Top ODI batsmen by Runs Added
(Those of you who don't care about cricket, which I think is everyone who reads this blog, please ignore.)
At the Cricinfo Stat Blog, there's been much recent discussion of how to rank the top ODI batsmen of all-time. Most of the methods proposed seem arbitrary. One useful post was by Ric Finlay, who gave his top 30 batsmen according to some weird "Bat Quality" index, but more intriguingly also gave runs scored by these batsmen, adjusted "according to batting conditions and the quality of the opposition bowling".
The generally accepted way to compare baseball batters across eras is to look at runs or wins above replacement (please, don't mention Win Shares). So I tried to convert the ODI adjusted runs to "Runs Added" for these 30 batsmen. "Runs Added" is meant to represent the additional runs scored by the player's team when batting due to the presence of that player, compared to a borderline "replacement level" player.
For matches from 1996 onwards, I (arbitrarily) set the replacement player as having an average of 30 and a strike rate of 70; the strike rate was lower for matches prior to 1996. We can think of a top player as adding runs in two ways: by scoring more quickly than a 70 SR, and by averaging over 30 (thereby preserving wickets for the final push). I (arbitrarily) set an extra wicket in hand as worth 10 runs, which I think is close to right for most of an innings.
I then adjusted for schedule, since many more ODIs per team are played nowadays compared to, say, the Eighties. So I weighted the stats of each player so that each would play 30 innings a year (fewer if the player was not always selected for whatever reason).
Finally I made a (really arbitrary) adjustment for the higher SD in performance before 1984: in the early years, there was a much larger gap in ability between the best and worst ODI players, so it would be too generous to give the same credit for Runs Added back then as now. This adjustment should be made (see Dan Rosenheck's arguments for doing so for baseball somewhere on the Baseball Hall of Merit site).
I got:
(Batsman - Projected Runs - Runs Added)
Richards - 17260 - 6780
Tendulkar - 16835 - 4293
Jayasuriya - 14886 - 3042
de Silva - 13556 - 2846
Lara - 13730 - 2801
Gilchrist - 9799 - 2621
Zaheer Abbas - 9547 - 2537
Javed Miandad - 15259 - 2483
D. Jones - 9031 - 2315
M. Waugh - 11169 - 1874
Mohammed Azharuddin - 11949 - 1837
Haynes - 15452 - 1625
Symonds - 6009 - 1611
Bevan - 9102 - 1606
Saeed Anwar - 9024 - 1572
Greenidge - 11621 - 1449
Cronje - 7438 - 1399
Klusener - 4716 - 1305
Inzamam ul-Haq - 12280 - 1271
Pietersen - 3567 - 993
Dhoni - 3567 - 960
Ponting - 12036 - 900
Hayden - 6642 - 900
Hussey - 3153 - 786
Kirsten - 7369 - 717
Mohammad Yousuf - 9458 - 706
Kallis - 10896 - 655
Ganguly - 10822 - 422
Dravid - 9699 - 323
Chanderpaul - 9903 - 246
That's a clear winner. Viv has 68% more Runs Added than Sachin, 123% more than Sanath. (As a comparison, in my dodgy Test match Runs Added calculations, Bradman has 136% and 143% more than Garry Sobers and George Headley, ranked second and third respectively.)
My major, major assumption is that each batsman's ability is constant throughout his career. This is, of course, false, but the ideas I have to fix this would be very hard work.
Does the system undervalue latter-day grinders like Chanderpaul and Dravid? Probably, but for teams that don't get bowled out a lot, it's hard to argue they're anywhere near as valuable as speed demons like Jayasuriya and Gilchrist. Forget replacement level, if your grinder falls down dead before the innings starts, it's only 7.6% of your resources, according to Messrs Duckworth and Lewis.
Does the system undervalue finishers like Bevan and Hussey? I guess it depends on what you're trying to measure. No doubt Bevan and Hussey were/are talented players, but were/are they really that valuable to Australia? With 4 down and 10 overs to go, a wicket is only worth 2.3% of your original resources. Why play for the not out? Hitting another 30 runs and seeing the innings through to the end is probably less valuable than slogging another 20 off 10 balls and trusting the tail to bat out the remaining overs.
Does the system overvalue old-timers like Zaheer Abbas or even Mohammad Azharuddin? Probably, but how much? Should there be an SD adjustment after 1984? (I mean, there should, but will it make that much of a difference?)
