Jayadevan vs Duckworth-Lewis

Ten years ago, I did a detailed comparison of the Duckworth-Lewis (D/L) and Jayadevan (VJD) methods to set targets in ODI matches interrupted by rain, bad light etc. Eventually, I concluded that both methods did the job rather well, but, if I had to pick one, I would pick the VJD method.

It is now time to revisit the comparison because it looks as though ICC is about to review the two methods again, and has in fact invited Jayadevan to Hong Kong on June 27 to present his method.

So what’s changed in the past decade? A great deal, but the principal difference is that ICC now accepts the use of computers to reset targets. This has allowed both Duckworth-Lewis and Jayadevan to significantly improve their models. This makes the two methods harder to describe, but they also now give much better results.

I have described the D/L method earlier on this blog. The key idea is to reset targets based on a combined resource percentage that considers both ‘overs remaining’ and ‘wickets in hand’. The underlying mathematics is pretty complex and I won’t even attempt to explain it here.

Jayadevan’s VJD method is equally clever. Before the D/L method was introduced around 1999, the two popular methods doing the rounds were the ‘norm’ method introduced by a South African mathematician and the ‘most productive overs’ method apparently suggested by Richie Benaud himself. Both methods had serious weaknesses (chiefly in their inability to handle one or more interruptions in the middle of an innings) although they were based on logically sound premises. The VJD method essentially combines these two methods in a way that also solves the middle-of-the-innings interruption problem. I have often marveled at Jayadevan’s ingenuity, and can only attribute it to his engineer’s sixth sense.

It would again be tedious to describe Jayadevan’s method. A very technical explanation appeared in the respected Current Science journal; my own version for dummies is also available. In brief, Jayadevan identifies two routes along which an ODI innings can progress: the ‘normal’ route, taken as long as the innings is not interrupted, and the ‘fast’ route that you must take after the interruption occurs. The VJD method describes how to reset targets by a judicious interplay of the ‘normal’ and ‘fast’ routes.

Most analysts have chosen to compare the two methods by looking at the targets they set in various ODI match interruption situations. In 90-95% of the cases, the D/L and VJD targets are very close; you could choose either! The controversy therefore centres on the last 5-10% of the ‘bad’ cases.

What are these ‘bad’ cases? One situation that bothers D/L is if Team 1 scores a lot of runs (350 or more) and Team 2 gets only 20-25 overs to bat. It is well-established that Team 2 is better off scoring at just 5 an over and ensuring that they hold on to their wickets (don’t lose more than 2). D/L’s Professional Edition (which necessarily requires a computer) solves this problem to a large extent, but this is an instance where VJD still fares better.

As someone watching D/L closely for over a decade, it is easy to see how much the method has improved … but this in turn means that the D/L method, when it was first introduced, was hardly a paragon of perfection! Even now D/L faces searching questions at the way it tries to bulldoze its 50-over method for T20 matches.

And while the D/L method flourishes with a strong ICC backing, and with everyone proclaiming that there is no better alternative, Jayadevan has soldiered on with little or no support. When ICC agreed to a computer-based system (which allowed D/L to correct one major weakness), Jayadevan too took advantage of ICC’s decision to iron out deficiencies in his own method.

I believe it is now time to give Jayadevan’s VJD method a chance at least for two years. When the VJD method came up for review at the ICC last time, it was summarily dismissed by ICC’s expert David Kendix (who has devised the ICC ODI and Test rankings). One of his remarks was: “some of the scenarios you submitted purportedly demonstrating the unfairness of D/L were, in fact, based on incorrectly calculated D/L targets”. Jayadevan was also asked if the VJD method could match the “D/L system’s flexibility with regard to any foreseeable changes in playing regulations.”

The first remark makes me angry! How is one expected to compute D/L targets when access to the D/L methodology and software is so fiercely protected? The coefficients used in the D/L equations are never revealed (apparently to “protect commercial confidentiality”; in contrast, all VJD coefficients are out there in the open) and D/L’s Professional Edition is very, very hard to obtain! And, as it turned out, Jayadevan had checked with Duckworth himself that his calculations were correct.

The second remark makes me laugh! The D/L method never had the flexibility to model powerplays; only the VJD method can model the field restrictions during the first 10 overs of an ODI (or first 6 overs of a T20 game), and, in principle, the other power plays.

I therefore await ICC’s verdict after June 27 with great interest. I would be personally delighted if the baton passes from the D/L method to the VJD method. I have no reasons to believe that this change would cause cricket any hardship, or inconvenience the cricket fan even by a teeny-weeny bit.