Wednesday, 24 June 2015

Results of Permutation tests on Cauchy-Schwarz

Following on from my previous post, I have to say that the results have been quite disappointing - in all the tests I have conducted so far I have been unable to reject the null hypothesis. These tests are still on-going and I may yet find some gold, but it is looking increasingly unlikely and so I won't bore readers with the results of these negative tests. Unless something drastically changes I am planning on abandoning the Cauchy-Schwarz matching algorithm, at least in its current form.

For those who are interested, the test I am conducting is the data mining bias adjusted random permutation test, which is the file on my Data Snooping Tests page on Github.

Soon I shall be going away for the summer and although I will continue working on a borrowed laptop, I am not sure what internet access I will have and so this post may be my last substantive post until some time in September. I am taking a lot of reading with me, all my Octave code and data and I have loaded up my R installation with lots of interesting packages to play around with some new ideas, so hopefully there will be some interesting new things to post about in the autumn.

Tuesday, 23 June 2015

Cauchy-Schwarz Matching Algo Revisited: More Tests

Some time ago I blogged about my Cauchy Schwarz inequality inspired matching algorithm and some tests of it here and here. More recently I have come across a nice paper about developing and back testing systematic trading strategies here, courtesy of the aggregating site, and being motivated by the exhortation in said paper to conduct hypothesis driven development and separate evaluation of each component of a strategy, I have decided to perform some more tests of the matching algorithm.

The above mentioned tests were of the Effect size of differences in means between random matches of price and algorithm matched prices for 5 bars following a test point, with the test statistic being the Cauchy-Schwarz value itself. This was intended to be a test of the similarity of the evolution of price after any particular test point. However, a more pertinent test is whether this similarity can be exploited for profit, and doubly so since I intend the matching algorithm to select training examples for my machine learning development of a trading system. If there is no potential to extract profit from the basic selection of matched training examples, it would be naive to expect any machine learning algorithm to somehow magic such profit from these same examples.

The first (set) of test(s) I have in mind is a simple Monte Carlo Permutation test, which will be the subject of my next post.