It's the Numbers that Count in an Algo

The reports below are for another one of my algos. This one is for the DAX using 18 minute bars,  It was created using the last 3 years of data.

Interesting stats in the extract below, the last almost 6 months of trading,  is that has a nice average trade profit that can handle any slippage that is likely to occur, the Profit Factor is robust and the win rate is unlikely to make me turn off the algo after 3 losing trades in a row. Also, the drawdown is quite acceptable. When I look at the month by month stats I see that the profits are spread evenly over each of the almost 6 months.

 

However, all this is not enough. Its not hard to create a set of stats that show a great historical result. It's easy! Just curve fit the algo to the data and it looks great. What I do after creating an algo is the robustness tests, more than seven of them. The final one is the walk forward analysis. Here is the WFA summary that was the final PASS to go live.

I use three platforms for my algo trading: NinjaTrader, MultiCharts and TradeStation. All are great platforms but have some differences in how they do the various testing and optimizations.