I use a simple money management method in my trading, which requires that I always use a stoploss, and that my trading size is always supported by a fixed amount of money. For example, I’ll only trade 1 standard lot if I have at least $10,000, and 1 mini lot with at least $1,000, and so on. I programmed this method into my EA, and did some backtests over the last three years data. The results were fantastic, and authentic as well, since I manually checked every one of the 600+ trades meticulously. I was overjoyed with the prospect of becoming a millionaire in less than 5 years.
<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" /><o> </o>
Then I divided the data into 3 parts by trading year, and did the tests again. This time I noticed sth unusual. The 1<SUP>st</SUP> year’s profit was 18 times the original capital with a relative drawdown of almost 60%, the 2<SUP>nd</SUP> year’s profit quickly dropped to only 1.5 times the original capital, the relative drawdown also dropped to around 40%, the 3<SUP>rd</SUP> yrs testing showed more or less the same result.
<o> </o>
This looked like curve fitting to me. Backtesting found the best possible set of parameters that goes with my variable lot sizes. Although the result looked fantastic, and every trade really could have happened, but the chances that it’ll happen again are indeed very slim.
<o> </o>
I’m retesting my system with 1 standard lot size. The result won’t be as fantastic at all, but I’ll have more faith in it. I’ve wasted many hours doing backtesting that only gives misleading results. I hope others won’t make the same mistake as mine.
<o> </o>
Happy trading and backtesting. J
<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" /><o> </o>
Then I divided the data into 3 parts by trading year, and did the tests again. This time I noticed sth unusual. The 1<SUP>st</SUP> year’s profit was 18 times the original capital with a relative drawdown of almost 60%, the 2<SUP>nd</SUP> year’s profit quickly dropped to only 1.5 times the original capital, the relative drawdown also dropped to around 40%, the 3<SUP>rd</SUP> yrs testing showed more or less the same result.
<o> </o>
This looked like curve fitting to me. Backtesting found the best possible set of parameters that goes with my variable lot sizes. Although the result looked fantastic, and every trade really could have happened, but the chances that it’ll happen again are indeed very slim.
<o> </o>
I’m retesting my system with 1 standard lot size. The result won’t be as fantastic at all, but I’ll have more faith in it. I’ve wasted many hours doing backtesting that only gives misleading results. I hope others won’t make the same mistake as mine.
<o> </o>
Happy trading and backtesting. J