Walk-Forward Evaluation by Donald W Pendergast Jr
When you design a trading system, you want to see an extensive set of walk-forward tested results. Here’s how you can do that.
Every system trader wants to see a healthy set of historically backtested results that will engender a sense of confidence before they begin to consider trading a given system with real money. But the truly savvy system trader will also want to see an extensive amount of forward-tested results from the same system using previously unseen data. A trading system that passes muster in both back- and forward-test mode (also known as “in sample” and “out of sample” [OOS] modes, respectively) has a far more promising future in the real world of trading than one that simply looks good when backtested over historical data. Every serious system trader should demand that both sets of test data be made available to them before they fork over their hard-earned money for any system, no matter who the developer is.
In Figure 1 you see the forward-tested results for an emini stock index futures trading system that I developed last year. I originally tested it on six months of historical data and after fine-tuning it, I decided to run it in forward-test mode for the long haul to see if the underlying market concept of the system was truly valid.
To do so, at the close of each trading session (0930 to 1600 Eastern time, Monday through Friday) I manually entered the trades into Adaptrade’s Market System Analyzer (MSA) using the commission, slippage, contract size, and starting balance you see on the chart in Figure 1. It was been a long and grueling test, but after 229 trades in OOS mode, it appears that the system has proved itself, and despite having endured some lengthy periods of drawdown and sideways choppiness, it has gone on to make new equity highs in recent trading action.