Of sabermetrics and sacred cows: The future of cricket statistics

Major League Baseball (MLB) is a wonderland for stat-heads and the evolution of its statistical catalogue over the last 40 years has heavily influenced other sports.

Sabermetrics have started seeping into cricket.

If that continues the rate and probability metrics commonly used to evaluate cricketers could look different soon. They probably do look different behind the scenes of professional cricket teams.

If sabermetrics are a guide, where might cricket be heading?

Baseball’s sabermetric revolution can be traced back to the 1980s and writers like Bill James who recognised that not every hit, strikeout or fielding play was of equal value, as assumed by traditional averages.

To be fair, everybody knew this. Many simply preferred to believe that performance in high leverage or ‘clutch’ situations was because of an innate quality in certain players, which is certainly easier than updating antiquated statistics to better reflect performance.

James and the many writers who followed him knew that an elite player is more likely to succeed in every game situation. The statistics needed to change, and they have.

MLB is now built on sabermetrics.

These changes entered the zeitgeist through Michael Lewis’ book Moneyball and the subsequent feature film of the same name. A much better and contemporary source is Keith Law’s Smart Baseball, a seminal reference book about how and why the revolution came, and the role of key figures like Tom Tango and Mitchel Lichtman.

To say that sabermetrics were controversial is an understatement. They engendered something akin to a holy war between traditionalists and stat-heads.

The cause célèbre for both sides was the hall of fame candidacy of Jack Morris, a good but not outstanding pitcher who happened to produce the performance of his life for Minnesota in game seven of the 1991 World Series.

There’s been no holy war in cricket, but it’s early days. Most of cricket’s advanced metrics reside in the proprietary databases of national teams and wealthy T20 franchises.

When they eventually leak into the mainstream there’ll be some revisionism, and that’s a good thing. Cricket has its own sacred cows and Jack Morrises

The value of sabermetrics is not that they’re new, it’s that they answer questions traditional stats attempt to answer but can’t and, increasingly, use technology to provide more detailed answers.

How many runs is a fielder worth?

Cricket has Andrew Symonds’ mental abacus, baseball has the MLB Statcast system, a network of at least 12 cameras at all 30 MLB stadia tracking the movements of the ball and every player on the field.

Statcast generates a wealth of data about the velocity and trajectory of the ball and players in relation to it.

When combined with Lichtman’s Ultimate Zone Rating system, it assigns every player values for outs above average and runs saved.

Cricket does actually have more than Andrew Symonds. Most of the same technology that underpins Statcast is used as part of the decision review system and the graphics about velocity, trajectory and distance presented by broadcasters. It’s almost certain professional teams are using it in some way.

The fielding averages system trumpeted by Cricket Australia but, as far as I know not widely adopted, is essentially a zone rating metric – comparing every fielder’s performance in the same set of scenarios. The impact weighting system said to be employed by Australia’s national teams sounds like a similar method.

If cricket’s zone ratings and Hawk-Eye data have been joined together, it hasn’t gone public. The official word remains that quantifying runs saved by fielders in cricket is too difficult to do with any consistency.

This is another way of saying it’s too expensive. MLB annual revenue peaked at $10.7 billion (USD) in 2019, after which the COVID-19 pandemic caused a downturn.

Quantifying the revenue generated by cricket worldwide is almost impossible, but some estimates suggested total annual revenue in the vicinity of $8 billion (USD) prior to the pandemic.

The problem is one country generates a disproportionate share, with the Indian Premier League the biggest single contributor. There’s currently no chance of a Statcast-like system at every international cricket venue or generating consistent data about runs saved by fielders in every nation.

While a zone rating system sans technology is limited, it’s not without value, and not just for evaluating fielders.

How do we liberate bowlers from butter fingers and clodhoppers?

As the Grade Cricketer wisely remarked, “cricket is a sport largely made up of individual battles between you and your teammates.”

This is especially true when it comes to the performance of bowlers, who’ve relied on teammates for more than 60 per cent of the wickets taken in test cricket history. Who knows how many chances have been squandered by feckless fielders?

Baseball’s sabermetrics have mostly focused on isolating the performance of pitchers from the variable performance of fielders.

Fielding-independent pitching adjusts a pitcher’s earned run average – the equivalent of a cricketer’s bowling average – based on an assumption that the pitcher is supported by a league average defence.

