross the book via a personal recommendation. –  DangerMouse Oct 3 '12 at 14:34
    
@DangerMouse I dont doubt that its a great book I just thought it would be great to read a book review before ordering it. (And I almost surely will) Just to see if it is too theoretical or too "shallow" material. –  vanguard2k Oct 3 '12 at 15:17
    
It's very mathematical - you need good level of linear algebra (regression, QR decomposition etc). If you feel comfortable with Hamilton or ts analysis books then you should be okay. He talks about the maths as what to do in solving the problems and doesn't explain the maths in terms of teaching linear algebra. You may find some online excepts if you google carefully. –  DangerMouse Oct 4 '12 at 9:16
add comment

I'm also looking for something like this, a book with setups and examples of statistical tests specifically for trading strategies. I picked up "Probability and Statistics for Finance" (Frank Fabozzi Series). I'm still reading it but I have not yet encountered an example of an application to an actual trading strategy/backtest. However, it's relevance and examples in general finance does help give it a context for the statistics/probability concepts described.

share|improve this answer
add comment

The best book concerning statistical validation of objective technical indicators would be David Aronson's highly acclaimed Evidence-Based Technical Analysis.

While the first quarter of the book is spent establishing the distinction between chart eye-ballers and objective verifiable TA, the latter portion of the book is an absolute goldmine as it starts off with the basics of statistical inference and then covers the monte-carlo permutation method and bootstrap tests extensively. In the end, numerous technical indicators were tested on the S&P and none of them exhibited statistically significant predictive power, although the author did mention that combinations of such indicators could be lucrative.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.