You may have seen my predictions on Facebook for individual matches over the recently concluded T20I World Cup.
Starting with the Super 12, there were 30 league matches and 3 others (semi-finals and final). 4 league matches had no result, so there were 29 matches which had results.
I used a simple predictive tool, which was the latest ICC ranking available before the day of the match.
This gave 22 correct predictions out of 29, or 75.9 % accuracy. This included correct predictions for the NZ v Pak semi-final and the final.
Better than flipping a coin, anyway. And this does not involve making complicated estimates of the relative strengths of players and teams. As in predictions of economic statistics, the idea is to get a reasonably correct estimate based on the least number of inputs. That way we can say that this prediction method is successful, provided that one keeps track of the ICC rankings which are updated every one or two days in a tournament like this.
An attempt was made to predict whether a result would be close or one-sided depending on the difference in rankings and ranking points. This usually gave wrong predictions, but may be more useful in 50-over matches.
See how this system works at the next World Cup in 2023.
abn397 