Thursday 14 June 2007

Update: I Feel Validated... Somewhat

We all like to be validated. I'd like to give you an update on what I've been up to in terms of validating the COTs Timer system. As I've mentioned before, this trading system is a work-in-progress. That's one of the reasions I've urged caution about acting on my signals and recently switched from outright "buys" and "sells" to "bullish" or "bearish."

If you've been a regular visitor, you've seen that I'm often revising the setups and looking for ways to validate the system in general. The last point is important because of the small number of trades for many of the setups and the fact that the Commitments of Traders combined futures-and-options data goes back only to 1995. The question is whether past profits tell us anything about future results.

Happily, statistics give us a few ways to test this question and validate trading systems. One is the Student's t-test. It is often used to give confidence levels for trading systems and other data like survey results. Ideally, we would like a trading system to be profitable at a 95- or 99-percent confidence level. (In other words, 95 or 99 percent of the time, the system's average trade is profitable.)

A key variable is the number of trades generated by a trading setup. The more trades, the more confident we can be the future results will be profitable. Because of the small number of trades, I've found many of my setups are profitable only at the 90-percent confidence level at this point. A few are valid at the 95- or even 99-percent levels. Some still simply have too few trades to have a confidence level.

Thus, a setup needs at least 10 trades to have a confidence level, based on the number of trading variables in my system. Of if I were to drop my stop rule - which counts as two variables - a setup needs at least eight trades. Click the "Profit/Loss Results" link to the right to see the number of trades in each setup, keeping in mind that I haven't updated some of them in a while. (The "Last Updated" column tells you when.)

I plan to post some specifics for each setup soon and to keep testing the system in other ways. My apologies for not getting to this sooner. This testing has been extremely time-consuming, but it's also been an extremely valuable and educational process that's helped me find better setups in a couple of markets. I hope to post those details soon, too.

Many thanks to the readers who wrote in about these questions and were so generous with their time and sharing their insights. Your further input is welcome.

No comments: