Serving the Quantitative Finance Community

 
User avatar
Muzzex
Topic Author
Posts: 1
Joined: May 2nd, 2003, 10:19 am

Technical trading signals with statistical significance

December 16th, 2010, 12:57 pm

Hi,Starting a trading job soon, I'm looking for some theories to base my trading on. Background in statistics, so I have a built-in skepticism towards technical analysis in trading. My problem is that I will have even more problems using some sort of fundamental analysis to form a gut feeling for going long/short.Trying to setup up a trading rule based on technical analysis rules (moving average, oscillators or whatever), back-test and so on.... Seems to be the way to find a strategy.Problem is that testing enough rules, you will always find a rule that does well in a back-test. I'm aware of the data mining bias.How would you proceed? Would it make sence to test your strategy on different products/markets? Would be a rather strange coincidence if your trading strategy back-tests well in several markets?What is the strategies among people with sound statistical background? I just read "Evidence-based TA" by Aronson. He tests 6400 TA rules on SP500. With his data mining bias tests, he finds not one single TA rule with results significanlty better than a random rule.I understand that since a trading rule would have to give very good back-test performance to be able to be statistically significant in a data bias mining test among 6400 different TA rules. Even if you would have one super-rule that would give you 8% return on capital every year, you simply would not find it. And I don't expect to find a rule better than that./m
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

Technical trading signals with statistical significance

December 19th, 2010, 3:05 pm

you got hired despite the skepticim you talk about. well done (?) !which type of trading the compagny you are going to enter is involved in ? the toolbox ?
 
User avatar
endian675
Posts: 0
Joined: March 11th, 2008, 7:09 pm

Technical trading signals with statistical significance

December 22nd, 2010, 10:47 am

Can you define "not significantly better than a random rule" for me?
 
User avatar
farmer
Posts: 61
Joined: December 16th, 2002, 7:09 am

Technical trading signals with statistical significance

December 22nd, 2010, 11:41 am

QuoteOriginally posted by: endian675Can you define "not significantly better than a random rule" for me?I saw a headline the other day said something like "studies show echinacea does very little to cure cold." It should have said "Stop the presses, echinacea actually does something."QuoteHe tests 6400 TA rulesI did the same every 0.5 seconds for 36 straight months.
 
User avatar
endian675
Posts: 0
Joined: March 11th, 2008, 7:09 pm

Technical trading signals with statistical significance

December 23rd, 2010, 10:23 am

QuoteOriginally posted by: farmerQuoteOriginally posted by: endian675Can you define "not significantly better than a random rule" for me?I saw a headline the other day said something like "studies show echinacea does very little to cure cold." It should have said "Stop the presses, echinacea actually does something.".I meant in statistical terms. If the values in the P-value "Boot" and "MC" columns are the probability of observing a result at least as extreme as the one observed given that the Null Hypothesis is true, then I would say there are plenty of tradeable strategies in his results. Therefore I feel I must have missed something.