Predictive analytics in cricket: IND-SA group B match preview
Posted on Thu, Mar 10, 2011 @ 09:22 AM
It seems strange that this encounter between the two
strongest teams in group B does not really impact advancement to knockout. However it is still a very interesting match up and for both teams it will remain the first true test of their consistency in this tournament.
What makes it more exciting is that both India and South Africa have been challenged by the next best team in the group - England. While India have held their fort (if you can call a Tie holding fort), SA have lived up to their choke reputation by collapsing from 68 for 1 to a defeat while chasing less than 200.
Looking at historical records between IND-SA makes it even more challenging to predict a clear winner for the match, although SA have a slight edge having won 14 of the last 26 matches. However, are there any trends within these 26 games which we can extract using predictive analytics? Do these trends make sense given the small sample size? These are all good questions.
So here is what our predictive analytics exercise using cricket statistics (and RapidMiner) between the two teams tells us:
1. SA will win if they restrict Indian batting from building up more than two partnerships each at least 34 runs
2. IND has a very good chance of winning if there are at least two good partnerships worth more than 34 runs ONLY if these partnerships happen early in the batting order - preferably before the first 6 wickets.
Any Indian partnership that comes 7th wicket or later (Yusuf Pathan driven?) is pretty much useless as South Africa would have apparently sealed the match!
How good are these trend predictions? Check out the previous two predictions here and here - and judge for yourself!
Download the data set for FREE by clicking below (and if interested also sign up for visTASC, our FREE tool for choosing the right analytics technique for a business problem visTASC) below.
