Revealed: Vital statistic that England employ to help select Test line-up
‘Weighted average’ takes into account opposition and match situation for more complete analysis The metric is now the default way by which players are filtered on the selection app
Before the Trent Bridge Test, Jos Buttler had a Test average of 37 this summer. But, by the metric preferred by England, Buttler had an average of 60. Welcome to weighted average: the statistic that England use to get a more accurate depiction of players’ performances.
Traditional averages have always been impervious to context, so a player scoring a century at home to Bangladesh would get more numerical credit than one scoring 90 in India or Australia. The notion of weighted averages is taking into account when runs are made or wickets taken, and so provide a more complete understanding of a player’s performances – and, therefore, help inform selection.
Take Buttler before Trent Bridge. He had endured a bleak start to the series against India, making 0, 1 and 24. Yet, thanks to runs scored in the earlier series against Pakistan, Buttler was still England’s leading Test batsman of the summer based on weighted average; the metric gave him credit for making runs in onerous circumstances in a summer in which Test runs have been sparse.
Weighted average suggested that, even though Buttler’s scores had been unexceptional by the normal standards of Test cricket, he had performed well considering the quality of the bowlers and conditions. After his maiden Test century at Trent Bridge, Buttler’s weighted average this summer is now 68 – still the highest of any Englishman, and way clear of his batting average this season of 47.
Weighted averages were created by England’s analysis team in 2011, as a way of revealing more than traditional averages. Initially, the weighted averages for players were just calculated by specific requests. From 2016, weighted averages – for both first-class and Test cricket – have appeared automatically in the England selectors’ app after each match. Since Ed Smith became national selector this summer, weighted averages have become the default way by which players are filtered on the selection app.
In the interviews to be England national selector this year, candidates were asked to make a presentation on whether selection is an art or science. It has always been both. The embrace of weighted averages is not at the expense of character assessment and cricketing judgment; it merely First-class batting average in past 12 months Weighted average in past 12 months Weighted Test average this year
Has consistently made runs at his home ground Canterbury, where games have generally been low-scoring and favoured the bowlers. reflects a belief that, if numbers are to be used as part of the process, it makes sense to use the best possible ones. Weighted average is not a perfect metric – but nor does it need to be. To be of use, it simply needs to be better than old-fashioned numbers.
England believe that weighted averages more fully capture what matters in a player’s performance, just as the Oakland Athletics recognised on-base percentage was a better way of measuring a player’s performance than baseball batting average. Others are also trying to use averages in a more enlightening way: Dan Weston, a professional gambler and data Test batting average this year
Weighted average reflects that Buttler has made significant contributions in onerous circumstances: 67 in the second innings against Pakistan at Lord’s, when England had been 110-6, and 80* against Pakistan at Headingley – the only score above 56 by either side. Test bowling average in past 12 months Weighted Test average in past 12 months
Weighted Test average is larger than his actual average because, going back to the second Test against the West Indies last August, bowlers have dominated the seven home Tests Broad has played. analyst who has been a consultant for first-class cricket teams, has developed a similar system.
The calculation of weighted average is actually fairly simple. The metric takes into account the quality of the batsmen or bowlers faced – and, for bowlers, the quality of the batsmen they dismiss – whether the match was high or low-scoring, and if the player was at home or away. So, a batsman making 80 will get far more credit when it is the top score in a match – as Buttler’s unbeaten innings in the second Test against Pakistan was – than when it comes during a high-scoring draw. This summer, Keaton Jennings has a Test batting average of 20, yet a weighted average of 33, showing how the numbers adjust for a summer in which ball has dominated bat – and that, for all his uncertainty, all other openers have struggled, too.
A bowler will get more credit for taking top-order wickets during a high-scoring game than tailenders during a low-scoring one. In the past two years, Adil Rashid is rated the second-best England bowler on weighted average – below James Anderson. The metric gives Rashid more credit for his performances in India at the end of 2016 than a traditional bowling average; Rashid got 23 wickets at 37 apiece, but no other England bowler took more than 10 wickets. Less flatteringly, in the past 12 months Stuart Broad has a weighted bowling average of 39, compared to an actual bowling average of 32.
Weighted averages cannot spin straw into gold – but, England believe, they can help identify who is most likely to improve the side. County Championship records are notoriously unreliable as a predictor of how players will do in international cricket. England consider weighted averages of county players to be more revealing than traditional numbers.
In county cricket in recent years, Ollie Pope is first. Then there is daylight. And then it is Joe Root and Alastair Cook. In the past year, Pope’s weighted average in first-class cricket is 71, even higher than his overall record for Surrey; in the same period, Joe Clarke, also highly regarded, has a weighted average of 37. Pope’s weighted average aligned with England’s decision to give him a Test debut after only 15 first-class games.
Joe Denly is another county cricketer who performs well on this metric – he has a first-class average of 44 in the past year, but a weighted average of 52, showing he has performed well in arduous batting conditions at Canterbury. With the ball, Lancashire seamer Tom Bailey is among the leading English county cricketers.
Test cricket is so epic in scope, and the challenges between games and even within the same ones so multifarious, that there will never be one figure that perfectly records a player’s worth. England regard the current model of weighted average as a flawed first iteration.
Nathan Leamon, England’s performance analyst, is working on revising weighted average. For batsmen, the new model could take into account where a player bats in the order – Pope’s numbers came while batting at six; for bowlers, it hopes to be able to distinguish between favourable conditions for pace or spin.
In blackjack, optimal strategy brings no guarantees of winning – it nudges players into taking decisions that, on probability, increase their chances. This is the aim of weighted averages: helping make the process of selection more refined and so increasing the chances of obtaining better results.
Expected goals now appear on Match of the Day. Perhaps, as the use of data in cricket intensifies, weighted averages will eventually be used on Test Match Special, too.