Vijay's All-Time Windies XI

Vijay's All-Time Windies XI

Post by Vijay Kumar » Wed, 30 Jan 2002 07:49:32


A la Mike... Having read Tony Cozier's 50 years of WI cricket (written over 20
years ago, and finished reading about 18 summers ago), here is my take on the
best All-time WI XI, in batting order. I know stats are not everything, but even
on stats, this side is mighty impressive.

CC Hunte          44    78   3245   260    45.06    8    13    16
GC Greenidge     108   185   7558   226    44.72   19    34    96
GA Headley        22    40   2190   270*   60.83   10     5    14
IVA Richards     121   182   8540   291    50.23   24    45   122
                               32  2-17    61.37    0     0    161.5
G St.A Sobers     93   160   8032   365*   57.78   26    30    109
                              235  6-73    34.03    0     6    91.9
FMM Worrell       51    87   3860   261    49.48    9    22    43
                               69  7-70    38.72    0     2    103.4
DL Murray         62    96   1993    91    22.90    0    11    181 + 8
MD Marshall       81   107   1810    92    18.85    0    10    25
                              376  7-22    20.94    4    22    46.7
J Garner          58    68    672    60    12.44    0     1    42
                              259  6-56    20.97    0     7    50.8
CEL Ambrose       98   145   1439    53    12.40    0     1    18
                              405  8-45    20.99    3    22    54.5
LR Gibbs          79   109    488    25     6.97    0     0    52
                              309  8-38    29.09    2    18    87.7

(First line : Tests-Innings-Runs-HS-Avg-100s-50s-Ct + St.
 Second line: Wickets-Best Bowling (I)-Avg.-10WM-5WI-Strike Rate)
(Greenidge, Hunte and Headley all bowled in tests, but were not regular bowlers
in any sense of the term. Murray never bowled in tests)

The *real* tough choices were
- BC Lara. The man is a genius. Leaving him out was unthinkable. It speaks
volumes for the great history of WI cricket that there were enough there who
could replace him without much loss. Eventually, I chose Worrell for Lara,
giving us the most cohesive factor in any WI team, a captain par excellence, in
place of one who has often been seen as a divisive factor.
- Holding. The fourth best bowler ever for the Windies, behind Marshall, Garner
and Ambrose. Gibbs gets in because of the "variety" factor. Granted, Sobers &
Richards can both bowl a bit of spin as can Worrell, but 309 wickets *must*
count for something. And Steve the Bajan has already published and copyrighted
the theory of how overbowling and well-past-sell-date matches have tarnished
Gibbs' record.
- Weekes & Walcott. So, who do we lose to fit them in? Sobers? Headley? Looking
at the paucity of genuine #3s for India, the WI cupboard is embarrassingly
full in comparision. Weekes, Richards, Lara & Headley have all had successful
stints at #3. As has Walcott. Given Weekes' reputation as a bit of a weak-attack
bully (India at the receiving end, mostly), Weekes had to sit out. Walcott at
#5 was up against Sobers. No contest, really.
- Haynes. As the other half of the most prolific opening combination ever, he
deserves to be there just for completion sake. But Hunte's record as opener is
better, if anything. And Hunte's fielding was phenomenal, too. Alan Rae's stats
as opener are marginally better but Hunte's longevity weighs in on his side.
- Dujon. As a batsman-keeper, there is no argument for Murray really. Dujon
also has a better dismissals/match ratio. So? Murray has actually kept to
Marshall, Garner, Sobers and Gibbs. Dujon has no real keeping experience to
spinners. If Gibbs were to give way to Holding, Dujon should take over from
Murray. Actually, an even better case to be made for Jacobs on dismissals/game
alone, but his lack of experience to quality spinners counts against him.

A couple of pernniel favourites sit out - Lawrence Rowe for one. Roy Fredericks
for another. And if one has to leave out players of the caliber of Lloyd &
Kanhai & Nurse (to say nothing of Lara), the batting must be rich indeed. Or
bowlers of the class of Holding, Walsh, Hall, Roberts, Croft & Ramadhin. One
minor surprise, Roger Harper has the best average among WI spinners, ahead of
Ramnaraine, Ramadhin & Gibbs.

