tag:blogger.com,1999:blog-107568321062020427.post3375438094422201065..comments2023-09-19T00:48:06.643-07:00Comments on Intelligent Trading: Genetic Algorithm Systematic Trading Development-- Part 2Intelligent Tradinghttp://www.blogger.com/profile/17765336450326139518noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-107568321062020427.post-35765030437032675182010-06-13T22:03:04.721-07:002010-06-13T22:03:04.721-07:00Intelligent Trading said...
Hi Jozef,
GA is one o...Intelligent Trading said...<br />Hi Jozef,<br /><br />GA is one optimization procedure that is useful in evaluating large search spaces. If the permutation set of your input rule space is small, then yes, it might be simple enough to just run the sharpe ratio or suitable metric of some trading rules and use that as an optimization criteria using simple sorting. However, as the permutations of inputs and rules become large, some type of optimization method like GA becomes more necessary. <br /><br />Imagine you have a rule; if ma(m)>ma(n) then buy, and suppose m and n are swept over 30 variables each, that's 30^2 combinations and about ~1000 variable sets to sweep over. Then you have the < case to test, and buy and sell rule evaluations, so multiply by about 4. You are now sweeping over a set of about 4000 variables. But, what happens when you start to add additional boolean rules. <br />For example, if ma(m)[><]ma(n) [AND,OR] rsi(y) [><]rsi[x] then [buy,sell] elseif [var[t] [><]var [e]] then [buy,sell]... Do you see how the set of rules to evaluate is becoming impossibly large?<br /><br /><br />Yes, the example was meant to be very simple for intuition, as such a simple set of parameters could easily be evaluated via a sweep of variables over some range.<br /><br />I like your blog at first peek, and will look a bit more.<br />Thanks for stopping in.<br /><br />ITIntelligent Tradinghttps://www.blogger.com/profile/17765336450326139518noreply@blogger.comtag:blogger.com,1999:blog-107568321062020427.post-13246166769068827952010-06-13T11:20:56.285-07:002010-06-13T11:20:56.285-07:00hi,
I do not see the point of using GA to find th...hi,<br /><br />I do not see the point of using GA to find the best rule set. Would not it be easier to just simply calculate e.g. Sharpe ratios of different MA combinations in the in-sample period and then simply trade the one with the highest Sharpe in the out-of-sample? I want to say, that GA is just an optimization technique, such as pattern search,etc. Maybe you are just trying to show how one would go about coding the system using GA and you are showing just very simple example. However, I cannot imagine in what case such a procedure would be justified. I think however that I probably have tonnes to learn from you (as you are 10 years in financial markets and living from trading systems). Here is my blog: jrudy.wordpress.comAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-107568321062020427.post-72663228205415192502010-02-17T20:18:49.039-08:002010-02-17T20:18:49.039-08:00You are absolutely correct. I learned how to count...You are absolutely correct. I learned how to count a long time ago in binary, so sometimes I wrote out the truth tables quickly without verifying. It was a slip and has been corrected.<br />Thanks for your keen observation as that was a very important point!Intelligent Tradinghttps://www.blogger.com/profile/17765336450326139518noreply@blogger.comtag:blogger.com,1999:blog-107568321062020427.post-29615067307443126372010-02-17T19:25:01.542-08:002010-02-17T19:25:01.542-08:00Hey! Thanks for all the great posts. In Fig. 1, I ...Hey! Thanks for all the great posts. In Fig. 1, I think you meant the decimal 5 be represented by the binary string 101 rather than represented by the binary string 100.Bradnoreply@blogger.com