Bracket Voodoo

Optimized March Madness Brackets

get it
There are no images or videos added to the gallery.
Add to gallery



MakersThere are no makers yet
You need to become a Contributor to join the discussion - Find out how.
gmisra@gmisra · El Toro Labs
LOL, you should probably ask Johnny Dawkins about that. For the daily fantasy CBB players out there, we share your pain in watching Randle and Nastic go from strong plays to weak plays so quickly...
Shane WalkerHunter@shane_112 · CEO of
Automatically optimized March Madness brackets by Bracket Voodoo makes betting in your office pool easy. Developed by 2 Stanford data science graduates, I used them in 2014 to make 2 optimized pools, (1 of which won), in less than a couple minutes. They're revamped for 2015, and if @gmisra can confirm he's the maker, he can shed some light on their service. Amazingly sharp team (Metamarkets) and a very focused product perfect for March Madness :)
Shane WalkerHunter@shane_112 · CEO of
@gmisra @bradnull Again, great work guys! Common question: Are the brackets unique for every user? If unique, how do you differentiate between optimal (say return on money bet) vs popular? Are they different?
gmisra@gmisra · El Toro Labs
Great question. The set of "good brackets" isn't that big, so we can't guarantee uniqueness, but we do a couple of things to tweak your bracket to give you a unique chance of winning. But if everybody in your pool uses the site, you'll probably end up with a lot of overlap (which hurts our word-of-mouth reach, if any user acquisition experts have any tips for that problem).
Shane WalkerHunter@shane_112 · CEO of
@gmisra @bradnull 1 more question: What techniques from data science are you applying to sports?
gmisra@gmisra · El Toro Labs
Probably the most important technique we use is: good data hygiene! Through our experiments, we have found that black-box modeling techniques quickly plateau in terms of their quality. Instead, we model specific attributes of player, team, and game behavior, and then run millions of simulations to generate game outcomes. Once we have the game outcomes, we use mostly standard off-the-shelp optimization techniques (and some non-standard ones as well, but we're not ready to talk about those).