Vijay

 
 
 

Vijay's All-Time Windies XI

Post by gmurth » Wed, 30 Jan 2002 08:11:39

Quote:

> A la Mike... Having read Tony Cozier's 50 years of WI cricket (written over 20
> years ago, and finished reading about 18 summers ago), here is my take on the
> best All-time WI XI, in batting order. I know stats are not everything, but even
> on stats, this side is mighty impressive.

> CC Hunte          44    78   3245   260    45.06    8    13    16
> GC Greenidge     108   185   7558   226    44.72   19    34    96
> GA Headley        22    40   2190   270*   60.83   10     5    14
> IVA Richards     121   182   8540   291    50.23   24    45   122
>                                32  2-17    61.37    0     0    161.5
> G St.A Sobers     93   160   8032   365*   57.78   26    30    109
>                               235  6-73    34.03    0     6    91.9
> FMM Worrell       51    87   3860   261    49.48    9    22    43
>                                69  7-70    38.72    0     2    103.4
> DL Murray         62    96   1993    91    22.90    0    11    181 + 8
> MD Marshall       81   107   1810    92    18.85    0    10    25
>                               376  7-22    20.94    4    22    46.7
> J Garner          58    68    672    60    12.44    0     1    42
>                               259  6-56    20.97    0     7    50.8
> CEL Ambrose       98   145   1439    53    12.40    0     1    18
>                               405  8-45    20.99    3    22    54.5
> LR Gibbs          79   109    488    25     6.97    0     0    52
>                               309  8-38    29.09    2    18    87.7

Good team. Some comments: I would have Walcott open with Greenidge. To
be followed by Richards, Headley, Weekes and Sobers. Also, considering the
pace strength flaunted by WI, I wouldn't "waste" a bowling slot on anyone less
than an unplayable spinner, so I'd leave out Gibbs and throw in Holding.
Although Walcott could keep, I'd have Dujon in there as specialist keeper.
Dujon Vs Murray is a close call as far as keeping itself goes, but Dujon was
a better batsman IMO. Regretably, I'm unable to fit in Worrell and Lara, more
so the latter who at his best is an unbelievably brilliant batsman. I could have
Lara over Weekes, but I don't want Lara batting at #5 and I don't want
Richards and Headley batting at positions other than #3 and #4 respectively.

-Ganesh

 
 
 

Vijay's All-Time Windies XI

Post by Yuk Tan » Wed, 30 Jan 2002 09:24:45



[snip]

Quote:
> A couple of pernniel favourites sit out - Lawrence Rowe for one. Roy
Fredericks
> for another. And if one has to leave out players of the caliber of Lloyd &
> Kanhai & Nurse (to say nothing of Lara), the batting must be rich indeed.
Or
> bowlers of the class of Holding, Walsh, Hall, Roberts, Croft & Ramadhin.
One
> minor surprise, Roger Harper has the best average among WI spinners, ahead
of
> Ramnaraine, Ramadhin & Gibbs.

The Bajan would also tell you that, if you were to include Ramadhin ahead of
someone like Gibbs, you'd only have 3 bowlers plus an all-rounder in the
side.

Cheers, ymt.

 
 
 

Vijay's All-Time Windies XI

Post by Uday Raja » Wed, 30 Jan 2002 13:26:10

Quote:

> CC Hunte          44    78   3245   260    45.06    8    13    16
> GC Greenidge     108   185   7558   226    44.72   19    34    96
> GA Headley        22    40   2190   270*   60.83   10     5    14
> IVA Richards     121   182   8540   291    50.23   24    45   122
>                                32  2-17    61.37    0     0    161.5
> G St.A Sobers     93   160   8032   365*   57.78   26    30    109
>                               235  6-73    34.03    0     6    91.9
> FMM Worrell       51    87   3860   261    49.48    9    22    43
>                                69  7-70    38.72    0     2    103.4
> DL Murray         62    96   1993    91    22.90    0    11    181 + 8
> MD Marshall       81   107   1810    92    18.85    0    10    25
>                               376  7-22    20.94    4    22    46.7
> J Garner          58    68    672    60    12.44    0     1    42
>                               259  6-56    20.97    0     7    50.8
> CEL Ambrose       98   145   1439    53    12.40    0     1    18
>                               405  8-45    20.99    3    22    54.5
> LR Gibbs          79   109    488    25     6.97    0     0    52
>                               309  8-38    29.09    2    18    87.7