Anyway, all of this should be thrown out once David Barry runs his simulations.
At the Cricinfo Stat Blog, there's been much recent discussion of how to rank the top ODI batsmen of all-time. Most of the methods proposed seem arbitrary. One useful post was by Ric Finlay, who gave his top 30 batsmen according to some weird "Bat Quality" index, but more intriguingly also gave runs scored by these batsmen, adjusted "according to batting conditions and the quality of the opposition bowling".
The generally accepted way to compare baseball batters across eras is to look at runs or wins above replacement (please, don't mention Win Shares). So I tried to convert the ODI adjusted runs to "Runs Added" for these 30 batsmen. "Runs Added" is meant to represent the additional runs scored by the player's team when batting due to the presence of that player, compared to a borderline "replacement level" player.
For matches from 1996 onwards, I (arbitrarily) set the replacement player as having an average of 30 and a strike rate of 70; the strike rate was lower for matches prior to 1996. We can think of a top player as adding runs in two ways: by scoring more quickly than a 70 SR, and by averaging over 30 (thereby preserving wickets for the final push). I (arbitrarily) set an extra wicket in hand as worth 10 runs, which I think is close to right for most of an innings.
I then adjusted for schedule, since many more ODIs per team are played nowadays compared to, say, the Eighties. So I weighted the stats of each player so that each would play 30 innings a year (fewer if the player was not always selected for whatever reason).
Finally I made a (really arbitrary) adjustment for the higher SD in performance before 1984: in the early years, there was a much larger gap in ability between the best and worst ODI players, so it would be too generous to give the same credit for Runs Added back then as now. This adjustment should be made (see Dan Rosenheck's arguments for doing so for baseball somewhere on the Baseball Hall of Merit site).
I got:
(Batsman - Projected Runs - Runs Added)
Richards - 17260 - 6780
Tendulkar - 16835 - 4293
Jayasuriya - 14886 - 3042
de Silva - 13556 - 2846
Lara - 13730 - 2801
Gilchrist - 9799 - 2621
Zaheer Abbas - 9547 - 2537
Javed Miandad - 15259 - 2483
D. Jones - 9031 - 2315
M. Waugh - 11169 - 1874
Mohammed Azharuddin - 11949 - 1837
Haynes - 15452 - 1625
Symonds - 6009 - 1611
Bevan - 9102 - 1606
Saeed Anwar - 9024 - 1572
Greenidge - 11621 - 1449
Cronje - 7438 - 1399
Klusener - 4716 - 1305
Inzamam ul-Haq - 12280 - 1271
Pietersen - 3567 - 993
Dhoni - 3567 - 960
Ponting - 12036 - 900
Hayden - 6642 - 900
Hussey - 3153 - 786
Kirsten - 7369 - 717
Mohammad Yousuf - 9458 - 706
Kallis - 10896 - 655
Ganguly - 10822 - 422
Dravid - 9699 - 323
Chanderpaul - 9903 - 246
That's a clear winner. Viv has 68% more Runs Added than Sachin, 123% more than Sanath. (As a comparison, in my dodgy Test match Runs Added calculations, Bradman has 136% and 143% more than Garry Sobers and George Headley, ranked second and third respectively.)
My major, major assumption is that each batsman's ability is constant throughout his career. This is, of course, false, but the ideas I have to fix this would be very hard work.
Does the system undervalue latter-day grinders like Chanderpaul and Dravid? Probably, but for teams that don't get bowled out a lot, it's hard to argue they're anywhere near as valuable as speed demons like Jayasuriya and Gilchrist. Forget replacement level, if your grinder falls down dead before the innings starts, it's only 7.6% of your resources, according to Messrs Duckworth and Lewis.
Does the system undervalue finishers like Bevan and Hussey? I guess it depends on what you're trying to measure. No doubt Bevan and Hussey were/are talented players, but were/are they really that valuable to Australia? With 4 down and 10 overs to go, a wicket is only worth 2.3% of your original resources. Why play for the not out? Hitting another 30 runs and seeing the innings through to the end is probably less valuable than slogging another 20 off 10 balls and trusting the tail to bat out the remaining overs.
Does the system overvalue old-timers like Zaheer Abbas or even Mohammad Azharuddin? Probably, but how much? Should there be an SD adjustment after 1984? (I mean, there should, but will it make that much of a difference?)
Anyway, all of this should be thrown out once David Barry runs his simulations.
Labels: cricket statistics