Last time, Mateo shared how he determined that his strategy was curve-fitting.
Most traders give up on their idea at this point, but there are several ways to improve a strategy once you’ve developed it enough to recognize curve-fitting.
Here’s what to do if you find yourself in this situation, similar to Mateo’s.
There probably wasn’t a single rule you included in your strategy that caused all the curve-fitting.
More likely, it was a combination of rules that painted you into a corner, each contributing a different amount.
A good exercise here is to go back and start from scratch, that is, remove all your rules so you’re starting from a large set of unoptimized trades from your backtest.
Then, test each rule in order, watching what happens to the equity curve after you apply the rule.
Consider how robust each rule is likely to be going into the future. In other words, how predictive is the rule on your original trade signal.
At every point, be skeptical of the rule – insist that it “tell a coherent story” with your system’s trade signal.
If it doesn’t seem robust, skip it and go to the next rule.
Your mindset should be less about getting your strategy trading live at all costs, and more like “the data fooled me once already, I’m not going to let that happen again!”
Tomorrow I’ll explain the best ways to curve-fit, so you can do the opposite.
-Dave
