I've just posted a new free tool to WhiteHatCrew.com -- a Split Test Analyzer that calculates the statistical probability that one split test candidate is really better than another.

Whenever you run a split test, you need to be sure that you've collected enough data to be confident that the results weren't caused by random chance. I've seen many a split test where it looked like one version was clearly better than the other, only to see the results swing the other way with further testing. Having a tool to statistically analyze the results and tell you whether you're ready to pick a winner or need to continue testing will save you from jumping to the wrong conclusion in such cases.

I've been using this script to analyze my own AdWords and sales page split test for quite a while, and decided this morning to make it publicly available.