Good team. I especially like having Worrell in there as
captain. It turns out that 49.48 was the lowest
Worrell's batting average ever got to in Test cricket.

It's a tough choice amongst the bowlers. Marshall and
Ambrose must be in there, but the other two places are
up for grabs. If you do pick a spinner, I agree, Gibbs
over Valentine (Ramadhin being thrown out for obvious
reasons). I'm tempted to include Holding in place of
Garner, but with a spearhead of Marshall and Ambrose, a
support bowler like Garner may be just what's needed.
In any case, it's a choice I'd like to have.

 
 
 

Vijay's All-Time Windies XI

Post by Paul Robs » Wed, 30 Jan 2002 19:44:25

Ya know, I started to pick the worst all time XI.

But it ended up being the current team :)

 
 
 

Vijay's All-Time Windies XI

Post by John » Thu, 31 Jan 2002 00:20:36


Quote:
> A la Mike... Having read Tony Cozier's 50 years of WI cricket (written over 20
> years ago, and finished reading about 18 summers ago), here is my take on the
> best All-time WI XI, in batting order. I know stats are not everything, but
> even on stats, this side is mighty impressive.

Raw stats aren't everything.  Everyone knows you have to adjust for things
like Weekes bullying weak attacks, but I've actually done it.  Results are
iteratively adjusted for opposition, conditions, and home/abroad.  In the
Bayesian model, averages are not means but expectations; in this case, for
a player's performance against average opposition in benchmark conditions.
The peak averages are the expectation for a player at his best.  The side
would be more impressive, statistically, if it included Walcott.

                      Career  Adj        Peak
IVA Richards           52.77 1.0562  1976/07/08  59.06
GS Sobers              55.19 0.9945  1966/08/04  57.88
CL Walcott             50.93 0.9652  1955/04/11  56.29
FMM Worrell            48.71 0.9451  1950/01/14  54.42
GA Headley             52.57 0.9689  1935/03/14  52.58
BC Lara                49.07 1.0001  2001/11/29  50.32
ED Weekes              50.13 0.9062  1954/01/15  50.23
CH Lloyd               46.32 1.0392  1983/12/10  49.76

These are the middle-order candidates.  The adjustment is applied to the
raw averages before the posterior is computed.  Richards, for example,
scored his runs against better bowling than Weekes, and is credited
appropriately.  Headley's career is so spread out that he has no peak
according to the algorithm.  On career value, I'd take Headley at 3,
Richards, Walcott, Sobers; for peak, Richards, Worrell, Walcott, Sobers.
The other four make the second team.

CG Greenidge           43.83 1.0167  1984/04/28  45.85
RC Fredericks          42.09 1.0256  1976/08/12  44.48
JB Stollmeyer          40.87 1.0044  1949/02/04  42.97
CC Hunte               40.72 0.9301  1965/03/03  42.56
DL Haynes              40.99 1.0094  1993/04/16  42.54
LG Rowe                37.50 0.9082  1974/03/06  41.02
AF Rae                 39.48 0.9358  1950/07/20  39.92

The openers aren't nearly as strong, but better than middle-order bats
out of position.  Greenidge and Fredericks are clearly the top two.
Stollmeyer, Hunte, and Haynes are too close to call.

