Expert Blog

Thursday, March 11, 2010

Is there a method in the madness?

The day is not far when IPL team owners might have to think of analytics to help pick the best players for their respective squads.

One of the Indian Premier League’s more enduring puzzles is how franchise owners choose to evaluate T20 cricketers. Why would Mumbai Indians pay Jean-Paul Duminy $950,000 when a player of similar capability, Dwayne Bravo, is only paid about $150,000? Why would Chennai Super Kings pay Matthew Hayden $375,000 and Jacob Oram $675,000?

I have tried to analyze this business but failed to ‘fit’ an adequate mathematical model to explain this behaviour… although there is indeed some method in this apparent madness.

For example, it is easy to understand why Indian players are more expensive. Every IPL team must have at least seven Indians in the playing eleven, and, therefore, at most four foreign players. From a demand-supply perspective, Indian players are scarcer and therefore end up with a higher price tag.

We have also argued in an earlier blog that all-rounders and perceived match-winners (e.g. an explosive Sehwag or Pollard) command higher prices. ‘Glamorous’ players too are paid more because they bring in more spectators and add greater sheen to promotional campaigns.

However, it would indeed appear that a very large number of players were bought on subjective criteria: competition and ‘izzat ka sawaal’ situations between bidders, eye-catching performances on the eve of the auction (Ishant Sharma before IPL1 and JP Duminy before IPL2), name called out at a fortuitous time during the auction, and poor management of the bidding kitty. There might therefore really be more madness than method in the business at the moment.

We must also remember that franchises had to bid for a three-year period; so one mistake got multiplied three-fold and looked three times sillier (e.g., buying Manoj Tiwari for $650,000).

But franchise owners are not likely to keep looking silly – remember that they are all very successful business persons. So very soon (perhaps in 2011 itself, when current contracts run out), franchise owners will start getting analytical. They are going to ask: “Can you come up with a computer program that will tell me how much a T20 player is worth? If I type ‘Michael Clarke’, will your program tell me how much should I pay him?” (even if it can’t model the Lara Bingle factor).

Soon after IPL1 ended, I worked with a group of four bright interns from Indian Statistical Institute, Kolkata, to create exactly this kind of computer program. It worked pretty well: when I typed ‘Kevin Pietersen’, I got an answer of ‘$900,000’.

We know that Pietersen was eventually sold for $1,550,000, but remember that our model only calculated his on-field worth based on his past T20 and ODI record. It seems reasonable to assume that the extra $650,000 was for the ‘Pietersen brand value’ (remember he was still England captain and had been widely appreciated for returning with the England team so soon after 26/11).

I am absolutely sure that a plethora of cricket analytics will soon invade the marketplace: the volume of T20 data is growing, and the volume of money being invested in cricket is growing even more… so this is something that is just waiting to happen! In fact I feel certain that we can spin a very useful prediction model based on the Castrol Index.

Such models will also help verify a large number of interesting hypotheses. This year, we see fast bowlers like Shane Bond and Kemar Roach being sold for upwards of $750,000 because the hypothesis doing the rounds is that fast strike bowlers who knock out the top opposition batsmen early are potential match-winners. Is this hypothesis indeed valid? A robust mathematical model could help us prove or refute it.

Monday, March 08, 2010

The importance of openers in one-day internationals

Arvind Iyengar explains why opening combinations are crucial to the success of teams in the 50-over format.

On Castrol Cricket’s Expert Fan Speak section, Ashan Vijay asked an interesting question about the growing importance of openers in ODIs. We decided to take a deeper look at the numbers and interpret what they tell us.

First, we consider ODIs played in the past 2 years. The table below outlines the matches played and Win/Loss ratios for the top 10 teams in ODIs. Australia heads the list with India in second place.

Next, we consider average and strike-rate of all openers for these sides over the past two years. The Batting Momentum,a comprehensive measure of overall batting performance which incorporates the effect of average and strike-rate has also been computed. Here’s a number for you - in the past 24 months, India has used 9 openers in total with Sehwag opening in 38 matches, Gambhir in 34, Tendulkar in 25 and six others with 10 or fewer opening appearances (kudos if you can name all six!).

The Men in Blue top the charts while Australia, the team with the best Win-Loss ratio, is in fourth place. For the stat geeks, the correlation between the Win-Loss ratio and Openers Average is 75%, and the correlation is 69% with Batting Momentum. Overall, that would seem to indicate that the openers tend to have a somewhat significant impact in determining team success, but you don’t necessarily need to have the best openers in order to win.

Of course, there is an inherent bias in some of the above numbers - India tend to play more matches in batting-friendly conditions at home compared to Australia which plays in tougher batting conditions on average, which makes the absolute numbers hard to compare. In order to truly measure the impact of openers over and above the rest of the team, we look at the average of openers with respect to the average of the rest of the team. For instance, India’s openers average 45.20 in ODIs in the past two years while the entire Indian team has averaged 40.52 in the same time span, indicating the Opener Average to Team Average ratio is 1.12. This gives us a sense of how valuable openers are to the entire squad.

Australia is last in the above table! Similar numbers with Strike-rates and Batting Momentum reveal that Sri Lanka, India and New Zealand rely most heavily on their openers while Australia, England and Bangladesh do not. This Opener/Team ratio then ends up having close to no correlation on Win-Loss ratio which seems to suggest that openers alone, relative to the strength of the entire squad, have a limited impact on a side’s overall success rate.

In short, it’s a no-brainer to say a team is likely to get better if they get better openers, all else remaining equal. But teams that rely heavily on their openers will not necessarily be more successful, Sri Lanka is a great example of that. Good starts help, and while they may be a key determinant of success in T20s, in ODIs well begun is far from half done!