DataMasher is a bit more cryptic, but much more powerful - I'm worried about people drawing the wrong implications from the simple analyses, but it's interesting in a "data mining, damn the statistics and causality" kind of way. [snip] I'd have a tough time chosing between ThisWeKnow and DataMasher, and I really hope both stick around after the award thing is over.
Thanks! And yes, Datamasher, for one, will be here after the contest is long gone.
We were concerned about appearing statistically valid when we knew very well it wouldn't be in most cases, so we chose to encourage discovery and discussion over rigor. We have ideas of things we'd like to do for it, but it was literally built over a few weekends after hours, limiting what we could do by the contest deadline.
The cool thing about the contest is that all the applications are open source, so if a contestant can't maintain an application you like, the code will still be available for anybody who wants to perpetuate it. I'd encourage people not only to check out and vote on the finalists, but also check out the other entries that didn't make the final round--there were some good ideas in there, too.