Yesterday, I addressed Matt’s question about the rabbit hole of trying to model perfect execution in a backtest and what not to do.
Today, I’ll share a way to include meaningful data in your backtest that’s highly correlated with execution while still keeping the super simple “one entry and one exit” backtesting approach.
It will come as no surprise to Strategy Cruncher users that the answer is to add custom columns to your backtest.
What Matt’s really trying to do is capture profit from symbols trading with less volume than most larger traders will notice.
I like this approach because the default for a lot of systematic traders is to use a hefty minimum threshold for volume in their strategies. That is, they only consider the trades in the thickest symbols.
(I bet your stomach would drop when you see how thinly traded some of the symbols are in some of my models!)
Of course, there is a limit to how thinly traded it can be, but there are many tricks you can use to trade thin names.
When you’re running a backtest, you should include columns that describe a symbol’s thickness and then you have a good idea of how low to go and a reference point to adjust to real-world trading results.
Here are some column ideas you can use to capture a symbol’s “thickness”:
- Average Volume
- Relative Volume (recent volume versus average)
- Spread (ask – bid) (raw value and in terms of ATR)
- Very recent volume
- Very recent volume as a percent of today’s volume
Notice that a lot of these columns will apply not only to execution, but could very well be predictive for your strategy more generally.
With the Strategy Cruncher, there’s no reason to limit the number of columns you add since it will tell you which ones are the most predictive. (Some users have over 300 columns in their backtests.)
Do you have any suggestions for other “thickness-based” columns? Hit reply and let me know.
-Dave
P.S. A lot of these columns are included in Trade-Ideas backtests by default, so TI users will see how to improve their strategy using the Cruncher in seconds. Click here to get started:
