Why T20 is a disfigured caricature of cricket

I have previously argued that T20 is not really cricket. That was based, in part, on the simple observation that giving teams ten wickets to play with over 20 overs against nine fielders and five bowlers (each limited to four overs each) skews the contest hopelessly in favour of the bat. I used the example of Chris Martin (Test average 2.36), who was arguably the worst batsman of his generation. Martin was dismissed 52 times in 615 deliveries in 104 Test innings. He faced 11.8 deliveries per dismissal. Eleven Chris Martins would have fallen two balls short of surviving 20 overs on average.

Typically, the counterpoint to this view has been to simply say that T20 is "different". The more obvious differences lie in the creation of the World Twenty20, played twice as often as the 50-over World Cup; and the purchase of the biggest players and the biggest commentators, and the presence of the biggest sponsors, in the IPL. These aspects of T20 are supposed to grant it the legitimacy of being cricket.

Today, when most people watch T20, I'd wager that they think they are watching cricket. The more sophisticated T20 aficionados may concede that T20 is cricket in miniature. This is at least an attempt to affirmatively describe how it is different. These arguments also come with the usual questioning of motives and suggestions to the effect that one is free to stay away if one doesn't like T20. Let's set that nonsense aside, because cricket is important and wonderful.

In Test cricket, there are measures of merit that hold true irrespective of the outcome. It is possible to say what a good spell or a good ball is. It is therefore possible to say when a batsman has had some good fortune, or when a bowler has not. I'm not talking about umpiring decisions, but rather about things like a bowler beating the bat consistently on a given day, or another bowler getting a few cheap wickets. These measures of merit hold true irrespective of the outcome, and do so irrespective of the individual fan's belief in those merits, because in the long run, the former bowler is likely to end up with a superior record to the latter bowler. Line, length, pace, control, footwork, variation, mystery balls, the limits of mystery balls, stamina (mental as well as physical), concentration - all these things are discernible in Test cricket without as much as a glance at the scoreboard. A bowler can deliver a horrible over, full of half-volleys and long hops, that ends up as a maiden simply because the batsman kept hitting his cover drives and square cuts to two brilliant fielders. But it's still a horrible over.

I've been trying to work out what the measures of merit in T20 are, how we might tell a good T20 player from a bad one without looking at the numbers. So far, I've concluded that this is not possible in T20 because it features an overwhelmingly skewed contest between bat and ball. The argument that T20 is cricket in miniature is an example of the scaling fallacy. It's like saying we could have giant, human-sized bugs simply by scaling them up in the three cardinal directions. We couldn't, because their exoskeletons would be just too heavy, assuming that other things (like gravity) remain constant. Scaling involves unpredictable shape-shifting and occurs along many dimensions, not all of which are known at the outset.

T20 is not cricket in miniature, just as ODI cricket is not cricket in miniature. Yet, there are evidently some measures of merit in ODI cricket that can be identified irrespective of the outcome. For one thing, given the length of an ODI innings, batsmen have to construct innings. They can't take chances on the third ball. Batsmen also have to overcome different conditions - the ball gets old, it reverses (or it used to before they started using two balls), the pitch wears during the course of 100 overs. The cost of getting out is significant, given that teams have ten wickets over 50 overs and would like their batsmen to face most, if not all, of those overs. Since batsmen have to construct innings, a good line and length is possible beyond the perfect yorker. There are substantial periods in ODIs where fielding teams are not resigned to conceding one run every time the batsman doesn't miss the ball. A somewhat even contest is possible most of the time. This is not evident in T20.

I will provide some of the numerical evidence that forms part of the basis of my conclusion.

As a start, I'm going to compare T20 to Tests and ODIs. I looked at two measures. First, how often does a batsman who scores a particular number of runs (or runs in a given range) end up winning, losing or drawing games (or simply winning or losing them)? Second, how often does a bowler who takes a particular number of wickets end up winning losing or drawing games?

Test cricket
Test cricket is designed to be won by bowlers. A player who takes seven wickets in a Test is more likely to end up on the winning side than a player who scores 200. Taking wickets is a sure route to winning Tests. Bowlers who take seven, eight or nine wickets in a Test match end up winning nearly 60% of the time and avoid defeat about 80% of the time. These bowlers are three times more likely to win than they are to draw. Batsmen who score 150 to 199 runs in a Test are also likely to win or draw about 80% (but the likelihood of the result being a win is about equal to that of it being a draw). This holds true across a range of match aggregates for individual batsmen. Taking more wickets makes both draws and losses less likely, while scoring more runs only makes losses less likely.

ODI cricket
Things are different in ODIs. Batsmen have a greater match-winning role. This, as I will show later, also means a diminished role for bowlers. Getting bigger scores makes ending up on the winning side more likely, as does taking more wickets. The general shape of the contest, when limited to 45, 50, 55 or 60 overs at the outset, shifts slightly in favour of the bat. Playing substantial innings matters, as does taking wickets. There is a potential match-winning role for both batsman and bowler in this format.

