There was an error in this gadget

Wednesday, 10 July 2013

'Moneyball' in Cricket?


On the day that the Ashes starts I thought it was about time to do a post on Cricket. I will be writing my weekly round up of the week's sports tomorrow and will be including some thoughts more directly related to the Ashes there. 

Billy Beane. Doesn't look like Brad Pitt.
The rise of advanced statistics in cricket has been a long time in coming. Similar in many ways to cricket, baseball has been at the forefront of statistical analysis in sport since the mid 1990s. Made famous by the film ‘Moneyball’ starring Brad Pitt, baseball had its statistical revolution, known as sabermetrics, under the guidance of Sandy Alderson and Billy Beane, the general managers of the Oakland Athletics. Sabermetric principles focused on analyzing players using objective evidence measured from in-game activity. By recording every action that each player took in each game, Alderson and Beane were able to create either entirely new statistics to rate players or use the existing statistics alongside new qualifiers to give a more accurate picture.[1] In turn this allowed them to build up a far more detailed picture of which players were the most effective to a team’s success then their competitors. The principles of research and analysis allowed Beane and the Athletics to remain relevant for long periods despite having one of the lowest payrolls in baseball.[2] Given the similarities between the two games, why hasn’t cricket followed the same trend until recently and where might we begin to see similar analysis begin to creep into cricket?

For many years Cricket had been dominated by the most simplistic statistical thinking; Averages and strike rate (and additionally for bowlers – economy rate) were the only statistics taken into any sort of consideration. The problem with these statistics is that they take no account of game situation, pitch conditions, overhead conditions, quality of opposition and type of opposition (i.e., left arm spin, right arm seam etc). For example, whilst Kevin Pietersen averages a quality 49 in test cricket, his average against left arm spin is a modest 38. Stuart Broad has an overall bowling average of 31.93, which improves to 27.51 in England but balloons to 43 in Asia (including two tests against enthusiastic minnows Bangladesh). Would it not be sensible to consider picking other another batsman who doesn’t exhibit this weakness when facing quality left arm spin bowling.[3] Broad’s has struggled in the sub continent for a while now and a large enough sample size in evidence to suggest that an alternative should be found by England. Even these simple manipulations of the statistics seem to be beyond England’s selectors who seem to prefer a rigid team selection to a more squad-based system where players are picked according to their various strengths.

Alan Wells is dismissed by Curtley Ambrose.
Is it better to be picked once and dropped than
never to be picked at all?
Until recently even these basics statistics were often ignored if an international selector managed to watch a player put in a stellar performance.[4] Small sample sizes were ignored and any player on a decent run of early season form was considered for the England team with no thought given to temperament, conditions and quality of opposition. Despite players like Vaughan and Trescothick being successful additions to the England squad, a policy of picking players purely on the basis of recent good individual performances, rather than long-term weight of achievement, largely led to the shambles that was English Cricket circa 1990-2000. One cap wonders like Mike Smith, Gavin Hamilton, Simon Brown, Joey Benjamin, Neil Williams and Alan Wells were all unlucky to only be given one chance, (or maybe they were just lucky to get a chance at all), though none endured the humiliation that Ian Blackwell suffered in being dropped after his only test match for the truly inept… Liam Plunkett.[5] All were unfortunate victims of the horrific lack of consistency endemic in England’s selection; a policy that closely resembled teaching kids to swim by just chucking them into a pool and allowing those not naturally gifted to drown.

Only the arrival of Nasser Hussain, Duncan Fletcher and the new standard of professionalism that they brought with them saved English cricket from the inconsistent selection that had plagued it for so long. Alongside the consistency required to build a strong team, Fletcher, Hussain and, later, Vaughan used technology to help England improve and analyze their game. The rise in professionalism and analysis has coincided with a rise in the England team’s fortunes. England has sports analysts, Nathan Leamon for tests and Gemma Broad for ODIs, whose sole role is analyze data and come up with plans to combat opposition players.

