Some people have complained that there isn't enough analysis/strategy on this blog, I can oblige:
The following is a graph of all D-1 college basketball games played in the last 3 years:
The X-axis represents the difference in points each Home Team is supposed to beat the Away Team by. This is predicted using data on pace(# of possessions/game) and efficiency(points/per possession) - putting the two together you get points/game for each team.
The Y-axis represents how the Away Team actually performed in comparison to these expectations.
Without knowing shit about CBB, you can see a couple of things:
1. This model has all data points spread around in a pretty even circle. ie, it is accurate and will be contributing highly to my sundance steak/new wardrobe/we still need a surround sound system in the house/thai hooker fund.
2. The circle is off-centered, to the right, showing the worth of Home Court Advantage. In this case, the circle is shifted ~5.1 points. Should Coach Dawkins offer to play all his home games at somewhere random in exchange for being spotted 5 pts? It doesn't matter. They are equal. Stanford would be pretty pissed off it couldn't sell tickets though.
Credit to Keith for helping me make graph.
Edit: To clarify, when people try to predict things in sports they are trying to come up with the fair line, where the distribution is even on all sides.(ie. When I say "The Blazers should win by 11 tonight", I am really saying, ~50% of the time they will win by <11, ~50% the will win >11, in a roughly symmetric distribution on both sides). This is why the circle is important, because it does just that. In contrast, here is an example of a model that is not predictive:
No comments:
Post a Comment