Alright, this is just for fun because it’s March. I’ve set up a tourney challenge on ESPN that you can get into with this link. In addition, I’m going to give away a yearlong subscription to RotoViz to whoever wins opening weekend. Got that? So you don’t have to win the whole bracket, just be in first place after opening weekend (what used to be called 1st/2nd round and now is idiotically called 2nd/3rd round). This also opens up the possibility that more than one person can win a subscription, as ties are more likely early in the tournament.
To help you pick your games, if you’re interested I also have a March Madness app below that will let you throw any two teams in and get a predicted point spread for a neutral court game based on looking at historically similar games. More work went into this app than I’d like to admit, but by the time it was done it was actually much simpler than when it started. When searching for historically similar games, it heavily weights towards finding similar games among teams that had similar Generic Point Spreads. Generic Point Spread is probably a little like Simple Rating System in that it just creates a huge linear model with every game point differential as a dependent variable and every team as an independent variable and then it finds the Generic Point Spread that reduces the squared errors. There are some other variables that are contained in the similar game search, like strength of schedule, pace, offensive rebound percentage, turnover margin, and effective field goal percentage, but again it’s heavily weighted towards Generic Point Spread. Even though the similarity search is weighted towards Generic Point Spread, some schools consistently play the same type of basketball (even across seasons) that they end up being similar to themselves (i.e. Duke).
If you wanted to simplyifyhow you think of the app, you might think of it as a more efficient version of the mental process that goes “How did Team X do against teams of similar quality to Team Y?” Then the only thing we’re doing in addition to that is adding in teams that are similar to Team X and Y to the mix.
Feel free to use this app or any of the other models on the internet. I will say that my model doesn’t like Indiana nearly as much as the other stuff I’ve seen on the internet. Losing three out of their last six games really ended up hurting Indiana’s Generic Point Spread in my model. But the other thing you should keep in mind is that college basketball is a high variance game. We know this intuitively based on the upsets that occur every March, but when you look at the historic game comparisons this will be even more apparent.
RotoViz is not responsible for however you might choose to use the point spread feature of this tool.
Because I’ll need your email in order to set up your subscription, if you’re in the lead after opening weekend, you’ll have to shoot me an email so I can add you to the subscriber list. My email is fantasydouche (at) gmail (dot) com.