Tuesday, 24 March 2015

Bleg: Quantifying the value of bowling variety

A common critique of the current English ODI side is that they suffered from having a sameness to their bowling attack--a series of right-handed fast-medium swing bowlers. It's an interesting question. Obviously, diversity in bowling styles can only take you so far: there is a limit to how much aggregate quality a team would be prepared to sacrifice in order to increase diversity, but it is not clear that there is any value to diversity at all. The question was raised recently in the following tweets:
There are some obvious ways in which diversity might improve a team's bowling as a unit. First, having the style of bowling change from over to over might make it harder for batsmen to settle into a rhythm. Second, if some bowling types are more effective wicket takers against right-handed batsmen and others against left-handed, then style diversity might help stop one batsman running away with a game. But these are big mights. The twitter thread above led to this request:

I'm up for the challenge, but I can't see obvious solutions to three conceptual problems:

1. As @CricketFanBob asks, how do you define variety?

If you look at the cricinfo player profiles, you will find that Bill O'Reilly and Clarrie Grimmett, who played in the same Australian test team, are both classified as "legbreak googly". This is true, but this simple classification does not tell you that O'Reilly was an unusually fast spinner who liked to bowl with the wind, while Grimmett was a more-classical flight-into-the-wind legbreak bowler. Similarly, player profiles will tell you that Joel Garner and Malcom Marshall were both "right fast", but the difference between quite fast delivered by a 6'8" bowler and extremely fast delivered by a 5'11" bowler is probably quite substantial. Maybe, however these examples are sufficiently rare that simple cricinfo categorisations are sufficient. But...

2. ...What is the best way of aggregating these bowling-type classifications into a measure of variety.

Is RFM, RFM, LF, RM, SLO more diverse than RFM, LF, RM, SLO, SLO? I think so, but how do you quantify that. And above all,

3.... How do you assess what performance a given set of individual bowler abilities would be expected to produce in order to assess whether variety (or its absence) can explain some of the difference?

In particular, how do you control for the endogeneity that, for example, a spin bowler in spin-friendly conditions will probably a) be in a team with other spin bowlers to take advantage of those conditions, and b) likely to do better than average because of those conditions, making it difficult to infer any value to diversity that might exist.

I have some ideas, but I suspect that the number of variables needed would exhaust the useful degrees of freedom. Any thoughts?


  1. I enjoyed your article. Many people have made comments along the line of how NZ could do better if it overcame its "little investment in the higher value R&D" . Nearly always these are folk who are not about to put their own money behind their infinite wisdom. If it would be so easy to do, why aren't these business commentators making the investment and, as you put it, picking up the easy $20 notes?

  2. An interesting starting point might be to see if there is any evidence of certain types of bowlers fair better against certain types of batsmen. In Baseball most batters do better against opposite handed pitchers.

  3. ballsintherightareasTue Mar 24, 09:32:00 am GMT+13

    Firstly, thank you so much for posing the question. This is one of many cricketing 'common knowledge' assertions that everyone seems to believe but which has perhaps never been tested. If we get any closer here to finding out, I'll be delighted.

    Regarding point 2, perhaps rather than trying to test for all possible permutations of variety, it would be easier to take a few common examples and test those on their own. For example, does having a left arm seamer rather than all right arm seamers help a bowling attack?

    And regarding point 3, I reckon this depends on the format, so perhaps it would be best to focus on one format only at first - eg ODIs. The performance measure I'd suggest would be the average runs scored per over by the opposing team. I am conscious that data may need to be taken from cricinfo's statsguru for this, so perhaps one way of doing it would be to take a random sample of left arm seamers and run queries to find out their team's performances both with and without them in the team, over the period of their international careers?

    You guys might be able to suggest something more sophisticated, though. Has been about 20 years since I studied any stats!

  4. It would be interesting to know if the L v R thing is as strong in cricket as in baseball. But even if it is, and the opposition had a mix of left-handed and right-handed batsmen, would that suggest that it would be optimal to have a mix of left-armed and right-armed bowlers? Maybe, but I'm not sure.

  5. I agree that it would probably be best to focus on variety within the seam bowlers, and just L versus R at first (rather than trying to separate pace from swing, etc.) And yes, I think the focus should be on ODIs. It seems obvious that variety is more important in test cricket, simply because conditions change as a match progresses. You need bowlers to take advantage of the shine, others to soldier on with the older ball, a spinner to take wickets on the last day if the pitch starts to turn, and, in Wellington, someone as strong as an ox to bowl into the wind all day.

    My preferred measure of performance would be to look only at the first innings and then consider a bowler's net reduction in the first-innings WASP. (One could use DL instead, but naturally, I think WASP is a more-accurate measure. )The key thing is to weight runs and wickets according to the game situation and to give the credit where it is due. If a bowler has a very low RPO because the batting team are consolidating following the loss of early wickets, most of the credit should go to the bolwer who took the wickets--A WASP or DL measure will do that. I think I would then look at specific bowlers and separate out their games as a function of the number of overs bowled by pacemen of the same handedness in the game. I'm still not sure I would have enough data to reach a conclusion, but I'll give it a go at some point.

  6. And one cautionary note to saying that the value of variety is obvious in test cricket. I recall fondly the 3rd 1976 NZ v IND test
    in Wellington. NZ had gone with three pace bowlers and their (at that time) best ever spinner--Hedley Howarth--in the previous two tests. Pacemen 1 and 2 had bowled well, but paceman 3 had not. They brought in Lance Cairns to bowl into the Wellington wind, and everyone thought that paceman 3 would be the one dropped. Instead, the selectors eschewed variety and dropped Howarth going with an all pace attack. The captain seemed to agree with the pundits rather than the selectors, because paceman 3 from the previous tests was the 4th bowler used. But he took 7/23 in the second innings, and never again was he anything other than the first bowler used in any test for NZ. His name, of course, was Richard Hadlee. So much for the value of variety!

  7. Paul Romer on Econtalk recently made the point that economists tend to assume all the $20 notes have all been picked up, but in his experience studying growth over decades he has been surprised at how often people keep finding new ones. I think this is doubly true in the modern economy where almost any industry is open to disruption.

    I thought Davies' essay was great. I wish Bollard had spoken about reducing the deterrents to saving, and setting up a charter city in the BOP, for example.

  8. I'm not sure I agree. If you put aside the agricultural sector, NZ is pretty backwards business wise and kiwis don't like change and seem to only care about cricket and rugby as their "intellectual" talk points. A nation of averages pretty much.

  9. This is a bit off topic, but seems like the best place to ask.

    If WASP had replaced D-L, what total would NZ have needed to chase in the semi?

  10. Under WASP we would have had to chase 6 runs more. Not a big difference, although every run counts in such a close game. In general, DL is at its worst when more overs are lost in the second innings than the first, and one team is well ahead at the time the rain comes. The approach of adjusting the target in proportion to the resources lost then has the effect of reducing some of the advantage that the team on top had achieved. In the SAvNZ game, both teams lost the same number of overs, and so this problem wasn't an issue. I hope to do a post on this soon.