Welcome Your IET account
Finch inspects his bat

Algorithms and analysis at the 2019 ICC Cricket World Cup

Image credit: Cricket World Cup, Getty Images

As the 2019 Cricket World Cup builds up to its finale in July, it seems that techno-nerds are fast becoming the game’s new anoraks.

Cricket fans of yesteryear diligently filled in their scorecards and referred to the sports pages or their Wisden Almanac if they needed to know something about the match or one of the players. These days, a new generation of fans use their smartphones to access and understand cricket’s finer points.

And this, don’t forget, is a sport with more quirks, conventions, nuances and ambiguities than a Tory party leadership contest. Even diehard fans don’t understand all the rules.

For a start, what if a cricket match is interrupted by rain?

We all know that this happens. Some of us even know why. Bowlers can’t grip the wet ball, might slip and injure themselves. A wet playing strip makes the ball do strange and unpredictable things, which makes it hard for the batter. And no good for the TV executives who rely on lots of fours and sixes to maximise their audiences and advertising revenues.

Limited-overs World Cup games have to be completed in one day. So when it rains, playing time has to be reduced and teams’ scores modified. And this is when things start to get a bit tricky.

The official International Cricket Council (ICC) system for doing this is Duckworth Lewis Stern (DLS), named after the statisticians who invented it. The precise workings are a closely guarded intellectual secret. But what we do know is that after a rain break DLS converts the number of overs remaining and the number of wickets lost into a resources remaining figure. As overs are completed or wickets fall, the remaining resources fall. From this, DLS calculates par scores, from which teams’ actual scores are revised and matches decided.

 “In machine-learning language, this is a parametric model,” says Jehangir Amjad, a cricket-mad lecturer in machine learning at Massachusetts Institute of Technology. “DLS tries to model various aspects of a game using equations and from this determine which team is in the ascendency at the time of the interruption.”

When Amjad ran tests, however, he found that DLS had a statistically significant bias in favour of the team chasing the revised target.

Amjad has his own system, which, he says, predicts not just the final score and which team is in the ascendancy, but also what might have happened during every subsequent ball, had the match continued without interruption.

“We’re trying to turn what happens in the game into concrete mathematical formulation,” Amjad says. “From this we try to work out how many runs would have been scored and wickets lost by the batting team, assuming nothing majorly different were to happen.”

This technique, Amjad says, is called robust synthetic control, and is mostly used in economics, health policy and political science. Synthetic control is a statistical method for evaluating the effects of an intervention. Amjad explains that when examining a problem this way, limited or missing data is not an insurmountable obstacle.

Sometimes, though, there’s so much rain that a cricket match has to be abandoned. It’s what spectators hate the most. Unless, of course, their team benefits.

Like in the 1992 World Cup group match where England bowled Pakistan out for a measly 74, the rain came, didn’t stop, and the one lucky point Pakistan got for the abandoned game was enough, it turned out, to get Imran Khan’s ‘cornered tigers’ through to the semi-finals. They ended up winning the semi and the final.

The effect of luck on cricket matches is one of the sport’s greatest imponderables: the source of many discussions all over the cricket-playing world, and just as many arguments.

Researchers from Indian Institute of Technology Madras Gyan Data Pvt. Ltd, a local IT start-up, have been working with cricket website ESPN Cricinfo to find a way of quantifying the effect of luck on cricket matches.

The researchers identified two major luck events for a batter – a dropped catch and an edge for four, and a third generic luck event that covers any other fortunate occurrence. That could be a bowler beating the bat often without getting a wicket, a wicket taken because of uneven or unpredictable bounce, or an umpiring error.

‘We’re trying to turn what happens in the game into concrete mathematical formulation.'

Jehangir Amjad, lecturer in machine learning, MIT

Professor Raghunathan Rengaswamy from the IIT Madras department of chemical engineering, explains that the algorithms used play out different scenarios to evaluate the impact of luck events by considering what has previously happened in past matches.

