That is exactly how MS used to have it. I think they had the numbers on the reality of how many people said yes and I suspect it was terrible. So they forced

That is exactly how MS used to have it. I think they had the numbers on the reality of how many people said yes and I suspect it was terrible. So they forced the issue. Unfortunately not for the better. I wonder if they are really wasting resources though. Would a small variance really hurt what they think is going on? Do they really need a million plus number of samplers? If I remember my stats classes, once you get past 1k or so your well into diminishing returns territory.