Winning the toss on a cloudy morning

Saturday, May 15, 2010

Our expert throws light on the pros and cons of the Duckworth-Lewis method in the T20 format of cricket.

In an ICC World Twenty20 match on May 3, England scored an impressive 191 in 20 overs, but still lost to West Indies who scored 60 in 6 overs to win by the Duckworth-Lewis (D/L) method.

West Indies won because their captain Chris Gayle did two things right: (a) he chose to field after winning the toss on a wet morning and (b) he launched a batting blitzkrieg when he came out to bat because he realized that rain was imminent.

Gayle chose (a) because it is now common knowledge that the truncated D/L method for T20 significantly favours the team batting second, and he chose (b) because things turn even more favourable for the team chasing if they have scored a few quick runs (preferably without losing wickets) before the rain comes down.

Imagine that it had rained before WI started the chase. WI would then have had to score 66 in 6 overs to defeat England’s mighty total of 191. It is a lot easier to hit 11 per over with batsmen free to throw their bats … than average 9.55 over 20 overs. That’s why the D/L verdict appeared unfair.

As it turned out, WI came out to bat, scored 30 for no loss in 14 balls … and then the rain came! This lowered the WI winning target further: they now had to score only 60 in 6 overs to win.

The trouble with the truncated D/L method (that assumes that the evolution of a T20 match is like the last 20 overs of a 50-over ODI match) is that the fine balance between ‘overs remaining’ and ‘wickets in hand’ breaks down, and the ‘overs remaining’ resource becomes much more valuable than the ‘wickets in hand’ resource. And if the team chasing has ridiculously few overs to bat (5 or 6, as in that Eng-WI match), then the D/L method sets a very low target because it attaches most of the value to ‘overs remaining’ ignoring ‘wickets in hand’.

What is the way out? The sensible recommendation would be to restrict the minimum overs required for a result to 10 (or even 8). If that is not possible, then we should NOT truncate the D/L curves; instead we should pretend that a T20 match evolves exactly like an ODI match but at a much faster pace.

By suitably ‘treating’ the full D/L curves, it would be possible to set a target of about 75 runs in 6 overs if a team is chasing 191. This seems far more realistic.

There is a second problem with the D/L method for T20: it does not factor in the advantage of field restrictions in the first 6 overs. In 50-over ODI matches, this advantage is compensated because of the greater likelihood of losing ‘valuable’ wickets. But, in the truncated D/L model, wickets are far less expensive and the chasing team is better served by going slam-bang even if a wicket or two falls in the process.

A reasonable solution to the D/L method’s 20-over dilemma might therefore be to come up with redrawn D/L curves, and, since the new targets might be rather stiff – especially if the match has a severe curtailment – we might suggest that the 6-over field restrictions will apply irrespective of the duration of the interruption.

With this solution, the decision to bat second would be based on genuine cricketing considerations and not irrelevant meteorological ones.

In-form Aussies have something to worry about

Sunday, May 09, 2010

The way Australia played (and won against) India in the Super Eights was superb. Our expert delves into the game and wonders if the Aussies will peak too early.

Australia dismantled the Men in Blue in what was billed as a clash of the titans. Take away Rohit Sharma and the rest of the Indian batsmen managed just 46 runs between them; Harbhajan Singh and Mr. Extras were the only ones to reach double figures. It’s a promising sign for Australia who are now, by popular consensus, the favorites to win it all this year. However, results from recent T20 tournaments indicate that Australia might be in danger of peaking too early.

One caveat to begin with - a team may not have to worry about winning consistently if there is a wide talent gap between teams. For instance, if Australia played the Netherlands in 10 straight games, it wouldn’t be a surprise if Australia won them all. Similarly, in the MAGS era (that’s the Matthew- Adam-Glenn-Shane era, you know who I’m talking about!) the talent gap between the Aussies and the rest of the world was fairly high and Australia were expected to win a significant percentage of Test and ODI matches they played.  In today’s world, there is little disparity between the top eight international teams, particularly in T20s in which results are unpredictable. Even in the IPL, there isn’t much difference in skill levels between the eight franchises. And it’s precisely in these types of competitions with roughly equal teams that peaking too early has been a problem.

Case in point - the World T20 2009 won by Pakistan. The team that peaked too early - South Africa, who had three consecutive dominant displays heading into the semis. Sure they were undefeated, but let’s get deeper into the numbers - they beat New Zealand by just 1 run in their second match and had a poor outing with a bat at the start of the tournament. They really started to peak in the Super 8 stages, beating England by 7 wickets, the West Indies by 20 runs and India by 12 runs - all of which were comfortable victories. And then they lost to Pakistan - the Proteas peaked too early. Sri Lanka suffered a similar fate in the final right after two emphatic wins - 48 run victory against New Zealand and a 57 run win against the West Indies.

In general, the odds of playing consistently well in over four consecutive matches against competitive oppositions is extremely hard and the team Castrol Index (CI)) does a good job of illustrating that for IPL 3. The average CI for a team across all matches was 133 for the tournament. That number by itself may be hard to understand, the key here is to understand a team’s performance relative to the average. For instance, a great team performance will have lead to a CI score of about 170 or more- a relative score of 1.25 or more with respect to the133-baseline. A poor performance results in a CI score of about 100 or less, a relative score of 0.75 or less

In IPL3, there were 9 dominant ‘streaks’ of 3 or more matches - a streak being number of consecutive matches with a CI score above 133 or a relative score above  1. Of these, five were 3 match streaks and four were 4 match streaks- there wasn’t a single streak longer than 4 matches! And what happened at the end of streaks- teams tended to come up with a poor performance and in some cases that started a losing trend. The two charts below highlight the most significant streaks (the ones in which teams were truly dominant and had an on average relative score of about 1.2 across the streak)

Deccan, Mumbai and Rajasthan all dropped after a three match streak

Bangalore and Delhi peaked too early

These graphs indicate that all these ‘hot teams’ or ‘momentum teams’ or ‘in-form teams’, whatever you want to call them, all had a big loss at some point. The Delhi Daredevils are a perfect example of a team peaking too early and losing steam when it mattered - they won their 6th match in the tournament by 17 runs, the next three by 40 runs, 67 runs and 37 runs respectively. And then they lost to KKR by 14 runs which started a 3 match losing streak and an eventual knock out against Deccan. Similar story with the other teams, though some had their setbacks early enough to bounce back.

The implication for Australia - they better be careful about getting hot too soon and fizzling out at the end. The odds suggest they are probably better off losing one of their early matches or at least being tested early. A danger of dominating too much too early is that you never really get pushed and complacency may set in, which can be a real danger in a competition of relative equals. That’s probably the silver lining for Indian fans after the trashing on Friday - at least they aren’t peaking too early!