So for you who know statistics as I am bit rusty. How do you calculate when your strategy’s sample size is statistically significant?
it’s a complicated subject (as you know!)
in your case, with a 41% win-rate, to have 95% confidence that you’ll detect any swing of 20% higher or lower in the win-rate (i.e. to be 95% confident that your win-rate is actually, reliably, something between 33% and 49%) you’ll need a minimum of 310 samples in the data-set
with a win-rate closer to 50% than that, you would need fewer samples to achieve that 95% confidence-level
with a win-rate further away from 50% than that (i.e. lower) you will need exponentially more
(you can probably find a free PDF copy floating about online of the Michael Harris book i mentioned, in which all of this is explained, if that helps)
Can you give me the formula where 310 samples are calculated?
Degree of freedom