Dave,

I liked this article, thanks!

One more thing: this optimization relies on the assumptions that only a small percentage are winners. It might be interesting to research which query is faster if everyone is a winner, and what percentage of winners is the tipping point when both queries run for the same time.

What do you think?