January 16, 2014

How does one measure T20 proficiency?

The traditional data range of cricket statistics is far too narrow and its focus sometimes irrelevant to the requirements of the shortest format

We know Aaron Finch is a good T20 batsman but it's time the number crunchers measured factors like "run-out involvements" and "strike ratio" © Getty Images

Casting a critical eye over Network Ten's new Big Bash League coverage this summer, a number of things have struck me. First and foremost, the broadcast is a refreshing change-up from Channel 9's now relentless pursuit of vaudeville. Recent Australian greats Ricky Ponting and Adam Gilchrist have made a winning transition to the commentary box, and the dry, unscripted additions of Viv Richards have added something unique. In most senses including local ratings it has been a winner.

Still, and this isn't just down to Ten or any other T20 broadcast network, I think we're all missing a trick as fans. T20 statistics, based almost entirely on the same parameters as traditional cricket, are all but useless. Every time the broadcaster flashes them up I'm taken by how little they tell us about the shortest format. They also add nothing to the discussion surrounding individual players.

It's almost perverse really, that a game with such a long-standing fetish for numbers has done so little to adapt and shift the debate around short-form cricket. It puzzles me slightly that the rapid enthusiasts of T20 aren't taking steps to change this, because it could be central to building appeal for the game among traditionalists that goes beyond mere entertainment.

Mostly we're left to debate the merits of technical innovation in a purely physical sense: laps, chips, ramps, slow bouncers et al. The fact is - and I would call myself an engaged agnostic when it comes to T20 cricket - we should be doing better in measuring what it is that makes a T20 player great, good, average or otherwise.

How is there not yet a standardised, widely adopted stat column for dot-ball percentage for bowlers and batsmen? In many instances it would tell you more than strike rates and economy rates do. Two batsmen may score 15 runs from 15 deliveries, but one of them might have hit two sixes and then otherwise have robbed their partner of strike, killing the momentum of the innings and potentially opening the door for the bowling side.

What's clear is that the traditional data range of cricket statistics that we see most commonly is far too narrow and its focus sometimes irrelevant to the requirements of T20. Teams and coaches know this. It's why legions of analysts and number crunchers have slowly migrated into the management ranks of top teams around the world. But their intention is to decode player data to provide performance analysis and a winning edge, not to enhance the knowledge of fans.

Cricket fandom boasts a fair portion of self-described nerds but too few of the statistically minded ones are really drilling into this stuff. All of us are guilty in that sense. Maybe in the relentless procession of cricket we just don't stop long enough to consider it.

Here's a superficial example: the commonly available stats for the current Big Bash season will tell you that two of the better opening batsmen, Aaron Finch and Usman Khawaja, average 52.40 and 54.00 respectively. Finch edges Khawaja on strike rate, 148.02 to 118.24. Finch has 23 fours and 12 sixes, while Khawaja has 20 fours and has not yet cleared the rope in two fewer games.

Beyond highly subjective criteria, which basically boils down to aesthetic preferences, prejudices and an often ill-defined notion of "X factor" that pervades T20 media analysis, the average stat sheet or telecast graphic doesn't tell us much that could objectively establish who is the better opening batsman. That is always the case of cricket statistics to some degree, but the brevity of a T20 game makes it the most suitable for quantification through naked numbers than any of the three formats. The permutations and infinite variables of 50-over and Test cricket can make a liar of even the most brilliant statistician.

Would a stat as simple as dot-ball percentage reveal the chinks in Khawaja's armour that make him seem less effective to the naked eye? If that stat was segregated into ten-ball intervals it certainly would tell us he's a slow starter. "Run-out involvements" might also be of interest, as well as "strike ratio", telling us which of them turns the strike over regularly and who leaves their partner to watch on anxiously in search of a single or to calcify at the other end.

As ever, baseball can show cricket a thing or two. There are many fans of that sport who would argue that the incursion of advanced statistics and sabermetrics has taken a little joy and romance away from the game, but it's also been a boon for numbers nerds and analysts wanting to establish the cold, hard truth about a given player's capabilities. Yet instead of providing cricket and other sports with a "lightbulb" moment, the Moneyball revolution really hasn't done much more for cricket than prop up lazy journalistic analogies, a great deal of which entirely miss the point of the book and the tale it tells.

T20 cricket still awaits its own breakthrough. For a game so endlessly played, watched and analysed, we're still ridiculously low on the capture and dissemination of data that could help us understand it with greater clarity. Ed Cowan hinted at a stats revolution on these very pages two years ago, but still we plug away with our averages, our strike rates and our six counts.

T20 cricket is well past its tenth birthday now. We really should start treating it like a grown-up.

Russell Jackson is a cricket lover who blogs about sports in the present and nostalgic tense for the Guardian and the Wasted Afternoons. He tweets here

Comments have now been closed for this article

  • Harvey on January 17, 2014, 6:52 GMT

    Another interesting way to see the value of a player is their performance in wins vs losses. Take for example Ponting and Watson in tests.

    Watson averages 36.33 with the bat. In wins 37.39 and losses 30.44. With the ball, 31.83 overall, wins 27.64, losses 30.30. Ponting averaged 51.85. Wins 59.46, losses 32.83. Tendulkar and Kallis were similar to ponting.

    I can already hear people saying.."but Watson is an allrounder".

