So I tried posting this on /biz/, didn't really work out: >>>/biz/18755435
Maybe I can get an answer about the math here. Anyways, copy paste:
I backtested a simple strategy on a chart as far as it would go, and I got a certain winrate. The problem is, I only got 154 entries (so it's a sample of 154) over 11 634 candles. I wanna know if that number of 154 is representative of anything, statistically. And using this http://www.raosoft.com/samplesize.html, it would seem that it's representative of a population of 223 if I choose the default margin of error of 5% and a confidence level of 95%.
I understand that this is all meaningless, cause 223 is way to small of a population to say anything about a trading strategy, but humor me just about the theory please. I want to know how would margin error and confidence level affect the winrate in theory? Let's say the winrate is 60%. Am I right in interpreting these results when I say the winrate is somewhere between 55 and 65, with a certainty of 95%?
If I am, what would it mean for a hypothetical system to be tested with a 95% certainty, instead of 90 or 99.99? The difference is really not that intuitive to me, can this level somehow be translated in broadening the range of +/-5% even more?
Maybe I can get an answer about the math here. Anyways, copy paste:
I backtested a simple strategy on a chart as far as it would go, and I got a certain winrate. The problem is, I only got 154 entries (so it's a sample of 154) over 11 634 candles. I wanna know if that number of 154 is representative of anything, statistically. And using this http://www.raosoft.com/samplesize.html, it would seem that it's representative of a population of 223 if I choose the default margin of error of 5% and a confidence level of 95%.
I understand that this is all meaningless, cause 223 is way to small of a population to say anything about a trading strategy, but humor me just about the theory please. I want to know how would margin error and confidence level affect the winrate in theory? Let's say the winrate is 60%. Am I right in interpreting these results when I say the winrate is somewhere between 55 and 65, with a certainty of 95%?
If I am, what would it mean for a hypothetical system to be tested with a 95% certainty, instead of 90 or 99.99? The difference is really not that intuitive to me, can this level somehow be translated in broadening the range of +/-5% even more?
