List member Simon replied and asked a question that crosses the minds of all systematic traders at some point or another. (Name used with permission.)
Simon:
How do you determine if a model has a statistically significant number of trades?
Dave:
My thinking has evolved on this over the years.
I used to think: “The more trades, the better.” In general, I still think this and always try to create the most trades possible in a strategy.
However, as your intuition and “strategy taste” improve over time, you can be more confident in strategies with fewer trades.
If you understand the workings of your model well, with a strong signal, you can develop the faith to trade a strategy with size even if there are a small number of trades.
There have been multiple times over the years when I’ve shared the details of one of my strategies with smart people with PhDs in statistics.
I remember one of them told me, “Yeah, there’s nothing there. It’s not statistically significant.”
This is a strategy that generates multiple 6-figure years in profits.
Yet, it’s not “statistically significant” if you apply all the fancy t-tests and such.
Here’s the thing.
The only person you need to convince that your strategy has edge is… yourself.
Specifically, your future self who is trying to find the confidence to continue trading through a system’s drawdown.
Would that person find more confidence if there were more trades or fewer in the backtest?
So, start with more trades than you think you might need, but eventually you might find that fewer trades can work when you develop more intuition.
Great question, Simon – thanks for sharing with the list.
-Dave
