Duckworth-Lewis vs Jayadevan

What happens when an ODI cricket match is interrupted by rain?

We use the Duckworth/Lewis (D/L) method to reset targets.

How does the D/L method work?

Well, no one really understands completely, although rediff.com did try its best to explain the business to its readers some six years ago.

But doesn't the D/L method keep changing?

Not really, it's essentially the same method, although there was a minor improvement in 2002, and a major improvement in 2003 (after which we need a computer to calculate the D/L targets).

And wasn't there another method to reset targets proposed by an Indian?

Yes, by V Jayadevan (VJD) in 2001. In fact rediff.com explained the VJD method, carried a detailed comparison of D/L vs VJD and concluded that the VJD method could be marginally better.

Then what went wrong?

The VJD method was never officially forwarded to the ICC before its 2001 review. So it was decided to continue with the D/L method.

And what's the latest on this D/L vs VJD question?

Well, there was another ICC review in end-2004. This time the VJD method, which too now contains several significant improvements, 'officially' made it to the finals. But it's going to be dumped again!

Why?

Because the ICC review, conducted by Englishman David Kendix, concluded that the method proposed by fellow-Englishmen Duckworth/Lewis was better.

How did it reach this conclusion?

ICC's not telling too much; it has sent some sort of communication to Jayadevan, but seems unwilling to discuss the matter further.

Is the ICC verdict valid?

It's certainly very convenient for the ICC. The D/L method is now being used for over six years. D/L in 2005 is certainly better than what it was in 2001. And both the D/L and VJD methods give equally reasonable targets in most cases. I suppose that's enough to convince the ICC!

Where does this leave Jayadevan?

At the wrong end, as usual! We aren't treating Jayadevan well. We aren't respecting his ability and effort. Here's an earnest Indian engineer from Thrissur who's come up with a remarkably good method. India's most respected science journal, Current Science, reviewed and published his method, the papers and magazines wrote about the method … but the Board of Control for Cricket in India failed to promote the VJD method with the seriousness that it deserved. And the ICC, one suspects, is trying its hardest to bury the VJD challenge to D/L. Even though the VJD method made it to the final shortlist, ICC didn't invite Jayadevan to present his method! ICC didn't really try hard enough to understand the VJD method.

I suppose the ICC thought: if the new D/L method is now even better, why rock the boat?

You have to be fair. If the D/L and VJD methods both passed ICC's eligibility criteria, both must receive a fair hearing. Cricket fans worldwide (and there are more in India than the rest of the world put together!) must be told how this evaluation was made, who won and why. You can't decide these things behind closed doors in London or Dubai.

How, indeed, would one make such a comparison?

We can consider several criteria: how does the method handle different kinds of interruptions, is the underlying mathematical model valid, is the method sufficiently flexible etc., but, finally, it's all about the target being fair and reasonable in every conceivable situation.

And who wins based on this criterion?

As we remarked earlier, it's really a close contest between D/L and VJD. In fact, the ICC itself has commented: "On the key issue of fairness of results, neither method enjoys a consistent advantage over the other." 

But there are examples from real match situations where the VJD targets indeed appear to be fairer. The ICC has chosen not to comment or react to these examples.

Where then does ICC feel that the VJD method loses out?

Among other things, they've written that D/L outperforms VJD because they are concerned about the "validity and hence robustness of the statistical analysis underpinning VJD's formulae, and, consequently, the confidence that can be ascribed to the results generated".

Isn't that kind of vague?

They're suggesting that there might be situations where the VJD method can go wrong. But they haven't come up with specific examples … they haven't established how, in such instances, D/L would indeed outperform. One expects more rigour, and detail, in a comparative study of this importance.

Where do you see the VJD method having some advantage?

Generally speaking, VJD does a little better in the first half of the innings. This is because only VJD takes into account the changed scoring pattern during the first 15 overs when field restriction rules operate. Also D/L still can't shrug off its G50 weakness; in low scoring matches, where it might be an adventure to reach 150, D/L's assumption of an average score of 235 will set an unreasonably stiff target for the chasing team. The ICC expert, one presumes, has been completely silent on the G50 rule.

The ICC is now talking of changing the first 15-overs field restriction rule.

Yes, it's likely to be a first-10 and later 5+5 overs' field restriction rule. If implemented, this rule may marginally reduce the VJD advantage. In fact, probably realizing this, ICC has explicitly rated the D/L method to be "more flexible to include changes in playing conditions in future".

So what's the big change between 2001 and 2005?

The ICC now agrees that a computer can be used to reset targets. This has allowed both D/L and VJD to make significant improvements. D/L, in particular, gains because the computer corrects its recognized weakness of setting below-average targets for teams that hold on to their wickets while chasing an above-average target.

And what's the big constant?

We continue to be unfair to Jayadevan. We ignored him in 2001, and we are set to banish him again in 2005 without a fair hearing. This is patently unjust to Jayadevan, and to his many Indian admirers.