Would you rather follow a trading strategy that is simple to understand, or one that backtests well?
I've been making a momentum-based stock selection strategy where I basically fit a bunch of weights to different parameters in order to rank stocks. So for example I might take 0.3 * [performance last month] + 0.5* [performance last year] + 0.2 * [price range] and so on. I've tried fitting different positive weights to these parameters and gotten some good results in back-testing.
Now, I've also tried allowing for negative weights which produces results that are very unuintuiative (like -7 * [parameter one] + 2 * [parameter 2] -4 * [parameter 3] ) so it's not very easy to see what the strategy is per say (although it produces comparable outputs to the one with positive weights).
Which of these strategies would you rather follow? One that you can look at and understand easier how the stocks are getting ranked, or one where it's harder to understand the ranking but the backtesting results are far better?
(2)2 Comments
Sorted by latest first Latest Oldest Best
Fitting "a bunch of weights to different parameters in order to rank stocks" is optimization which is also called curve fitting. There are software programs which will do this process and all of them have the benefit of hindsight.
When attempting to derive a trading strategy, utilize the first half of the dataset. Then apply it to the second half. If you're lucky, it will perform well on that as well. Truth be told, such a strategy is not likely to be robust because the periodicity of each security is different.
As for your specific question, the goal of a trading strategy is to make money not to derive one that is easily understandable.
Two considerations:
Past performance does not guarantee future results. Just because your model backtests well does not mean it will perform as well (or worse, or better) in the future. So it's not as sure a thing as you make it out to sound (and therefore likely not worth the extra overhead involved vs a simple investing strategy)
Complexity. Even if your model is successful in predicting future performance, if it is hard to understand then it is easy to get wrong. If your strategy is correct about "buy as many shares as you can of ABC tomorrow, hold for 18 days, then sell", that doesn't matter if you get confused and buy ABC today (before a price drop), hold for 20 days, then sell (after another price drop), or if you buy XYZ instead.
Simpler strategies may not always produce the same results, but leave far less room for error.
Terms of Use Privacy policy Contact About Cancellation policy © freshhoot.com2025 All Rights reserved.