May 7, 2010

Two cursed statisticians and T20!

Frank Duckworth and Tony Lewis were cursed yet again; no their crime was not doing Mathematics and Physics and Mathematics and Statistics respectively at esteemed British Universities. Their mistake seems to have been developing the Duckworth-Lewis system for one-day international cricket – which determines what aggregate the chasing team should have after every over given the wickets they have used up. In last week’s T20 WC match between England and the West Indies, the method had to be employed again and West Indies who had by then got off to a blistering start finished off the target they needed with balls to spare. England who had put up a solid overall total stood shortchanged and the weakness of the method was yet again exposed. England captain Paul Collingwood was not pleased and said as much. 

There are two ways of looking at a method like the D & L one – in a scientific (or Mathematical) context and in a sporting context. But before I do that I am tempted to relive for you all the sort of absurdity that prevailed before the D&L method came in just to show that the method, for all intent at least, was intended to ensure fair play to chasing teams and prevent absurdities like the 1992 world cup semi-finals between England and South Africa. I do not think even the average cricket fan is likely to forget the match.

South Africa needed 22 runs off 23 balls (or something) when the first of three (if I am right) mini-interruptions caused by the weather happened. After the first of these interruptions, the balls game down to 19 and after the second the number of balls came down to twelve or below that. Strangely enough, the bulk of the runs to be scored never came down! I still remember the expression of daze across faces in both English and South African dressing rooms. But the best, rather the preposterous, was yet to come. After the third delay, South Africa needed 22 runs of ONE BALL! After two decades of isolation from the sport due to Apartheid, South Africa’s first World Cup campaign ended in a comedy of errors – or at least in the tragedy of Statistics or Mathematics whatever you want to attribute the humungous succession of blunders to. I am sure the dispassionate observer as well as the average Tom Doe who came to the match laughed out loud, more from the entertainment provided by the calculation rather than the match. If the South Africans had felt miffed, or downright murderous, you could only understand!

I still do not know what went wrong: whether there was a system in place or none at all is anybody’s guess. But it was in the wake of such unbelievable stupidity on the cricket field – in a world cup semi-final of all matches – that the necessity for a sound statistical method that would be a guide to chasing teams was felt profoundly. From 1996, we have had the D/L method and to say the method has been controversial is an understatement.

I would not go into great details about what the method actually does because though it sounds complicated it is not: you can google about it to find out more. However, the method is grounded on two types of numbers regarded as resources on a cricket field for a batting team – the number of overs and number of wickets used up. Few would question the validity of considering the number of overs batted and wickets lost in the process as resources (relative to the score of the opposition) in a one-day international. Statistically too – and I am not qualified to go into statistics proper – I am sure the system has its strengths. But the problem lies in the fact that the system cannot mimic one variable in cricket, or for that matter, sport – unpredictability. It was precisely this that England found out against the West Indies as well last week.

A team that is on a rampage can suddenly fall apart; and then a team whose middle-order fails in pursuit of a modest target can still reach home in a close chase courtesy gritty efforts by tail-enders. Not for nothing is it said in cricket that no match is over till the last ball is bowled. Chases are often complicated, in other words made thrilling, by conditions of the pitch as well. In the recent IPL, a city like Bangalore often assisted the chasers because of the dew factor later in the evening whereas in Delhi, and less spectacularly in Navi Mumbai, the pitches slowed down considerably in the second half making chasing a difficult prospect. Admitted, nobody asks statisticians to come up with a system that factors in any of the things mentioned in this paragraph for it is not their business to design for the unpredictable (or simply put, what is not statistically predictable). Yet that unpredictability in fact is the very essence of any sport, the substance of all enthralling spectacles which animate the human imagination.

Viewed objectively, therefore, it seems to me that the D/L method may not be as flawed as commentators, teams, bowlers, fielders and fans make it out to be; it is not good enough, sure, but for reasons intrinsic to cricket and yes statisticians and the game’s administrators should (have) be(en) aware of it. I am not a math person, perhaps my co-blogger here may be able to help, but I cannot think of too many methods that can turn out to be tailor-made for the game. Particularly keeping in mind the fact that T20 is a fast-paced affair, one may have to give up the statistical ghost because 3 balls can tilt the game this way or that way. I do not think any algorithm can predict that. That’s the way sport goes; that’s the way life goes. Sometimes, you win games you had no business winning and other times you are shocked into submission. It is all a part and parcel of the excitement.   


vEnKy said...

You are correct srini we cannot put in place a system which exactly predicts the outcome (unless they are fixed, which they are not)

As far D/L method goes, it is the best bet for us. I've read that this method was put to test in finished matches and found out majority of the times the system was right.

There is also a blame that it gives chasing team the advantage. Its again a case of Fox and the grapes, the losing party always complaints.

Unless someone comes up with a system better than this D/L should stay.

Govind Paliath said...

In fact, there is a viable alternative. A mallu named Jayadevan came up with it in 2001 or so. Comparisons revealed it to be as good and in some cases better than D/L. Even the ICC reviewed it but choose to continue with D/L. I daresay if Jayadevan were an English bloke with decent credentials rather than an engineer from some unknown coll in an obscure part of India(obviously my personal opinion about it being obscure differs), his method would have been adopted!Before you jump the gun and accuse me of supporting one of my own read this too:

Srini said...


I was not aware of it. Thanks for the link, will check it out.

And dude, I am not someone who minds a mal supporting a mal or a Barmy supporting another so long as there is merit attached to the support. And I have hardly questioned your merit inclinations. :D


Let's just hope for the good of the game that there IS a viable alternative. As GP says, if it is Jayadevan's, then one has to implement it.