Portfolio Versus Monte Carlo Returns

Strategies might be attractive on the surface but fail moving forward. Why? Here we discuss the qualities of effective trading strategies, and the many traps one should avoid.

Portfolio Versus Monte Carlo Returns

Postby Overload » Wed Feb 22, 2006 2:44 pm

The Portfolio Selection in StrataSearch is an important and necessary tool, since it’s the only way you can get an indication of how your system would have performed under real money-management conditions. As an example, suppose your trading system uses the Russell 2000 sector, and in a unique rally created 975 buy signals in a single week. Would you have purchased all of them? Probably not, and even if you had, your trade amounts would have been so small for you to be able to distribute you account’s cash into 975 positions. The Portfolio Selection solves this issue by assuming you’ll only buy the first trades available until you’ve filled your portfolio up through the designated Portfolio Size.

But there is a drawback. What if the stocks selected into the portfolio aren’t representative of the overall signals? Maybe, by chance, a few big winners were selected into the portfolio that perhaps might not have been selected under just slightly different circumstances.

The Monte Carlo Average Annual Returns can help provide an answer. By looking at the Monte Carlo AvgAnnReturn for the associated Portfolio Size, you can get a good indication of how alternate selection scenarios would have played out. If the AvgAnnReturn of your Monte Carlo evaluation is negative or significantly lower than your Portfolio’s return, then you should consider this a red flag.

It’s important to note that there may be justifiable reasons why a Portfolio return would be significantly higher than a Monte Carlo return. For example, based on the entry formula, the first trades selected in a wave of signals might perform better than the trades at the end. More specifically, the trades at the beginning of a rally may perform better than the trades at the end of a rally. Thus, the Portfolio selection may in fact be an integral part of the trading system, and there are solid reasons why the Portfolio returns are better than the Monte Carlo returns. But such a situation would require manual confirmation.

Likewise, a Monte Carlo return significantly higher than the Portfolio return may be an indication that the Portfolio might do better than the back test moving forward. But again, there may be reasons why the Portfolio selection performed as it did.

Regardless of the causes, the Monte Carlo returns are a helpful tool in confirming the robustness of a trading system. And, in particular, negative or poor Monte Carlo returns warn of a possible problem in the robustness of the trading system.

Posts: 2248
Joined: Wed Nov 30, 2005 12:14 pm

Return to Curve-Fitting and Other Pitfalls