In yesterday’s post on Amibroker performance, I shared some steps to minimize the time it takes to run a backtest.
It’s worth digging into the details and figuring out how to maximize your use of whatever backtesting tool you’re using.
If you can reduce the amount of time you’re twiddling your thumbs waiting for a backtest to complete, that’s more time you can spend doing the fun part – optimizing your strategy.
Part of that process is figuring out the strategy to start with.
Remember: your first backtest shouldn’t be an attempt at your “final answer” – it should be a strategy with minimal filtering to use as a starting point.
Most often, after optimizing with the Strategy Cruncher, the version I end up going live with has a fraction of the number of trades as the original backtest – sometimes less than 10%.
The backtest mentioned in yesterday’s post had over 134k trades over 5 years.
That’s on the very top end of the number of trades I’ve ever started with for a strategy.
It’s more art than science, but choosing the right starting point for a strategy ends up saving a lot of time in the long run.
The more trades in your backtest, the more memory your machine has to use to add custom columns and complete the portfolio phase to merge all the trades at the end.
Another important consideration is that with 134k trades over 5 years, there has to be a lot of noise in that starting point (noise as opposed to signal).
The higher the signal-to-noise ratio in your starting point, the better your strategy will end up being – and the easier your optimization process will be.
You’ll get better and better at choosing good starting points the more you do it – it takes practice.
-Dave
