The curves seem to look a bit deceiving, probably something in how our eyes work - switch to the 'count' tab instead of the 'percent' tab. It's much easier to see then just how much 'flatter' the founder curve is compared to the non-founder curve, and this flatness pushes up the deviation, but it pushes up the average far more and puts the confidence interval well out of range of the non-founder curve.

Judging based off just the visual curve is not a great idea as our brains are wired up for matching patterns and not for working out these kind of differences properly. It's why math is such a good idea when making decisions.

you're kidding right? Because I can make the non-founder curve just as "flat" by yanking up the scale on the percentage graph to 10,000%. The reason why it looks "flat" is because when you go to "count" the website scales both curves the same y-axis. I think you mean to say something like "kurtosis" - which is, if anything, obscured to the eye by the process of flattening by scaling.

And math is not a good idea when you make assumptions like normality of curves which are absolutely not normal. In this case, using the t-test to calculate a p value.

I think reification of poorly done math is a bigger problem than math by eye.