Feeds:
Posts

## Live blogging: 20/20 Australia v. South Africa

WA v. Victoria, the first Twenty20 match played in Australia; from the flickr page of Jordan Brock.

Here’s a reverse chronological transcript of my live-blogging the 20/20 match. Actually there’s not much match commentary, because there are people on the television for that, but rather some ruminations on how the statistic(s) used for quantifying batting performance in this form should be modified.

22:34 Well, this is heading for an early finish, so I might sign off unless something extravagant happens. This has been a fun and interesting experiment, which would be improved by my next time announcing it in advance so that readers can leave comments in real time. As it is, I hope there will be some comments after the fact about either the blogging or the suggested statistics. And I do promise to blog more science for  a while—I take it from the lack of comments that people want more of that. And more pictures, too.

22:29 What about doing away with LBW in 20/20? Batsmen have no incentive to protect their stumps with their legs because you can’t get anything other than singles doing that. Given that it lends itself to disproportionately many controversial dismissals, there can’t be anything wrong with taking that pressure off umpires. Not that I’m sad to see J. P. van Damme Duminy walking back to the pavilion canvas chairs.

22:24 Have put a copy of the IPL averages for players with more than ten matches in comma separated value format (for easy importing into Excel/Numbers/Matlab) here.

22:12 The only thing poor about that Mike Hussey catch is the result of it inspiring Mark Nicholas to spout poetry. The Betfair trading must have be swinging sharply to Australia. I think that Crown Lager ad with the chap tapping his bottle is about the only good thing they show between overs; does that mean that Crown Lager thinks I’m their demographic?

21:58 Those skin cancer ads are really scary! Kids, wear your suncream.

21:57 The CricInfo Statsguru list of IPL Batting Averages is a useful find. There are about sixty players that have played more than ten matches, which is fairly good numbers, though not enough to eke out how their averages are distributed; it looks pretty uniform, really. Strikes rates certainly have a central peak though, so I’m going to call them Gaussonian for now. I’ll post some plots after the match. Query for the time being: is it possible to construct from these two random variables the quantity that is unambiguously most nearly Gaussian? More importantly, is that quantity special in any useful way? I suppose ‘usefulness’ here is imbued from agreement by folks watching cricket that any player with a higher ‘batting statistic’ than any other is a better batsman.

21:45 Do the distributions of averages or strike rates obey a particular form? Obviously a Bell Curve would be lovely, but as Taleb tells us, we just like to imagine them so we can invent narratives for ourselves. Is dimensional analysis any help: avg (runs/innings), strike rate (runs/balls). Don’t really see an ‘obvious’ way to combine them, so normalising them individually either to a single ‘average’ quantity or through a distribution (so each is a z-score) and only then combining them seems to be most clever.

$C = \frac{\mathrm{Average}}{\bar{\mathrm{Average}}} + \frac{\mathrm{SR}}{\bar{\mathrm{SR}}}$

Here ‘+’ represents a general binary operation.

21:32 Australia looking good first-up thanks to the RTA Speed Blitz Blues’ Bracken. And de Villiers is now both out and injured. Two birds, one stone, as it were. Poor chap.

21:19 There is a further subtle point with regard to combining raw averages with strike rate. In 20/20, because of the relatively brisk rate-of-play, a high average (runs/innings) almost certainly indicates a high strike-rate, while a high strike-rate could indicate anything. So there’s a running correlation, or rather, the variables are related in a non-linear fashion.

21:15 Who is this person who is singing? Not Emma Kirkby, that’s for sure. Apparently she [clarification: the singer, not Kirkby; JBJ 21:51] is on fire.

21:0814 182 it is. Should try computing these numbers for other sides. Some points that come to mind: taking SR/100 for the fifty-over form doesn’t seem to make sense when the representative figure seems more like 80. Certainly some scaling between the two forms is necessary; a factor of 2.5 somewhere feels right.

There are units problems too, and a dimensionless statistic would be optimal. C. H. Evans pipes up again, suggesting dividing the average by the, umm, average average of batsmen for each form, and the strike rate by the average strike rate, so that each is dimensionless. There’s some uncertainty about what to do after that to combine the two figures.

21:03 Things dying away for Australia here, but I think 180 will be enough because the MCG takes up about half the green-space in Victoria. Having attempted some quick calculation of modified batting ‘averages’ (20/20 first; 50/50 second):

	        avg	balls   SR	B = avg*SR/100
RT Ponting 	39.37	228	138.15	54.389655
MEK Hussey 	22.83	106	129.24	29.505492
DA Warner 	89	43	206.97	184.2033

RT Ponting 	43.24	13816	80.43	34.777932
MEK Hussey 	57.13	2873	85.52	48.857576
MJ Clarke	42.46	5120	79.62	33.806652
BJ Haddin	28.66	985	78.57	22.518162

20:44 I don’t see any reason they couldn’t have ~30 players playing for three distinct Australian teams, one in each of the forms. That would do away with complaints about players having to play ‘too much cricket’.

20:39 Warner’s out! That’s a shame for everyone, but we’ll have forgotten after his next few innings, which promise to be just magnificent.

20:38 Further comment re: batting in this form. Clearly Ricky Ponting, who is generally agreed to be on song wrt. batting at the moment, is not effective in this form. He’s a strokemaker, not a smasher; he doesn’t have the minarets of sinew for it. Michael Clarke should be captaining this form. D. Hussey out now—Cameron White would be the sensible next in… and it is!

20:34 Paul Reiffel is third umpire! What’s next? Don Burke as groundskeeper?

20:32 Well I’m a little late in getting started, unlike, say, David Warner, who is off to a flyer on debut, having never played a first-class match before (i.e. even at state level). What a champ. Across the couch, C. H. Evans makes the obvious point that average not the critical figure in this form of the game, as strike rate is so important, as well as the less obvious point that indeed there is some slide in this between the fifty- and twenty-over forms.

My proposal for the accurate single batting statistic is:

$B = \mathrm{Average} \times \frac{\mathrm{Strike\ rate}}{100}$.

Comments? Suggestions? The commentators indicate that 100 from 50 is the fastest in this form, and I think 100 from 42 is the fastest in any form. No it’s 100 from 50.