random trading / efficiency of the market

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
docSmith
Contributor
Contributor
Posts: 2
Joined: Fri Jul 16, 2004 10:08 am

random trading / efficiency of the market

Post by docSmith »

There is this old idea from LeBeau and Lucas (see for instance http://www.streetstories.com/lb_futures96.html ) to decide upon the efficiency of the market by using backtesting with some kind of random trading strategy. You may check whether a specific trading strategy performs better than random by performing hundreds of random runs specifically tailored for the strategy to test. The random runs should trade with the same frequency, use the same time frame and the same money management rules, but scramble the entry bars of the positions and the symbols actually traded. Perhaps it is a good idea to test long and short positions separately.
You may see what I mean by downloading the evaluation version of Wealth-Lab (http://www.wealth-lab.com/cgi-bin/Wealt ... wnload.htm , I am not related in any way to that company) together with any of the two chart scripts BetterThanRandom3 (http://www.wealth-lab.com/cgi-bin/Wealt ... c?id=12487 and http://www.wealth-lab.com/cgi-bin/Wealt ... m?id=31882 ) or BetterThanRndTst104 (http://www.wealth-lab.com/cgi-bin/Wealt ... m?id=31833 ). This way you may check for more than thousand trading strategies (including your own, if you implement it using Wealth-Script) whether they perform better than random for a specific period and watch list of symbols. BetterThanRandom3 performs real random simulations where BetterThanRndTst1.04 uses some kind of bootstrapping technique.
Since these tests are performed by interpreted scripts in Wealth-Lab, they are quite slow. So I would be interested whether this kind of functionality is supported as a built-in feature by some other trading software. I would also be interested, if anyone knows about other scripts (for any software) doing something similar, so that it would be possible to compare the results.
stancramer
Roundtable Fellow
Roundtable Fellow
Posts: 60
Joined: Sat Feb 07, 2004 10:34 am
Location: Washington DC

Post by stancramer »

You might be interested in this thread. Free software!
viewtopic.php?p=3917
docSmith
Contributor
Contributor
Posts: 2
Joined: Fri Jul 16, 2004 10:08 am

Post by docSmith »

Thanks for your hint. But the random strategy discussed in viewtopic.php?p=3917 is not related to a specific strategy to test. What I want to do instead, is to run a specific strategy to test, then run many hundreds of random runs tailored to test this strategy. Then I may count how much percent of the random runs are better than the strategy to test. If it is below 2.5%, my strategy is signifigantly better than random.
Post Reply