Saturday, December 22, 2007

Modified batting averages

Over at the Cricket Fans' Forum I saw this article, which describes some of the work of two economists, who have come up with a way of adjusting the batting average to take into account the consistency of the batsman and the importance of his scores in each individual innings. I don't think that the former adjustment is warranted at all. The latter adjustment has some merit, but I'm not convinced about it — a good batsman in a weak team would be unduly rewarded, and vice versa.

One modification to the batting average that does make sense is an adjustment for the bowling strength of the opposition. I've implemented this in a pretty simple way. For a given batsman's innings, I weight the runs scored by a average of the bowlers' bowling averages, where this latter average is weighted by how much each bowler bowled.

Perhaps an example would be useful. Suppose that a batsman from Team A scores 100, and in that innings Team B used three bowlers. Bowler A, whose bowling average is 25, bowled 20 overs. Bowler B, whose average is 25, bowled 10 overs. Bowler C, whose average is 30, also bowled 10 overs. The weighted bowling average would then be 0,5*20 + 0,25*25 + 0,25*30 = 23,75.

(Once again, I've used the career averages. This makes the numerics much more stable and easier to perform, but means that the modified averages can change when a batsman's not playing.)

This is similar in spirit to the modified bowling averages below, though somewhat less elegant — you'd expect that a captain would often give the most overs to whoever happens to be bowling the best that day, regardless of what his career average is. Still, it's a relatively simple idea and it seems pretty fair.

Here are the top 25 Test batsman according to this modified average (qual. 20 innings):


name matches inns no runs avg adj runs adj avg
Bradman,Donald 52 80 10 6996 99,94 6321,22 90,30
Hussey,Michael 18 29 7 1896 86,18 1639,18 74,51
Headley,George 22 40 4 2190 60,83 2261,05 62,81
Walcott,Clyde 44 74 7 3798 56,69 3903,64 58,26
Weekes,Everton 48 81 5 4455 58,62 4243,46 55,84
Jackson,Stanley 20 33 4 1415 48,79 1606,10 55,38
Hobbs,Jack 61 102 7 5410 56,95 5218,26 54,93
Sobers,Garry 93 160 21 8032 57,78 7590,09 54,61
Hutton,Leonard 79 138 15 6971 56,68 6613,95 53,77
Sutcliffe,Herbert 54 84 9 4555 60,73 4020,51 53,61
Nourse,Dudley 34 62 7 2960 53,82 2932,81 53,32
Barrington,Ken 82 131 15 6806 58,67 6096,42 52,56
Chappell,Greg 87 151 19 7110 53,86 6899,34 52,27
Tyldesley,Ernest 14 20 2 990 55,00 939,58 52,20
Pollock,Graeme 23 41 4 2256 60,97 1924,28 52,01
Ponting,Ricky 111 184 26 9408 59,54 8111,16 51,34
Ranjitsinhji 15 26 4 989 44,96 1116,44 50,75
Pietersen,Kevin 32 61 3 2993 51,60 2934,97 50,60
Kallis,Jacques 110 187 30 9114 58,05 7917,01 50,43
Lara,Brian 130 230 6 11912 53,18 11272,23 50,32
Ryder,Jack 20 32 5 1394 51,63 1358,58 50,32
Worrell,Frank 51 87 9 3860 49,49 3856,59 49,44
Steel,Allan 13 20 3 600 35,29 831,70 48,92
Harvey,Neil 79 137 10 6149 48,42 6193,82 48,77
Compton,Denis 78 131 15 5807 50,06 5656,85 48,77


(Includes all Tests up to Test 1852, the second between Sri Lanka and England in the current series.)

Most of the usual suspects at the top of the list take quite a hit, with Headley and Walcott the two exceptions. The biggest major falls are from Eddie Paynter and Wally Hammond, who come in 36th (47,27) and 37th (46,86) respectively. Stanley Jackson leads the charge of the 19th century brigade (though he also played in the early 20th century). Allan Steel, who played all of his Tests in the 1880's, has one of the biggest increases, going from 35,29 to 48,92. Steel is also pretty cool because he made nice round numbers: 600 Test runs, and 7000 first-class runs.

Some other notables (usual average in brackets): JJ Lyons 44,37 (27,07); WG Grace 42,91 (32,29); Arthur Shrewsbury 46,32 (35,47); Victor Trumper 41,73 (39,05); Chris Martin 1,97 (2,00).

Feel free to request any others.

Comments:
Interesting David.

Have you read the blog on cricinfo regarding the extended batting average? It is a modified average and takes into account 'not out' innings with a different perspective. Here's the link:

http://blogs.cricinfo.com/itfigures/archives/2007/12/tackling_notouts_and_answering.php
 
Hi Q. Narayanan's extended batting average is seriously flawed. There are two underlying assumptions to it. The first is that the recent innings of a batsman give a good indication of what his next innings will be. This isn't all that true - I have some graphs on the topic here, which show that while it might be a good indication for some players, it isn't for others. And generally, any effect of "form" is quite weak.

The second problem is that it doesn't take into account the current score of the batsman. Once again I have graphs. :) They are here. Narayan's extended batting average can change with a score of nought not out! It's not realistic at all.
 
I completely agree ith your assessment David. Love your graphs.
 
Post a Comment

Subscribe to Post Comments [Atom]





<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]