When Jon Lewis, head coach of England’s women, revealed recently that team analysts had used artificial intelligence to inform selection for last summer’s Ashes, it sounded like another nail in the coffin for the traditional talent scout, armed with binoculars, notepad and gut instinct.
But the use of software run by the London-based firm PSi to simulate 250,000 match-game permutations is just part of an ongoing revolution in the way cricket coaches and selectors make their decisions.
Lewis, who concluded that off-spinner Charlie Dean should come into the T20 side – a move that played a role in England’s stirring comeback against Australia – was only too happy to let the boffins do their work.
Artificial intelligence (AI) has, of course, become an abused term, not least because of a now familiar doomsday scenario: one day soon, it will put us all out of a job. Paul Hawkins, founder of the Hawk-Eye ball-tracking technology, defines AI as ‘making computers do things that humans consider intelligent’. He adds: ‘Washing machines were once considered AI.’
And without reams of data for computers to crunch, the concept would be obsolete. Yet Hawkins has played a role in the next great wave of data collection in the English game, thanks to his iHawk GoPro camera, which since the start of last season has been attached to the jackets of umpires in county games, allowing a close-up assessment of each ball: how much it swings, seams or spins, how fast it travels, and where it passes the bat and the stumps.
Jon Lewis (above), head coach of England’s women, revealed recently that team analysts had used artificial intelligence to inform selection for last summer’s Ashes tournament
Charlie Dean proved an AI success story after sparking England’s comeback against Australia
For Stafford Murray, a former junior squash champion who now heads up the ECB’s data analysis team for England’s men, the technology has been ground-breaking.
‘We can now start giving this in-depth, contextualised information to the selectors,’ he says. ‘We can tell them: “You know what, if we put these players into certain international conditions, then the likelihood is they’re going to succeed more than if we just look at averages and traditional stats.”’
Data from iHawk has already made its presence felt. Last summer, it confirmed that the fastest bowler in the county championship was Worcestershire’s Josh Tongue, who took a five-for on Test debut against Ireland, then removed David Warner and Steve Smith twice each in the Lord’s Ashes Test.
Hampshire’s unheralded John Turner was also catapulted into international white-ball contention because of iHawk’s readings of his pace.
And while Brendon McCullum and Ben Stokes are less enamoured of number-crunching than their white-ball counterparts, Matthew Mott and Jos Buttler, the decision to take Lancashire’s uncapped spinner Tom Hartley on England’s Test tour of India was backed up by iHawk data that showed his release point was similar to the home side’s own slow left-armer Axar Patel, who had tormented England on their previous visit.
For Mott and Buttler, the use of data to inform ‘match-ups’ – the prospect of one player’s success against another – is deeply ingrained, to the extent that Freddie Wilde, the white-ball teams’ highly rated analyst, sits in on selection meetings. His research into what England can expect next month, in terms of opponents and conditions, when they defend their T20 world title in the Caribbean will be crucial.
Even so, Murray insists: ‘It’s not about making their mind up for the domain experts, the coach and captain. It’s making the conversation informed by data. We’re not data-driven or data-led: we’re data-informed.’
Central to the approach is the idea that batting and bowling averages tell only part of the story. ‘The data I love is trying to find someone’s true impact and quality,’ says Rob Key, the ECB’s managing director of men’s cricket.
Brendon McCullum and Ben Stokes are less focused on number crunching in the men’s side
‘Work out who that bowler is that might not get wickets on a particular day but actually created more pressure than anyone else, and because of them, bowlers at the other end have the wickets.
‘Or the batter who has played through the most brutal spells without maybe the return of a hundred on the day, but without it, the side would have capitulated. It’s about finding the players in domestic cricket that show the qualities required for international cricket and this type of data is beneficial.’
Two years ago at Old Trafford, Zak Crawley ground out 38 in 36 overs against a high-class South African attack, helping rescue England – one down in the series – from 43 for three. The praise he received from team-mates felt at the time like a bid to boost the confidence of a struggling player. Yet his innings, which laid the foundations for hundreds from Stokes and Ben Foakes, ticked the ‘impact’ box that is now a buzzword for Murray and Key.
If the data generated by iHawk shows, for instance, that a seamer is adept at breaking through with the old ball on a flat pitch, then their impact in a Test in Australia is likely to be greater than a seamer who thrives only when the ball is fresh and the pitch green.
Murray explains how the concept of impact can help drive selection. ‘Firstly, where appropriate, our philosophy is to help inform decisions in a predictive manner, so future-facing. We look into what will happen, having learned from what has happened. ‘Secondly, the data is often post-match, and this data is essential for debrief and reflection, but where possible we want to inform decisions and conversations in the moment, to have real-time impact during matches.’
‘Finally, it’s about using these techniques to ensure we’re gathering data on the whole pathway of players to ensure the best talent comes to the top. And that’s where the measurement of quality and impact is really big for us. It’s a massive data science project.’
He adds: ‘When we say impact, it’s boiled down to: what did that spell, or the certain action of a player, have on the likelihood of winning? How did it change the probability of success during that period? There’s a lot of maths behind that, but that’s basically what we’re doing.’
Murray laughs when the analogy is put to him, but he concedes that, broadly speaking, this approach mirrors the expected-goals calculation in football.
AI is less new than many realise. Back in 2010, England’s then analyst Nathan Leamon, a Cambridge maths graduate, devised a system called Monte Carlo, which used existing data to simulate the outcome of matches. WinViz, the win predictor devised by Leamon’s CricViz company, does something similar.
And Leamon caused a stir during a T20 series in South Africa in late 2020 when he placed numbered cards on the dressing-room balcony as an aid to Eoin Morgan’s on-field decision-making. England’s stance on data, though, is the same as it was back then: it is there to assist, not dictate.
The decision to take Tom Hartley to England’s Test tour of India was supported by iHawk data
‘Predictive analysis is going to be a big one to watch over the next six to 12 months because it’s really coming into play now as AI methodologies and technologies evolve,’ says Murray.
‘I’m not Bill Gates, but the way I see AI impacting most for us is to have more real-time information, enabling us to predict better and to measure the pathway in a more sensitive or in-depth way, to better inform selection.”
‘But you can’t replicate what the skipper can see, feel, smell, taste out on the pitch. There’s always contextual stuff going on that the numbers won’t measure. And that’s why it will never, ever replace the decisions of those people. And that’s the fun. If you could measure everything, it would be a bit boring.’
The analysts aren’t out of a job. Yet.