so in you mind if I play a fighter with +8 to hit and 1d8+9 damage, and you play a fighter with +8 to hit and 1d8+9 damage after 1 fight we will be equaly effective...
or can that variance you mentioned matter?
What?
Of course the variance matters. It just...
Look. Just...
It seems to me that you are saying this: that statistics may make predictions, but that variance is so high and the sample size so small that these predictions are not useful in terms of people's actual experiences at the table.
This is not true.
First of all, the variance is not as high as you think. It doesn't take hundreds of attack rolls to start matching the predictions pretty well, it takes a few dozen.
Second, even if the die rolls at the table don't exactly match the predictions, the predictions are still useful and important because they tells us the shape of what we should expect to see as the die rolls vary.
Let me give you an example- lets say we had a raffle. There are thirty ENWorlders in the raffle, and everyone has one ticket, except me. I have two tickets. So everyone has a 1/31 chance of winning the raffle, except me- I have a 2/31 chance.
Now it would take us dozens of raffles to start to show that I have twice the chance of winning as the rest of you have. And there's a fair chance that some lucky guy will get drawn two or three times, totally obscuring my better odds.
But I STILL have twice the chance that you all do! Two tickets are STILL twice as good as one ticket! Even if we only do one raffle, and there's no chance at all of demonstrating experimentally the accuracy of the prediction that I have a 2/31ths chance of winning, that's still what I had.
The predictions don't stop mattering because the sample size is small. 2 is still twice 1.