is this curve fitting? check these results out...

hi,

i have a simple system which uses 3 variables: the period of a simple moving average and the size and direction of the previous bar. it yields the following results over 18 FX pairs aggregated over 10 years (after spreads are accounted for):

number of trades: 45,273
profit factor: 1.07
sharpe ratio: 3.0

i have modified the period of the moving average as well as the size variable and the results don't differ drastically between tests (lowest PF was 1.04, highest PF was 1.09, lowest SR was 2.5, highest SR was 3.1). what is worrying me is that a low profit factor coupled with a high sharpe ratio is usually associated with curve-fitted results. however, the shape of the equity curve is too tempting...

can anyone offer advice?

Hi, on what platform did you conduct the tests? MT?

holy cow! you have found the holy grail!

as long as the EA is based on looking backwards for trade direction and not looking forward, the results should be fine. Why dont you try forward testing. You should be able to tell if the forward testing mimics backtest results.

i use a set of R scripts that i've developed over time to run my tests - a lot easier to deal with than MT if you don't use custom indicators.

Interesting! If it were MT, those results would have been clearly way off.

In your place I would turn to other backtesting platform, reliable one unlike MT, and confirm the results. Forward testing is an option, but it takes time and a few errors can occur.

According to your stats that is 250 trades per year or just over 1 trade per day. I hope this wasn't modeled on 1 day bars or higher or the results would be highly suspect. I'd check and recheck data snooping. It is very easy to unintentionally look at future data in R. A simple thing like the bar direction being shifted forward 1 period would produce results that would stay about the same even when changing the other 2 parameters.