Saturday, November 19, 2016

New T-Rank Stat

I added a new stat to the schedule section of each team's "T-Page," which I'm just calling "Team +/-" in lieu of something else occurring to me.  This is a basketball version of a football stat, as far as I know invented / popularized by Chase Stuart at, where they call the stat "Game Script." Noted Badger fan Nuclear Badger suggested the idea of bringing this stat to basketball, and we got it done.

In a nutshell, the stat is a team's average lead or deficit over the course of a game. For example, the Badgers' loss to Creighton the other day was a back-and-forth affair for much of the game, so even though the final score was Creighton by 12, the Badgers' average deficit over the course of the game was actually just 1.7 points.

On the other hand, Marquette lost to Pitt yesterday despite having a significant lead for most of the game. So, even though they lost by 3 in the end, their Team +/- for the game was actually significantly positive, at 2.6.

So this is kind of a cool "game control" stat that sheds some more light on the result than just the final score.

Perhaps as importantly, doing this stat has made me finally get around to start harvesting play-by-play information, which should open up the possibility to do some other interesting stuff as well. The only problem with play-by-play is that it is somewhat unreliable and it isn't available for every game. (For example, note that I currently don't have PBP for the Michigan / Marquette game above. But still, it'll be nice to have.


  1. A couple of other things on the stat - one of the things that Chase monitors for football are %run/pass - typically, the higher the game script, the higher the run%. The equivalent for BB is probably 3PTA%, but that also is influenced by what defense is being played (zone vs man). My guess is the more negative the game script, the higher the 3pta%.

    Another thing to potentially track is how many teams win with a negative game score - - just to show come back-ish type wins.

    1. So far the lowest game script in a win is Northern Iowa versus Oklahoma, -6.9, versus Oklahoma.

    2. That seems like it is a pretty significant comeback - you must have trailed for a long while to end up -6.9. That said, I think more data would be useful to know just how low that is.

    3. Can take a look at the progression through the game here. (Third column shows +/- for the second team). Actually looking at this helped me see there was a slight bug in the code, which I've fixed. Actual result now is -7.0

  2. Is +/- time-weighted? Or is it simply the sum of all leads/(deficits) divided by the total number of different leads/(deficits)?