Nashua New Hampshire
Posts: 32 since Sep 2011
Thanks Given: 6
Thanks Received: 26
|
I would like to find out what other traders feel is the best singular statistic for evaluating performance? There is no "right" answer here, it is partially a subjective decision. I'll start with one approach I have found to be very useful:
- Take the annual Net Profit, including a reasonable baseline commission, and divide by the draw-down. The higher the number the better the system. Here is an example:
1) Net Profit = $141,000 over 3 years = $47,000 annual Net Profit. Included $8 round-trip commission.
2) Draw-down = $12,000 historically.
Performance = 47,000 / 12,000 = 3.92. A value above 2 is a good system, so this is an exceptionally good system.
Here is my reasoning:
1) Draw-down gives the overall risk better than anything else. It is where a system had its greatest streak of losses in a row. Using Monte Carlo simulations you find that above 90% of the time 2 X draw-down is the absolute worst case. Therefore risk is directly proportional to draw-down.
2) Annual net profit averages the years of backtesting into one number. So if a system has a broad deviation over years it is spread over the average net profit. (I also always make sure there are small deviations for annual net profit year over year).
This ratio provides a good single measure for reward vs. risk of a system.
|