                      Career  Adj        Peak            W/M
J Garner               21.30 0.9506  1980/06/05  19.41  4.334
MD Marshall            21.21 0.9798  1988/06/02  19.45  5.337
ST Clarke              23.07 0.9467  1984/01/27  19.84  5.591
CEL Ambrose            21.41 0.9881  1993/01/23  20.02  4.862
CC Griffith            28.42 0.9360  1963/07/25  21.18  5.166
MA Holding             22.99 0.9363  1981/12/26  21.19  4.668
EA Martindale          26.87 1.0054  1935/01/08  22.31  4.648
HHH Johnson            29.43 1.0455  1948/03/27  22.43  7.562
CEH Croft              24.23 0.9325  1977/03/04  22.82  5.538
CA Walsh               24.64 0.9902  2000/06/29  22.85  4.604
LR Gibbs               28.54 0.9329  1961/01/27  22.99  4.993
IR Bishop              25.22 0.9832  1993/01/30  23.29  4.344
AME Roberts            24.35 0.9380  1975/01/11  23.42  5.273
WW Hall                26.89 0.9819  1962/03/07  23.92  4.571

Garner, Marshall, and Ambrose are obvious choices.  Clarke,
like Richards and Worrell, is boosted by excellent performance
in unofficial Tests.  Including matches of Test standard
regardless of Test status improves the estimates and thus
the log-likelihood, so I've counted them.  On career value,
Holding is probably the fourth bowler, but for peak it has
to be Clarke.  Martindale and Johnson are listed despite
careers perhaps too short for an all-time XI.  None of the
last five stands out from the others on peak, their batting
making up the fractions of a point between them.  The team
lacks a spinner, so the spot goes to Gibbs.

The statistics necessary to evaluate wicketkeepers are not
kept, so my choice of Dujon over Murray is largely subjective.

So, for career: Greenidge, Fredericks, Headley, Richards,
Walcott, Sobers, Dujon, Marshall, Holding, Ambrose, Garner.
Second XI: Stollmeyer, Haynes, Lara, Weekes, Worrell,
Lloyd, Deryck Murray, Roberts, Croft, Clarke, Walsh.

For peak: Worrell ahead of Headley, Clarke ahead of Holding,
Griffith and Gibbs ahead of Roberts and Walsh.

I'm never sure what criteria people use to select their All-Time XIs.
If it's career, you can't consider Gibbs.  If it's peak, perhaps people
don't realize how great Walcott became after giving up the gloves.

John

 
 
 

Vijay's All-Time Windies XI

Post by Pau » Thu, 31 Jan 2002 06:04:03

What about Leary Constantine?
 
 
 

Vijay's All-Time Windies XI

Post by Pau » Thu, 31 Jan 2002 06:04:04

What about Leary Constantine?
 
 
 

Vijay's All-Time Windies XI

Post by Yuk Tan » Thu, 31 Jan 2002 06:25:16


Quote:
> What about Leary Constantine?

Would probably make an Indian, Sri Lankan or NZ XI, but who do you suggest
that he should replace in the WI lineup?  Sobers?  Marshall?  Garner?
Ambrose?  This is probably the strongest bowling attack of any all-time XI
in the world, while batting-wise, 2 players who averaged in the mid-50s
(Weekes and Walcott) and at his peak, probably the best bat since Bradman
(Lara) can't make the team.

Cheers, ymt.

 
 
 

Vijay's All-Time Windies XI

Post by Ian Galbrai » Thu, 31 Jan 2002 11:07:17



:> A la Mike... Having read Tony Cozier's 50 years of WI cricket (written over 20
:> years ago, and finished reading about 18 summers ago), here is my take on the
:> best All-time WI XI, in batting order. I know stats are not everything, but
:> even on stats, this side is mighty impressive.

:Raw stats aren't everything.  Everyone knows you have to adjust for things
:like Weekes bullying weak attacks, but I've actually done it.  Results are
:iteratively adjusted for opposition, conditions, and home/abroad.  In the

Interesting. How exactly did formulate your adjustment figure, if you
don't mind me asking? Maybe if you use Viv Richards as an example.

