5 trades/year doesn't seen to be very statistically significant for me.

Hi Mithrandir

I would be a little bit concerned about the low number of trades. Is it enough data?

Thanks for your inputs mates. I used oversampling, with NumSampleCycles=4 and 6 but the number of trades per year doesn't increase. I attached the reports. How can I obtain a statistically significant result? Or is it filtering the bad trades for a specific period a tradeoff with getting enough trades for the result to be statistically significant?

I think I am more or less curve fitting but it is a 'good' curve fitting because I am adapting to the volatility of a specific period of time (in particular one which has low volatility thus not good for trend strategies) . Even though as an engineering student I am afraid of a not statistically significant result as you

Hey Mith

I'm an absolute newbie on the stats/testing side of things, however I would probably want to see how the revisions impact the previously profitable 2009 onwards period. Whatever you do, can't reduce profitability of this more recent period. I myself would be tempted to accept that the markets have changed, and a previously unprofitable strategy is now profitable, rather than attempting to make this strategy profitable for that period. If you take this path, then all you need is a way to identify conditions have changed enough for the strategy not to work, and stop the system, or alternatively see if you can train the system out of taking any or many trades during that period, so long as profitable periods stay profitable.

I am going to finish here by again stating that I know nothing, and I am sure that there are more knowledgeable people to comment on this.