Discussion:
Top 20s - Test players debuting before the Great War
(too old to reply)
David North
2003-10-04 06:48:06 UTC
Permalink
Batting
1 796.6 Jack Hobbs (Eng)
2 761.6 John Lyons (Aus)
3 748 Ranji Ranjitsinhji (Eng)
4 738.8 Frank Iredale (Aus)
5 673.9 Andrew Stoddart (Eng)
6 673.5 Herbert Taylor (SA)
7 667.4 Clem Hill (Aus)
8 651.2 Robert Abel (Eng)
9 649 Allan Steel (Eng)
10 645 Stanley Jackson (Eng)
11 639.8 Walter Read (Eng)
12 631.7 Reggie Spooner (Eng)
13 628.5 WG Grace (Eng)
14 627.8 Warren Bardsley (Aus)
15 626 Vernon Ransford (Aus)
16 622.9 Arthur Shrewsbury (Eng)
17 622.4 Reggie Duff (Aus)
18 602.3 Gordon White (SA)
19 600.8 Johnny Tyldesley (Eng)
20 590.7 Aubrey Faulkner (SA)

Bowling
1 800.8 Charlie Turner (Aus)
2 784.2 Frank Foster (Eng)
3 761.5 Tom Richardson (Eng)
4 734.7 Sydney Barnes (Eng)
5 722.8 Bill Whitty (Aus)
6 715.2 George Lohmann (Eng)
7 713.3 Bobby Peel (Eng)
8 688.5 Frederick Spofforth (Aus)
9 667.5 Joey Palmer (Aus)
10 634.4 Jack Saunders (Aus)
11 634 Monty Noble (Aus)
12 608.2 Billy Bates (Eng)
13 593.8 Johnny Briggs (Eng)
14 587.6 Jack T Hearne (Eng)
15 576.7 Tibby Cotter (Aus)
16 566.5 Charlie Blythe (Eng)
17 553.2 Hugh Trumble (Aus)
18 551.2 Jimmy Blanckenberg (SA)
19 525.3 Bert Vogler (SA)
20 514 Wilfred Rhodes (Eng)

All-Round
1 1260.7 Frank Foster (Eng)
2 1163.4 Charlie Turner (Aus)
3 1090.8 Monty Noble (Aus)
4 1083.9 Aubrey Faulkner (SA)
5 1061.2 Billy Bates (Eng)
6 1060.5 Allan Steel (Eng)
7 982.7 George Lohmann (Eng)
8 957.4 George Ulyett (Eng)
9 922.3 Bobby Peel (Eng)
10 914.7 Tom Richardson (Eng)
11 910.2 Johnny Briggs (Eng)
12 900.1 Jimmy Sinclair (SA)
13 897.2 Frank Woolley (Eng)
14 896.8 Wilfred Rhodes (Eng)
15 875.5 John Lyons (Aus)
16 862.8 Bill Whitty (Aus)
17 858.7 Billy Barnes (Eng)
18 856.6 Sydney Barnes (Eng)
19 855.6 Warwick Armstrong (Aus)
20 844.7 Frederick Spofforth (Aus)

Minimum 10 Tests

I have modified my rating system since I posted the top tens for 19th Century players.

The previous method was to take the average of each players PwC ratings from their 10th Test
onwards, on the basis that I had estimated that ratings tended to level off after 10 Tests.

On the data I have entered so far, bowling ratings tend to settle after 13 Tests (i.e. the average
rating for all players after their 13th Test is just higher than the average for all Tests from the
14th onwards). However, the batting ratings do not do the same until the 37th Test.

Apart from that, the method did not properly credit or penalise players for their performances in
their first few Tests, which is why, for instance, Clem Hill was rated well ahead in the batting,
despite having made only 30 runs in his first 6 Test innings.

The modified system uses a player's average rating after each of his Tests, including the first ten,
and then makes an adjustment to allow for lower initial ratings.