Without a run-saving metric, cricket can’t do much with bowling averages or economy rates. But it already has a much better evaluative measure for bowlers: strike rate.

If we have fielding averages – and if reports are true, weighted fielding averages – then it follows that cricket can adjust the strike rate of bowlers to a fielding-independent rate.

For example, Scott Boland took the 2021-22 Test summer by storm with 18 wickets at a strike rate of 27 over three Tests. 13 of Boland’s wickets required a teammate taking a catch and two were dropped off his bowling in Hobart.

Boland’s fielding-independent strike rate, assuming his teammates take 85 per cent of chances, is 24.

West Indies great Malcolm Marshall’s career was typical in so far as 60 per cent of his Test wickets were out caught. Assuming Marshall’s teammates – generally very good fielders and catchers – took 92% of chances, his already excellent strike rate of 46.7 would adjust to a fielding-independent rate of 44.4.

Which runs matter?

Like baseball supporters, we know that not every run is created equal. The last 150 runs of Brian Lara’s world record 400 not out against England at Antigua in 2004 were largely unnecessary and probably cost the West Indies any chance of winning.

Lara’s 153 not out against Australia at Barbados in 1999 is rightly regarded as one of the greatest innings in cricket history, made under extreme pressure in the fourth innings of the Test against the strongest bowling attack of the time.

Lara’s Test batting average makes no distinction between the two innings. The question is what to do about it. I’m not a fan of arbitrary endpoints or binary classifications.

Saying some runs are more valuable than others doesn’t mean the others are entirely without value.

In baseball, on-base percentage (OBP) and slugging percentage are widely accepted as the best traditional measures of a hitter. The former measures how many run-scoring opportunities they create, while the latter is an efficiency measure; how many bases they reach off their own bat.

The problem with OBP is that it doesn’t discriminate between a meaningless single in a 10-0 blowout and a game-winning home run in the ninth inning. Like strike rate in cricket, slugging percentage is a good measure of a hitter’s efficiency but, again, it lacks context.

Enter Tom Tango’s weighted on-base average, effectively OBP and slugging percentage rolled into one, with linear weights applied to each of the component parts. It’s not perfect, but it’s a much better way of measuring the value of every hit and walk.

It’s easy to apply a similar concept to test cricket. Runs scored close to parity are worth more than runs scored when a team is well ahead or well behind.

I’ve compared Joe Root’s excellent 2021 season, played mostly on the road in India, Sri Lanka and Australia as part of a weak England batting line-up, to Mohammad Yousuf’s record-breaking 2006 season, played mostly at home as part of a strong Pakistan team.

Root scored more runs when the tests he played in were close to parity, mainly because of his teammates. Yousuf scored more runs and scored them quicker, with the highest proportion of his runs scored between 100 and 200 runs from parity.

Who’s tipping the balance?

Who performs best in high leverage situations? The best players, of course. But anybody from Ricky Ponting to Gary Pratt can tip the balance of a Test match.

The furious debate that raged in baseball for years about players like Jack Morris and their supposed instinct for turning it on at key moments eventually led to a metric called win probability added, which is quite unlike most other sabermetrics.

Many sabermetricians consider it a courtesy to traditionalists; ‘Ok, if you insist ‘clutch’ performance is a discrete thing, here’s how you might measure it.’

Cricket already has the key component of a win probability added metric in WinViz, which calculates the probability of teams winning based on historical data and is sensitive enough to isolate the change in probability caused by individual players.

Before Pratt’s famous intervention at Trent Bridge in 2005, Australia was 2-155 in its second innings, still 104 runs behind England’s first innings of 477. At this point, Australia’s win probability was low.

For the sake of argument, let’s say it was 10 per cent when Damien Martyn punched the ball straight to Pratt and inexplicably called Ponting through for a single.

If Ponting’s subsequent dismissal reduced Australia’s win probability to 9.2 per cent, he’s debited 0.8 percentage points of win probability, while Pratt is credited with 0.8 percentage points. These credits and debits accrue across a player’s career.

Veteran slugger Albert Pujols is MLB’s active leader in win probability added, with 70.7 percentage points accrued since 2001. Ponting would undoubtedly be in positive territory as well and it’s harsh to blame him for the run out at Nottingham – it was Damien Martyn’s fault – but Ponting was out.

The moral of the story: no matter how sophisticated the metric, it’s impossible to escape the tyranny of your teammates.