Jimmy Anderson and Steve Harmison recently spoke on the Tuffers and Vaughan radio show and highlighted the advanced use of specific plans to individual batsmen that they faced. When specifically questioned by Mark Chapman as to how he would get out Ramnaresh Sarwan, for example, Anderson responded that Sarwan is LBW candidate early on and that he would try to “run on back into him.” To AB De Villiers the plan would be to “make him play with a straight bat” as he scores heavily with cross bat shots. 

His answers show that plans to get out different batsmen are created in two ways, firstly through analyzing any technical deficiency, (in Sarwan’s case above that he gets his front foot too far across early on and ends up playing round his pad) and secondly through Hawkeye pitch data and a batsman’s average when facing balls pitched in certain areas. De Villiers scores heavily when facing anything short pitched and so the plan is always to keep it up to the bat to force him to play straight. It must be noted that De Villiers is such a quality batsman that Anderson’s method of bowling to him is more of a way of restricting his scoring rather than targeting a weakness.

An example of the Hawkeye pitch map. This one appears to to show
Zaheer Khan's left arm seam bowling in a particular match to right
handed batsmen from both over and around the wicket. Whilst
conceding fewer runs going over the wicket, Khan has been more
successful going around the wicket as he has taken two wickets
shown by the 2 white dots.
The use of Hawkeye and similar ball tracking technologies is at the heart of the new advance in analytics in cricket. By being able to track the flight and pitch of every ball bowled in world cricket, statisticians are able to record the strike rate and average of each batsman in different pitching areas and finishing points.[6] For example analysis of the finishing points of a certain bowler could help Eoin Morgan’s shot selection outside off stump, the pitch map could allow Mitchell Johnson to locate the pitch (just a tad) more regularly and Jimmy Anderson can even end up seeing where his deliveries pitch.

Analysis and research into the game is continuing apace and its only going to accelerate in the future, so where is it headed? I feel that a squad system is more likely to become commonplace. Players will be used in a rotation policy slightly reminiscent of football. This will allow for squad depth in bowling and batting departments to cover for injuries and allow management to pick teams in a slightly more horses-for-courses way. These ideas are commonly used in county cricket with young players picked in short formats of the game to gain experience and experienced players picked in the more important county championship matches.[7] Bowlers will be on limited over counts similar to pitch counts in baseball in order to manage workloads. Other statistical elements will slowly work themselves into the game as captains and coach’s search for even more sophisticated ways of gaining an advantage.

Traditionalists fearful of the total dominance of cricket by analytics should not worry too much though. Cricket is a far more cerebral game than baseball and often a captain’s feel for the game will capture a wicket far more quickly than stubbornly sticking to a statistical plan that may not be working. Making use of the statistics in sensible ways is far harder than their creation, so whilst cricket may be appropriating statistical ideas from baseball the complexity of cricket will make it harder to totally analyze the value of each decision made.  Cricket is so complex that it will never be as comprehensively analyzed as baseball but their is definitely some work that can be done. Digital decision-making may be useful but sometimes a little analogue thinking can get you a wicket much more cheaply. Or, if you are an Australian, you can just punch someone in a bar and hope that will put him off his game.



[1] OPS –on-base plus slugging - is an example of using two old statistics combined together to form a newer, more accurate measure of a batters value to the team.
[2] For example in 2006, the Athletics finished with the 5th best record in Major League Baseball despite having the 24th lowest payroll of the 30 teams.
[3] For the record, in this case, I think they should consider it, then forget about it. KP is too good to get dropped. But they should definitely be considering things.
[4] Two famous examples where this strategy paid dividends were Marcus Trescothick and Michael Vaughan. Trescothick was famously picked for England after scoring 167 in a low-scoring match at Taunton in front of Duncan Fletcher, soon to be England coach, despite averaging in the low 30s in his career until this point. Trescothick went on to average 43.79 in tests and 37.37 in ODIs as a destructive opening batsman for England. Similarly Michael Vaughan also performed much better for England than Yorkshire with his test average standing at nearly 5 runs better than his first class average (41.44 – 36.95).
[5] Unbelievably JJ Ferris took 13/91 in the match and never played again!
[6] Only 3 of the international teams have statisticians at the moment – India, England and Australia.
[7] England are also following this route with their own T20 side. The large majority of the side is very young with only KP a regular in the test side.