When backtesting an EA what do you look for to qualify it as good? Win ratio? Drawdown? Profit factor?
Is a profit factor of 1.37 good? bad? ugly?
Ignored
The smothness of the curve and the drawdown in relation to the profit it made. But be careful: the backtester doesn't plot or otherwise record the drawdown that happens between two trade closings.
The curve should contain several hundred (better thousands) of trades during at least half a year, everything less than that is not statistically relevant in any way.
Often i see people posting profit curves consisting of only 5 or 10 trades. This is absolutely ridiculous. A simple coin flip could produce the same curves after just a few tries, even more so an EA with more tunable parameters (overfitting!) than there are trades in the equity curve.
The smothness of the curve and the drawdown in relation to the profit it made. But be careful: the backtester doesn't plot or otherwise record the drawdown that happens between two trade closings.
The curve should contain several hundred (better thousands) of trades during at least half a year, everything less than that is not statistically relevant in any way.
Often i see people posting profit curves consisting of only 5 or 10 trades. This is absolutely ridiculous. A simple coin flip could produce the same curves after just a few tries, even...
Ignored
Very good information 7bit, thanks.
I have done my own backtesting of a new EA I developed this month. and I did the backtest from 2006.01.01 until 2011.08.01. there were a few hundreds of trades, a acceptable smooth curve and a small drawn down (less than 20%)． However when I switch my timeframe from 2001 to 2006 it was up and down during they years and over 50% drawdown, even ending up a profitable balance in the end. so i guess for us who use EA to trade it is really important to learn the personality of a currency pair.
