Trading system statistically significant?

I've been curious about my results and whether it is a case of a good run of luck, or whether my results are significant.

So consider someone who has a sequence of 200 trades for example. He has a positive expectation based on his trades, but is this statistically significant? Because it is course possible that even if you were trading a losing edge, over 200 trades you could come out with profit and a positive expectation.

So what do people use to verify whether their results are significant? Or do people just assume that because it is working nicely now, everything is fine? For me personally I would like the comfort of knowing it is quite far beyond chance, which would verify that an edge is really present.

Well for a start, there's a difference between results being statistically significant, and them being forward looking. No-one's got a crystal ball.

So it depends on how exacting you want your statistical tests to be. How much do you know about statistics?

If your stop loss is 1000 points & your target is 1 point, you shouldn't read too much into 200 winning trades

You can only apply "statistical significance" to a certain extent when it comes to returns on financial markets. However, you must be careful in interpreting the results because what may seem to be statistically significant results could, in reality, be nothing more than mere luck. The primary reason for this is that most statistical models assume various probability distributions that have been known to not replicate the markets well. Of course there are some transformations one can do, but I won't get into that.

Even if you manage to compute the exact parameters required for a statistical model, there's a great chance your interpretation of the results may be incorrect.

If you're still interested in doing so, go ahead. Compute the returns of your portfolio over a given period of time. Plot a histogram to determine what sort of distribution it replicates. If you see a sort of "bell shaped" figure (which I doubt), then it should be okay to perform most statistical tests. However, if not, then take your results with a grain of salt.

Shakone said:

I've been curious about my results and whether it is a case of a good run of luck, or whether my results are significant.

So consider someone who has a sequence of 200 trades for example. He has a positive expectation based on his trades, but is this statistically significant? Because it is course possible that even if you were trading a losing edge, over 200 trades you could come out with profit and a positive expectation.

So what do people use to verify whether their results are significant? Or do people just assume that because it is working nicely now, everything is fine? For me personally I would like the comfort of knowing it is quite far beyond chance, which would verify that an edge is really present.

Click to expand...

Actualy some time ago while I was dealing with the same question, I came a cross a sample calculator, I used it since whenever i needed the make that kind of calculation.


First I assume the ''system'' Is rule based. every trade is taken strickly according to rules.

I played a little bit with your 200 trades sample . I wanted 99% accuracy
I assumed 85% win rate And I wanted to know how the next 100 trades would turn out if i placed the trades exactly ,the same rule based manner.(so i choose population size 300) confidence interval turnout to be 3.77 so you can be 99% sure that you next 100 trades winning percentage will be between 85-3.77and 85+3.77

You can find the sample calculator here: www.surveysystem.com/sscalc.htm