As an example, WG Grace has an average batting rating of 548.5 over his 22 Tests. The average of all
ratings for all players (up to 1914) up to their 22nd Test is 292.5, while the average including
Tests after the 22nd is 335.1. Therefore, Grace's average is multiplied by 335.1/292.5 to give an
overall rating of 628.5.

The all-round ratings are arrived at simply by adding together the players' overall batting and
bowling ratings.

Because the PwC ratings for all players are used in the calculation of each player's overall rating,
the ratings will change to some extent as more data is added, and players whose ratings are very
similar may change places on the list.

The lowest-rated batsman by far is Jack Saunders (Aus) with 23.6. No-one else has less than 80. The
lowest-rated bowler, of those who took at least one wicket, is Jack Hobbs (Eng) with 2.8.

A number players without the required 10 Tests would have higher ratings than those at the top of
the lists. In the batting, the highest would be Charles Bannerman (Aus) with 931.0 from 3 Tests, but
only Tip Foster (Eng, 818.0 from 8 Tests) would have a higher rating than Hobbs from more than 3
Tests. In the bowling, Frederick Martin (Eng) would score 1049.1 from 2 Tests; of those with more
than 2 Tests, Herbert Hordern (SA, 850.4 from 7 Tests), George Simpson-Hayward (Eng, 837.3 from 5
Tests) and Walter Lees (Eng, 813.5 from 5 Tests) would rating higher than Turner. The only
all-rounder who would rate higher than Frank Foster is Albert Trott (Aus/Eng) with 1492.7 from 5
Tests.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
John Hall
2003-10-04 10:10:10 UTC
Permalink
Post by David North
Batting
1 796.6 Jack Hobbs (Eng)
2 761.6 John Lyons (Aus)
3 748 Ranji Ranjitsinhji (Eng)
4 738.8 Frank Iredale (Aus)
5 673.9 Andrew Stoddart (Eng)
6 673.5 Herbert Taylor (SA)
7 667.4 Clem Hill (Aus)
8 651.2 Robert Abel (Eng)
9 649 Allan Steel (Eng)
10 645 Stanley Jackson (Eng)
11 639.8 Walter Read (Eng)
12 631.7 Reggie Spooner (Eng)
13 628.5 WG Grace (Eng)
14 627.8 Warren Bardsley (Aus)
15 626 Vernon Ransford (Aus)
16 622.9 Arthur Shrewsbury (Eng)
17 622.4 Reggie Duff (Aus)
18 602.3 Gordon White (SA)
19 600.8 Johnny Tyldesley (Eng)
20 590.7 Aubrey Faulkner (SA)
Bowling
1 800.8 Charlie Turner (Aus)
2 784.2 Frank Foster (Eng)
3 761.5 Tom Richardson (Eng)
4 734.7 Sydney Barnes (Eng)
5 722.8 Bill Whitty (Aus)
6 715.2 George Lohmann (Eng)
7 713.3 Bobby Peel (Eng)
8 688.5 Frederick Spofforth (Aus)
9 667.5 Joey Palmer (Aus)
10 634.4 Jack Saunders (Aus)
11 634 Monty Noble (Aus)
12 608.2 Billy Bates (Eng)
13 593.8 Johnny Briggs (Eng)
14 587.6 Jack T Hearne (Eng)
15 576.7 Tibby Cotter (Aus)
16 566.5 Charlie Blythe (Eng)
17 553.2 Hugh Trumble (Aus)
18 551.2 Jimmy Blanckenberg (SA)
19 525.3 Bert Vogler (SA)
20 514 Wilfred Rhodes (Eng)
All-Round
1 1260.7 Frank Foster (Eng)
2 1163.4 Charlie Turner (Aus)
3 1090.8 Monty Noble (Aus)
4 1083.9 Aubrey Faulkner (SA)
5 1061.2 Billy Bates (Eng)
6 1060.5 Allan Steel (Eng)
7 982.7 George Lohmann (Eng)
8 957.4 George Ulyett (Eng)
9 922.3 Bobby Peel (Eng)
10 914.7 Tom Richardson (Eng)
11 910.2 Johnny Briggs (Eng)
12 900.1 Jimmy Sinclair (SA)
13 897.2 Frank Woolley (Eng)
14 896.8 Wilfred Rhodes (Eng)
15 875.5 John Lyons (Aus)
16 862.8 Bill Whitty (Aus)
17 858.7 Billy Barnes (Eng)
18 856.6 Sydney Barnes (Eng)
19 855.6 Warwick Armstrong (Aus)
20 844.7 Frederick Spofforth (Aus)
Minimum 10 Tests
<Snip explanation of methods.>