[snip]

--
Ian Galbraith

"Mm, I wouldn't dream of interfering." Mark made for
the door. "Though I'm not at all sure I'd choose to
structure my most intimate relationship as a war.
Is she the enemy, then?"
- A Civil Campaign - Lois McMaster Bujold

 
 
 

Vijay's All-Time Windies XI

Post by John » Fri, 01 Feb 2002 01:23:26

Quote:

> Interesting. How exactly did formulate your adjustment figure, if you
> don't mind me asking? Maybe if you use Viv Richards as an example.

I'm not sure I can explain it concisely.  Certainly not using Richards.

Andy Ganteaume played one Test, scoring 112 in his only innings.
The mean of a large sample is usually close to the mean of the
distribution from which the sample is drawn, but with a sample
size of 1, the sample mean is meaningless.  His "career average"
should be what he'd be expected to average long-term.  If we knew
nothing else about him, maybe we'd guess 112.00, but we know he
was a Test cricketer.  The distribution of Test cricketer's batting
averages, heavy in the 20s, very light above 60, is our prior
distribution.  And he scored 112, so we combine the prior probability
his average was X with the probability of someone averaging X scoring
112 in a given innings to compute the posterior distribution.  The
expectation of the posterior is then the long-term expectation.

The other thing to consider is that not all 112s are equal.
Batting at home is easier, so the innings is discounted
4.25%.  This gives us an initial estimate of 36.05.

Initial estimates are also made for the bowlers he faced, including
part-timers brought on as WI amassed 497.  Ikin was probably not
bad enough to average 123.33 long-term, so again the expectation
of the posterior produces a better estimate than the average.
Laker and Wardle weren't yet Laker and Wardle in their debut series,
so performances long before or after the match in question are given
less weight.  The estimate of the combined attack's strength is
compared to the baseline.  In this case, it was 4% weaker.

Then the batting conditions are evaluated by comparing the runs scored
by both sides with the expected scoring for those players against those
attacks.  This match was a high-scoring draw with all four openers making
centuries.  Scoring was about 15% higher than would be expected.  The
program assigns 5% to good batting, 5% to poor bowling, and 5% to favorable
conditions, and discounts all batsmen's scores in the match by 10%.

This gives us a revised estimate for all the players.  The process is
then repeated until the estimates converge.  Ganteaume's final adjustment
is 0.827 and his final "average" is 34.59.

For someone like Richards, the adjustments are calculated for each match
as above.  They are then averaged, weighted by his scores in each match,
to determine the mean environment in which his runs were scored.  The
early part of his career looks like this:

  R  O  Adj    Cond
  7/ 2 0.9759 1.062  in India, Abid Ali/Solkar/Chandra/Pras/Venkat
192/ 0 0.8961 0.944  Abid Ali/Solkar/Bedi/Pras/Venkat
 62/ 2 1.1450 1.085  Ghavri/Madan Lal/Bedi/Chandra/Pras, improved attack
 52/ 2 1.3775 1.329
 40/ 1 0.8137 0.801
151/ 1 0.9651 1.104  in SL against a weak attack
 80/ 2 1.2711 1.309
  7/ 2 1.0379 1.021  in Pakistan
 10/ 1 0.8390 0.840
 12/ 2 1.0252 0.794  in Aus against a very strong attack
 12/ 1 1.1246 0.854  
 77/ 2 1.1983 0.918
 46/ 2 1.2896 1.038
131/ 2 1.0418 0.821
148/ 2 1.3701 0.996
142/ 1 1.0621 1.106  at home against India
150/ 2 1.0144 1.019
200/ 2 0.9266 0.967
 64/ 1 0.7913 0.828
295/ 2 1.0632 0.959  in England
139/ 2 1.3793 1.454
104/ 2 1.1636 1.044
291/ 1 0.7994 0.787

So, for example, his 291 in favorable conditions at the Oval
is discounted, while his 4 and 135 in Manchester are upgraded.
His runs against Lillee and Thomson in Aus earn extra credit,
while his 192* and 151 against subpar bowling earn less.  
Overall, more of his runs came in unfavorable conditions than
would be expected, so his career average of 51.xx is upgraded to
53.xx.  The prior gets thin around this point, so the posterior
expectation is a (perhaps conservative) 52.77.

