Since I had a lot of computational time over the night, I setup the optimizer to optimize my strategy exhaustively over the longest period of historical data that I could (01/01/2010 – 01/06/2013) in hope that I could find some way to create an algorithm that would perform well in the loss period identified above and still perform well in the good times.
I let the optimizer run for almost 14,000 tests and then was overwhelmed by the results. What am I actually looking for here?
Is it the strat that makes the most money? Well not really if it almost went bust at some point.
Ok so the strat that has the max lowest equity? Well, it has to make enough money to be worthwhile. What about the strat with the highest win/loss ratio?
So of course, I turned to the internet for research.
Some important metrics apparently, other than profit of course, are:
- Win/Loss Ratio = # Winning trades / # Losing trades … indicates if the trades are likely to win
- Profit Factor = Total Profit / Total Loss … indicates if this strategy makes profit or not.
- Profit Expectancy = Total Profit / # Profit Trades… i.e. average Profit per trade … indicates the expected profit per trade.
- Expectancy ratio = Profit Expectancy / Loss Expectancy
However, when looking into expectancy, I didn’t find it very useful. The optimized results that had the highest expectancy all failed to make money.