Predictive Analytics in cricket - predicting 2012 CB series finalists

Posted by Bala Deshpande on Wed, Feb 29, 2012 @ 09:10 AM

Just when fans had given up hope, the Indian cricket team tantalizingly positions itself for a chance to make the finals against Australia. If, and this is a big if, Sri Lanka prove to be pushovers against the Aussies, that is. On the other hand, the last league match between AUS and SL may be washed out, in which case IND may have to return home empty handed.

Analytics with Cricket Statistics

So, with the available data, we asked the question: what are SL chances of beating AUS in the last game on Friday? We looked at the ODI record for SL vs AUS in AUS to build this cricket prediction story. The big picture shows that SL have a weak record against AUS in AUS. They have a historical Win/Loss ratio of 0.30 over the years, but they have improved this record to 0.47 considering all matches played since 2002.predictive analytics cricket SriLanka Australia resized 600

Predictive Analytics with Cricket Statistics

If we dug a little deeper into the more recent matches, do we see any trend? What factors result in a SL victory? This is what we did with our trusty old RapidMiner decision tree for the data from the last 10 years, which amounts to about 25 matches (so the usual caveats about overfitting may apply). But, still it is interesting to see what the decision tree points to.

rapidminer decision tree srilanka australia ODI predictions resized 600Here are the key indicators for predicting the outcome of the Sri Lanka vs Australia game:

  • If SL bat first, they must score more than 270 to ensure an almost 100% chance of winning (4 out of 4 cases)
  • If the target for SL is less than 270, then they must ensure that they dont lose their top order (3.5 wickets), because if they do, their chances of winning are 2 out of 19 (about 10%).

So what will happen? Will SL shove IND out of contention by beating AUS? Well, we now know what must happen for that outcome to come to pass.

facebook like cricfore

Tags: predictive analytics in sport, data mining with rapidminer, cricket statistics