Henry Karstens recently posted a primer on testing strategies. A key indicator is what he calls "optimal f" which is related to the Kelly Criterion. Optimal f is supposed to tell you how much you should have in your account per standard sized trade. I computed this statistic for all of my trading of NASDAQ 100 futures since I opened my IB account - a total of 98 closing trades. I standardized the gains and losses for each trade by dividing the amounts by the number of contracts bought or sold. The criterion shows that I need $4906 per contract traded in my account. It would mean I could trade up to 79 contracts with my current net worth, or $2.85 million of underlying stock.
Another statistic that Karstens presents is the z-score which is a test of whether the average gain per trade is statistically significantly greater than zero. My z-score is only 1.04. Which means there is around a 15% probability that my system, the way I am trading it, doesn't make any money at all.
The bottom line, I think, is to take all trading test statistics with a big pinch of salt. The same data that says that there is a good probability that my results are random also could be construed as saying I should be trading with 7.3 times leverage.
No comments:
Post a Comment