Thursday, 1 January 2009

Looking Forward to: England's Tour of the West Indies 1

Probably the most valuable contribution sabermetrics can make to a cricket fan is via projections of performance. Then you can bet with more confidence!

Projecting players, especially in cricket, is minefield, though. There are too many variables. Perhaps the most important are the characteristics of grounds and of weather. While you can make an estimate of a ground's influence, weather is much harder to do. How on earth could anyone tell us the weather in the West Indies in February 2009? Even a long-range forecast would be too vague to be useful.

Nonetheless, as any aviation pioneer will tell you, 'small sacrifices must be made', so I'll risk looking mathematically incompetent and stupid, and fearlessly post some more projections, as I did a few days ago with Sri Lankan bowlers. In this case, I'm giving you a first run of batting projections for an England XI.

Ah, but already we hit a problem. Who is going to be in England's attack? Might the wicketkeeper change (again!)? Dunno. I've put some names in, and I'll update things as it becomes necessary.

Player          Raw Projection     Regressed
Cook 40.96 35.96
Strauss 32.58 29.50
Bell 38.03 32.77
Pietersen 48.52 36.17
Collingwood 38.34 29.49
Flintoff 24.48 9.81
Prior 33.05 29.79
Sidebottom 9.28 9.32
Anderson 6.43 5.71
Harmison 10.98 3.85
Panesar 3.53 3.11

Now for some comments. Most importantly, you must remember that these projections represent England batsmen against an average Test bowling side, not the West Indian one. We'll get to making adjustments for the West Indians later.

Secondly, why two columns? Well, I'm not sure whether I've regressed too much, given that I'm using all batsmen, and not adjusting between 'recognized' batsmen and 'tailenders'. Best view this entire methodology as 'work in progress'.

Thirdly, for some players I've incorporated their records in domestic cricket, to make up the sample size. I've made a guess as to how much this should be adjusted to reflect the higher standard of play in Tests. Again, this is still 'stick a finger in the air' methodology. I've done some work that suggests I'm in more or less the right area, but that more or less is quite big still.

Moving on to specific players, Prior looks pretty good, Flintoff could either be a marginally useful 6 or 7 or have fallen off a cliff, Pietersen regresses badly and Strauss (an early Cricketing Sabermetrics favourite way back to 2003) may be in need of replacement, more so than Bell. Also, that looks an uncomfortably long tail to me. I'd like to see at least one guy down there who averaged over 12.

They add up to either 286 runs raw or 225 regressed. Now, this is another reason why I've included both. When one looks at the individual totals, the regressions seem too low. When you add them all up, the raw seem too high. The average of the two is 256. Statistics being what it is, you'd do better to remember all three of those numbers, than just pick one.

No comments:

Post a Comment