T20
Bowlers who take three wickets in a T20 match end up on the winning side more often than bowlers who take zero, one or two wickets, and less often than bowlers who take four or more wickets. On the batting side, things are similar to ODI cricket as well, but the correlation between scores and win percentage is not as clear in T20 as it is in ODI.

Each interval in the ODI and T20 batting chart contains roughly an equal proportion of scores (1/8th of all scores in T20, 1/10th in ODIs). In each format, at least three-quarters of the scores made by batsmen are such that they are more likely to occur in defeats than wins. This is perhaps a useful threshold beyond which we can say that an innings has been built. The data for T20 cricket is noisier than the data for ODI cricket simply because fewer T20 internationals have been played.

But the threshold score - the lowest individual score that appears in wins more often than in defeats, such that almost all higher scores also appear in wins more often than in defeats - for T20 cricket is currently 29. In ODI cricket it is 42.

The overall economy rate in ODI cricket in the T20 era (from February 17, 2005 onwards) is 5.05, while the economy rate in T20 is 7.6. In order to reach the threshold score, an ODI batsman has to survive 50 balls on average without getting out, while the T20 batsman has to survive 23 balls on average, assuming that each scores at the average rate. In each case, the number of fielders, the rules for dismissal, the size of the boundary and the restrictions on bowlers remain the same. This illustrates the significant bias in favour of the bat in T20 cricket compared to ODI cricket. It also illustrates why ODI cricket has a place for the bowler, and why there are periods in ODI games in which bowlers genuinely have the upper hand.

I looked at the bowling average and strike rates of all bowlers who have bowled in international T20, ODI and Test cricket since the first international T20 game, between Australia and New Zealand in Auckland on February 17, 2005. Using this, I worked out the median bowling average and median strike rate for a bowler in each of the three formats. Then I divided the list of bowlers in each format into four categories:

1. Bowlers with average (AVE) and strike rate (SR) better (lower) than the median.
2. Bowlers with AVE and SR worse (greater) than the median.
3. Bowlers with worse AVE, but better SR.
4. Bowlers with better AVE, but worse SR.

This chart shows the share of the bowling in each format by bowlers from each category.

Good bowlers bowl more in Tests than they do in T20 or ODI cricket. Even though the limits on bowlers are identical in T20 and ODI cricket (a bowler has always been able to deliver no more than 20% of a team's total quota of overs in each form), better bowlers bowl more in ODIs than they do in T20 cricket for the period between 2005 and 2014. Weak bowlers bowl nearly as much as strong bowlers in the T20 game, while in ODIs they bowl about 3/5ths as many overs as the better bowlers do.

The economy rate can be derived here as well. The median bowler in international T20 concedes 7.56 runs per over. Sixty-two per cent of the overs in a T20 game are delivered by bowlers who concede less than this median economy rate. Interestingly, the corresponding figures for ODI and Test cricket from 2005 to 2004 are 5.14 and 69% (ODIs) and 3.38 and 76% in Tests. More profligate bowlers bowl most often in the format in which profligacy should, according to conventional wisdom, hurt a team the most.

This combination of facts is not surprising. The contest between bat and ball is more unequal in T20 than it is in ODI cricket. As a result, teams compromise on the quality of bowling most readily in T20. It is no surprise that the crack specialist bowler is an oddity in the standard IPL auction. These are dominated by hard-hitting batsmen and bits-and-pieces men who are also hard-hitting batsmen. Compromising on bowling is also a feature of bad Test teams - teams that tend to pick their fourth bowler for his batting ability and not because he is likely to get them wickets; teams chasing respectability rather than wins. In T20, this compromise is the rule, not the exception, followed by good and bad teams alike.

The three formats are different. From the point of view of the contest between bat and ball, given that the rules of dismissal and the number of fielders are identical in each format, shrinking the contest to 20 overs limits the role of the bowler to such an extent that the existence of the bowler in T20 XIs has become imperilled. T20 has eliminated those periods in the ODI format in which the batsman did not have the upper hand (due to the balance of resources). In some ways, even within ODI cricket we have seen steady changes to the rules, which have sought to either limit or eliminate periods in which batsmen do not have the upper hand. This occurred because of the original sin of establishing bowler quotas, which forced sides to play five bowlers. The pair of bowlers who bowled in the middle overs in most teams tended to be ordinary, and these middle overs became a stalemate, where the batting side was interested in preserving wickets and the bowling side content to concede about four or five an over. T20 has completed this segregation. Specialist bowlers are decidedly second-class, marginal figures in the new format. Their skill has ceased to matter.

If cricket is a balanced contest between bat and ball, then T20 is not cricket because it has marginalised bowling to a point just short of extinction. In doing so, it has also hurt batting. This situation can be remedied by changing the laws of the game. The details of these changes are for another post. But the laws of T20 will only be changed if the administrators are willing to accept it is not cricket. Until such time, it will remain a mediocre, disfigured caricature of a great sport.