“The runs a batter scores after a dropped catch count as lucky runs,” he says. “We can also predict what might happen if the piece of luck hadn’t happened.”

Cricinfo staff then used the analytics to identify key lucky moments in this year’s Indian Premier League games. They plan to do the same at this summer’s World Cup. Another IIT Madras algorithm, which measures how valuable a performance is in context of the particular game, considers the amount of pressure each player is under when they are batting or bowling.

Rengaswamy explains that cricket’s traditional performance metrics – averages, strike rates, economy rates – often don’t tell the whole story of how important a player’s performance is, because they lack context.

To name but a few: Kapil Dev’s 11 overs for 21 runs against the mighty West Indies in the 1983 World Cup Final, Craig McDermott’s 14 off six balls in the 1987 final. And then, in the very first final, in 1975, Rohan Kanhai’s 50 for West Indies against Australia.

On paper, these performances look like nothing much. But in the context of each game, they were key influences on their side’s eventual victory. This sort of thing happens a lot.

Cricket is richer than ever, and this drives media, sponsors and tech companies, as well as cricket boards and franchises, to look for new ways of maximising their returns.

Microsoft engineers are working with former Indian superstar Anil Kumble to produce a smart bat that will transfer data about the shots a batter plays to fans’ mobile devices. The India World Cup team are using STATsports GPS trackers to measure their players’ fitness ahead of the summer tournament, although they won’t be telling anyone outside the team if Virat Kohli strains his hamstring or MS Dhoni is carrying a few extra pounds.

Analytics company NV Play offers digital ball-by-ball scoring systems for clubs in New Zealand and the UK. CricViz, a cricket intelligence app aimed at fans, uses Hawkeye data modelling to provide real-time data about player performances during games.

“Researchers are looking into putting fine accelerometers under brand stickers on bats,” says Liam Sanders, a former England and Wales Cricket Board performance analyst who now works at the English Institute of Sport. “This technology can map the orientation of the bat in three different arcs and send fans live telemetry data about where the ball struck on the bat and a player’s bat speed, which essentially is the physics of hitting the ball out of the ground and what today’s fans want to know about.”

Is cricket, with its tech and its analysis, becoming too predictable? Rengaswamy says not. “All we are doing is helping to measure areas of unpredictability,” he explains.

Sanders adds that flooding cricket coverage with real-time data helps the sport engage with younger fans. “People’s attention spans tend to be shorter these days,” he says. “Do things in the same old way and people will just switch off.”

To this end, Amjad wants to develop an algorithm that can statistically isolate turning points in a game. Rengaswamy’s next idea is to develop an algorithm that will suggest the best players in a particular team for a particular game against a particular opposition.

Horses for courses is another of those peculiarly cricketing concepts, but it doesn’t always work.

In the first-ever World Cup match, back in 1975, India packed their side with mediocre medium-pace bowlers, thinking that they’d be more effective in English conditions than the team’s world class spinners, who were left out.

England proceeded to smash 334, a huge score back then, and India’s champion batsman, Sunil Gavaskar, was so annoyed that he blocked out India’s entire 60 overs for 36 not out in protest – he made no attempt to reach England’s score. Gavaskar’s team mates were not impressed. Neither were the crowd, nor the ICC, who were trying to sell the tournament to the public as an exciting new brand of cricket.

Gavaskar later admitted that he’d edged his second ball to England’s wicket keeper Alan Knott. But the umpire, who then didn’t have the benefit of TV replays and snickometer technology, didn’t see it and gave Gavaskar not out.

Now there’s a bit of good luck that didn’t work out well for anyone.

Sign up to the E&T News e-mail to get great stories like this delivered direct to your inbox every day.

Recent articles

Info Message

We use cookies to give you the best online experience. Please let us know if you agree to all of these cookies.


Learn more about IET cookies and how to control them