    True, but the individual performances in wins vs losses show the value of that player to the team's performance.

    The stats show that Ponting, Kallis and Tendulkar had a big influence on wins and if they didnt perform, they lost.

    It also shows that some players just skate by with mediocre performances that make little difference to wins and losses based on how they perform.

  • Sam on January 17, 2014, 6:29 GMT

    @alstar I think you are certainly on to something there. So by my reckoning, someone that averages 54 at a S/R of 148 would score those 54 runs in an average of ~36.5 balls. The expected value of those 36.5 balls would be 45.625 runs (36.5*(150/120)), therefore Finch would be producing on average a +8.325 runs per innings. This would have to be adjusted for the average extras per innings, but could certainly be a worthwhile statistic.

    Averages here would become useful, but would need to involve an adjustment removing not out scores to show how many runs a batsman contibutes to every innings they play as wickets in T20 cricket are not nearly as crucial as they are in longer forms.

    I think this is certainly a possible way of combining two statistics from yesteryear to form a statistic relevant to modern day cricket.

  • David on January 17, 2014, 4:08 GMT

    Here are two other possible batting measures:

    1. Value to the team: % of team score / % of innings balls faced (min. 10 balls).

    2. Average match situation score value (this is more complicated, and the suggested numbers are just gut feeling ... real statisticians could surely work out some more objective measures):

    Each match situation score value (MSSV) = runs x wicket down constant (WDC) x over introduced constant (OIC).

    The theory is that the openers should score more runs than anyone else - each wicket down increases the value of the runs scored (WDC). But, the later a batsman is introduced (in terms of overs), the value of his runs decreases (OIC). Runs earlier in the game provide a platform for a large score; runs late in the game may at best make a poor score respectable.

    My gut feeling constants are: WDC (proportionally increasing values): 0 = 1 1 = 1.1 2 = 1.3 3 = 1.6 4 = 2.0 5 = 2.5 6 = 3.1 7 = 3.8 8 = 4.6 9 = 5.5

    OIC: 0-25% = 1 26-50% = 0.8 51-75% = 0.6 76-100% = 0.4

  • David on January 17, 2014, 3:04 GMT

    It's not like some of the old measures aren't still useful. I'd have to question the value of Ave in such a short format, but SR is still significant ... and I don't think it really makes much difference how they got the runs - by running lots of 2s (like M Hussey) or by hitting lots of boundaries (like Finch) ... the end result is what counts.

  • Alex on January 17, 2014, 2:42 GMT

    Calculate a players value by the amount of extra runs they contribute to a team's score against Ave. In This BBL the Ave Team score is 150 (based on all the individual averages). Thus if you take into account the number of runs scored and the speed (Finch scores faster the Kwajah thus the hypothetical ave score of the other 9wkts is higher as they have longer to bat) at which they are scored I calculate that Finch is worth an extra 29 runs to his teams innings against Kwajahs 15. Maxwell is leading on 48. The same can be applied to the bowlers by calculating the average team score and then the bowlers contribution based on wickets & economy in reducing that score. Similar to Baseball's WAR or WAA statistic, except measured in runs and not wins.

  • Dummy4 on January 17, 2014, 1:14 GMT

    We need fielding stats too. Distance covered (using GPS), fumbles, runs conceded, runs saved, boundary runs conceded, shies at stumps, direct hits, dropped catches, assists etc. Should be recorded for each player but maybe more useful as a whole team or outfielder/infielder/catcher/keeper set of splits.

  • Jon on January 16, 2014, 17:54 GMT

    Finch edges srike rate? rounding about 148 to 118? Thats a runs off the bat score of 178 to 142 for twenty overs. Thats not even remotely close, anyone with half a brain can see Finch is a far better 20 over bat, even if he scored all his runs in sixes and dots and the other guy got them all in singles. The only batsmen you cant afford is a slow scoring bat who cant rotate strike, frankly a destructive six hitter not good at strike rotation like gayle you want the other guy to get singles not him, unless you are the fielding captain.

  • Jai on January 16, 2014, 16:31 GMT

    py0alb: Smart idea, but doesn't work, because the odds incorporate expectations. I.E. the odds when Dhoni needs 45 off 19 balls are ridiculously low for the situation because the market already expects him to do it....so change in odds only calculates how much better the player is then his market perception, and not his overall value.

  • Dummy4 on January 16, 2014, 16:11 GMT

    @GeckoGariack Nailed it, for me. You could probably use the expected value of one ball (following the Tweedie distribution ) as a measure for both batsmen and bowlers. Loosely, it would describe the proportion of scoring shots to balls faced/bowled * the value of the shot. The only thing missing, then, would be the expected duration of an innings. Batsman X scores at x per ball over y balls

  • LOUIS on January 16, 2014, 13:16 GMT

    @ThinkingCricket - It could well have been Gilchrist (I'm cribbing from memory). Thanks for the correction. And I agree that it isn't perfect, although I would suggest that any player averaging 9 (as per your example) across a career would be fairly obviously dismissed from consideration. But if it were used to compare established batting specialists/all-rounders, I still think there is some merit to it. Perhaps it would be more useful in ODIs, where the accumulating - high-average - top order player and the less reliable middle order destructive types can both add value.

  • No featured comments at the moment.