I hope that explains it.

John

 
 
 

Vijay's All-Time Windies XI

Post by Uday Raja » Fri, 01 Feb 2002 02:48:00

(details on Bayesian updating method deleted for brevity)

I find your method very interesting. Do you treat "out" scores
and "not out" scores the same, or do you make an adjustment for
that? And if so, what sort of adjustment?

There are a whole host of "detail" issues, such as the
differences in the empirical distribution (and hence the prior)
with batting position (e.g., the distribution of averages of 9,
10, and Jack should be heavy in the 5-15 region or thereabouts),
the "survivor bias" issue with higher quality bats, and so on,
but I certainly like the idea of updating based on likelihood. Do
you have a website, or have you posted "adjusted" averages for
players?

 
 
 

Vijay's All-Time Windies XI

Post by Ian Galbrai » Fri, 01 Feb 2002 11:05:12


snip explanation]

:I hope that explains it.

I got the gist of it, thanks John.

--
Ian Galbraith

"Only those whose lives are brief can imagine that love is eternal. You
should embrace that remarkable illusion, it may be the greatest gift your
race has ever received." Babylon 5

 
 
 

Vijay's All-Time Windies XI

Post by Moby » Fri, 01 Feb 2002 13:04:36

Quote:


> (details on Bayesian updating method deleted for brevity)

<snip>

Piggy backing.

There was something bothering me about the Andy G. adjusted average and I
think I've worked it out.

Surely the hundred on debut counts for a little extra than any old
hundred.

An adjusted average of 34 would stick Andy right at the bottom of the list
of players who made such a score on debut I think (blew-it would
be the only one lower??)

Is the result of 34 really a function of the score (100+), or is it much
more likely a simple function of the average of all test averages?

Perhaps the best way to measure the possible average of a batsman who has
made a solitary hundred is to measure it only against batsmen who have
also made hundreds??

Does that give a markedly different reading.

Moby
Rambling.

 
 
 

Vijay's All-Time Windies XI

Post by Sundar Subramania » Fri, 01 Feb 2002 22:36:59

I had posted a reply yesterday, but don't see it appearing. Here is a
repost.

Quote:


> (details on Bayesian updating method deleted for brevity)

> I find your method very interesting. Do you treat "out" scores
> and "not out" scores the same, or do you make an adjustment for
> that? And if so, what sort of adjustment?

The not out scores can be treated in a straightforward manner using
the idea of right censoring in survival analysis. The time-honored
Kaplan--Meier estimator (KME) of a distribution function, the
counterpart of the empirical distribution function, is then the
estimator of choice when censoring is present. To be pedantic, the
KME is strongly consistent (with a very very fast rate), asymptotically
normal, and asymptotically efficient in the sense of achieving the
semiparametric (information) bound. In short, one cannot hope to
better that estimate of a distribution function.

The 0.5th quantile of the KME of the distribution function can be used
as the "average". It should certainly be better than the usual average.

I toyed with idea of computing the median score of Sachin using the
KME two summers back, but then got bored with it. Any standard software
should allow one to analyze censored survival data. May be I should do
that sometime (may be next summer).

Alternatively, one might view this problem in a regression setting,
incorporating covariates like the strength of the opposition, impact on
the match, second innings or first innnings [just to please RK, that is
:-) ]
etc. A plethora of good fitting semiparametric models are available in
the
survival analysis literature, but there may be the problem of
mis-specification.

Quote:
> There are a whole host of "detail" issues, such as the
> differences in the empirical distribution (and hence the prior)
> with batting position (e.g., the distribution of averages of 9,
> 10, and Jack should be heavy in the 5-15 region or thereabouts),
> the "survivor bias" issue with higher quality bats, and so on,
> you have a website, or have you posted "adjusted" averages for
> players?

The beauty of adjusting through the idea of censoring is in its
nonparametric nature, i.e. the censoring distribution (as well as the
failure time distribution) is completely unspecified. One needs to
assume, though, that the failure time random variable and the censoring
variable are independent to obviate identifiability problems.

Best,
Sundar