Sunday, 7 February 2010

2010 Nagpur Test, India vs South Africa (Win Expectancy)

I tuned into the play until lunch yesterday, and I managed to catch all the wicket-taking excitement.

I left off mostly talking about win expectancies. In this case, one needs to tinker with the model a little, since the situation is rather unique. Here's how significant the first wickets were.

After Prince was the victim of an umpiring mistake, South Africa had boosted their chances of winning by +.007. (There's a joke there, somewhere.)

Smith, however, 'popped out', as they say in baseball, to the wicketkeeper, and reduced the chance to 19 per cent, a loss of -15.

Kallis and Amlah set up 'shot' and hey-ho, took South Africa's chances of winning to 100 per cent. Except I don't believe that. The model has spit out a 'porky'. I'll come back and report after I've done a bit of figuring. I think I know how to handle this.

Meanwhile, readers with longer memories might like to recall this match.


  1. Hi Paul,

    What do you base your win expectancy calculations on? Is the database available?

    Thanks, Dave

  2. Thanks for the comment, Dave.

    The database is not available at the moment.

    The expectancy is calculated by comparing the match situation in terms of runs and wickets against selected previous test matches with similar situations, and adding in a bit of regression to mean.

  3. Thanks Paul - I assumed that's how it was calculated. I'm in the process of gathering the info myself, but not being clever enough to write a program to poll scorecards on websites I'm having to do it manually.