Trading the Model: First Annual Report

It's been one year now since I started trying to trade the model - time for an assessment. I didn't trade much this month but did pretty well - a profit of $2691 or 11.95% on capital deployed. When I first started trading the model I had two good months followed by four losing months and now six winning months. The model itself only had one losing month in this past year (May):



The model and the market have similar standard deviations but the model returns far more - a sign of a very high alpha. My returns have been similar to the model but my volatility far higher. After the losing streak I cut back on the amount of capital I was using resulting in lower profits for given rates of return in the second half of the year. Obviously, I am very far from faithfully replicating the model, but as I'll show in the following, I think my performance is improving.



As this chart shows the model returned 102% for the year vs.23% for NDX and my performance of 76%. The model's return is entirely unlevered. Looking at a daily chart:



we can see that some short periods when the model got short correctly added a lot to returns. The only ways the model can add value is by shorting when the market goes down and by stopping out of long positions when the market goes down more than 1.25% in a day. Obviously the model makes mistakes too, but over time they are less important than its correct moves. I don't have daily data for my account, but I can show you trade by trade results for NQ futures contracts since November:



This shows clearly how my trading has been far more erratic than either the market or the model. The following chart compares my monthly returns to the model:



Months below the line show poor performance relative to the model and points above the line relatively good performance. The last three months have been three of the best by this measure. My performance has a beta of 1.67 to the model with a monthly alpha of -4.9%. This means that when the model returns zero I lose nearly 5%. But looking at the very small sample of just the last six months I have a monthly alpha of 5.3%. Relative to the market the results are clearer - I have a beta of -0.97 and a monthly alpha of 7%. This alpha is significantly greater than zero with a probability of 97%. So I do particularly well when the market declines. The model's beta in the last year is around -0.16 with a monthly alpha of 6.4%. So it too has a short bias.

In conclusion, I think I am gradually learning to trade with this model - my results are beginning to be statistically significantly positive and nonrandom.