Twenty20: Cricket’s great equalizer
It’s been nearly three years since international Twenty20 cricket kicked off in earnest, with the 2007 World Twenty20. And over five years have passed since the very first match, a trans-Tasman encounter in February 2005. In that time, which teams have adjusted well to the 20-over game? And which ones are still struggling to understand the format?
One thing is clear, even without looking at the data. Twenty20 levels the playing field — with the shorter matches, it’s easier for a Zimbabwe to surprise a team like Australia. But, as is my wont, I ended up running the numbers to compare a team’s ODI batting performance from 2006 to 2009, with its batting performance in Twenty20 matches over the same period.
The Duckworth-Lewis system comes in handy here. It provides a proven way to judge what the 20-over equivalent of a particular 50-over score is. The magic number is 58.6% (see this post for more info on D-L). That is, in a T20 innings, a team has 58.6% of the resources it has in a 50-over innings. That means that an ODI score of 300 is roughly equivalent to a Twenty20 score around 180, which sits well with my intuition. We can look at the equivalent Twenty20 scores for some ODI teams, and then compare to that team’s actual Twenty20 scores.
| Team | Avg ODI Score (95% interval) | Predicted T20 Range | Actual T20 Avg | ||
| Australia | 254 | 284 | 149 | 166 | 166 |
| England | 221 | 256 | 129 | 150 | 162 |
| India | 253 | 293 | 148 | 172 | 158 |
| New Zealand | 237 | 273 | 139 | 160 | 156 |
| Pakistan | 235 | 274 | 138 | 161 | 163 |
| South Africa | 251 | 298 | 147 | 175 | 169 |
| Sri Lanka | 241 | 267 | 141 | 157 | 156 |
| West Indies | 211 | 252 | 124 | 148 | 167 |
Overall, every team except India exceeds the median of the predicted T20 range — perhaps there’s something systematic about D-L underestimating the Twenty20 scores. Maybe the ODI model doesn’t map well to the T20 format. Or maybe Twenty20 specialists like Kieron Pollard are adapting to the format better than the ODI regulars.
Twenty20 has been a boost especially to the weaker ODI teams (England, West Indies). Both of them comfortably surpass the Duckworth-Lewis prediction in T20s. If you have any explanations for this, please post them in the comments.
Two possibilities I can think of:
- Duckworth-Lewis is just underestimating the prediction. One big thing the system doesn’t take into account is the strength of individual players. In a 20-over match, a single innings from a Gayle or a Pietersen can take the team to a good score, whereas ODI are slightly more of a team effort.
- Perhaps it’s just that the styles of batsmen like Pollard are suited to the wham-bam game, and can’t easily be adapted to play a 50 or 100-ball innings. This would undermine my methodology of predicting Twenty20 scores from ODI data.
Interesting! Especially since I thought India’s poor performance was due to their weak bowling and not necessary the batting.
But then again, maybe chasing huge scores due to a terrible bowling performance takes the toll on the batting.
Hi
I work with the Barmy Army and saw your post. For anyone travelling out to Australia this winter for the Ashes or just for some fun check The Barmy Army’s guide to the Ashes. It includes information on the best places to hang out, accommodation and things to see in Brisbane, Adelaide, Perth, Melbourne and Sydney. To down load your free copy, please go to http://tiny.cc/933cn
Aneesh,
That is one interesting comparison you have made, and I must add that is quite indigenious of you.
I agree with your second reasoning, and while there is no flaw with the first, I believe that statistics have a way of readjusting a sudden burst over a long period. Of course, I could be wrong.
However, there is one doubt that creeps in my mind. The ODI average that you have taken, is it for the entire duration of the existence of ODI’s? I believe this is the case, as an intuition. I believe you could have a more conclusive statistical comparison here, if you were to take the ODI average for the same years, in which T20 cricket has been played. (If that is not the case already!)
Thanks for the observation, Ankit. I did indeed use the same years for the ODI data:
“I ended up running the numbers to compare a team’s ODI batting performance from 2006 to 2009, with its batting performance in Twenty20 matches over the same period.”
I also used only data from the first innings for both matches (because obviously second innings would throw off the average), discarding any rain-shortened matches. However, I didn’t worry about the fact that the T20 players for a country could be different than the ODI players. I also didn’t account for where the matches were played (ie, the Indian team plays more matches in India, which typically has better batting pitches).
I’d be happy to share the dataset I used if you want.