StatSheet.com
Help spread the word:   Facebook-logo   Twitter-logo
StatSheet.comNASCARCollege BasketballHS BasketballNBACollege FootballBuild ChartsAnalysisForums


StatSheet.com ChangeLog

When is a basketball game statistically over?
September 25, 2008

Simple answer: when Bill James says it is!

Now for the longer answer. I've had a notion of creating a stat that indicates when a college basketball game is "statistically over". This would be the point in the game where the outcome is certain regardless of future events. While doing research for this, I found a formula developed by none other than the King of Sabermetrics, Bill James. Turns out Bill is a loyal Kansas Jayhawk fan, and he wanted to determine when a lead was "safe", so he came up with this:

  • Take the number of points one team is ahead.
  • Subtract three.
  • Add a half-point if the team that is ahead has the ball, and subtract a half-point if the other team has the ball. (Numbers less than zero become zero.)
  • Square that.
  • If the result is greater than the number of seconds left in the game, the lead is safe.
Before I started using this on StatSheet I wanted to test it out. For the close to 10,000 games I have in the StatSheet database that contain game flow data (ie, scores throughout the game), I wasn't able to find one that failed the Bill James test. None of the games had a different outcome than the one computed by Bill's formula. In the aforementioned article, there was at least one game his editor was able to find that broke the Bill James test. It was none other than the magical comeback by UNC when the Heels were down by 8 with 17 seconds to go against Duke in 1974.

There may be other cases, but it is clear his formula works most of the time. But that might also suggest that his formula is too conservative. This is where I plan on evaluating and possibly changing the formula if I can find other indicators that determine a game is over sooner. Bill just went by the score, but I have access to both team and player stats (seasonal and game-by-game) which may help predict even better. More to come on that!

For now, if a game is "statistically over" sooner than the end of the game (some games are so close they aren't over early), you'll see it highlighted. Here is an example: http://statsheet.com/mcb/games/2008/04/05/kansas-84-north-carolina-66

Posted by Robbie | Permalink | Comments


Discuss this Blog Post

Go to the Discussion Forum

Cool feature. I’d tend to agree the formula is a little too conservative. I’d rather look at when a game is a 95% or 99% lock, rather than a rock-solid 100% lock.
I’ve often thought that’d be a good way to measure how much one team dominated another for the purpose of computer rankings, rather than using point spread, since the point spread can be affected by when a coach pulls his starters, etc.
KJ
Spartans Weblog

 

Now that you mention, it would be interesting to superimpose a graph of the “lead safeness” on top of the usual game flow chart. I’ll work on that.
What's New?
This blog is the place to learn about all the new stuff I'm working on for StatSheet.

About StatSheet.com

Subscribe
 Via a reader Via Email

Twitter Updates

@StatSheet


    StatFix

    Have a suggestion?
    Let me know