Some thoughts:

It looks as though for all-round you just add the batting a bowling
figures. This produces misleading results, in that an outstanding
batsman or bowler may be high in the list even when their other
discipline is mediocre. Tom Richardson is a good example. 761.5 for
bowling and presumably 153.2 for batting, giving a total of 914.7.
(Turner and Barnes are other examples.)

If you averaged performances over the whole period, it explains why
Rhodes comes out comparatively poorly as a bowler. Great bowling figures
early on, but in the years just before WW1 he was playing primarily as a
batsman.

I'm surprised that Blythe and Trumble aren't higher amongst the bowlers.
Conversely, I'm surprised that Whitty and Blanckenberg are so high.

Amongst the batsmen, Lyons, Iredale and White are higher than I'd have
expected. Where do Hayward, Fry and Maclaren rate? What figure would you
have got for Grace if you'd ignored Tests after say 1890? (Of course,
his best years were over even by the time of his first Test in 1880.)
--
John Hall

"I don't even butter my bread; I consider that cooking."
Katherine Cebrian
Cricketislife!
2003-10-04 13:13:10 UTC
Permalink
On Sat, 4 Oct 2003 11:10:10 +0100, John Hall <***@jhall.co.uk>
wrote:>
Post by John Hall
<Snip explanation of methods.>
It looks as though for all-round you just add the batting a bowling
figures. This produces misleading results, in that an outstanding
batsman or bowler may be high in the list even when their other
discipline is mediocre. Tom Richardson is a good example. 761.5 for
bowling and presumably 153.2 for batting, giving a total of 914.7.
(Turner and Barnes are other examples.)
If you averaged performances over the whole period, it explains why
Rhodes comes out comparatively poorly as a bowler. Great bowling figures
early on, but in the years just before WW1 he was playing primarily as a
batsman.
I'm surprised that Blythe and Trumble aren't higher amongst the bowlers.
Conversely, I'm surprised that Whitty and Blanckenberg are so high.
Amongst the batsmen, Lyons, Iredale and White are higher than I'd have
expected. Where do Hayward, Fry and Maclaren rate? What figure would you
have got for Grace if you'd ignored Tests after say 1890? (Of course,
his best years were over even by the time of his first Test in 1880.)
John, you cud have cross posted as well like this! Now the usc guys
can read ur thoughts as well.
Uday Rajan
2003-10-04 13:39:40 UTC
Permalink
Post by John Hall
It looks as though for all-round you just add the batting a bowling
figures. This produces misleading results, in that an outstanding
batsman or bowler may be high in the list even when their other
discipline is mediocre. Tom Richardson is a good example. 761.5 for
bowling and presumably 153.2 for batting, giving a total of 914.7.
(Turner and Barnes are other examples.)
Apparently, PwC multiplies the two ratings and divides by 1000. I
won't attempt an example, because I have a nasty affliction to
taking the square root of the result instead.
Post by John Hall
I'm surprised that Blythe and Trumble aren't higher amongst the bowlers.
Conversely, I'm surprised that Whitty and Blanckenberg are so high.
Amongst the batsmen, Lyons, Iredale and White are higher than I'd have
expected. Where do Hayward, Fry and Maclaren rate? What figure would you
have got for Grace if you'd ignored Tests after say 1890? (Of course,
his best years were over even by the time of his first Test in 1880.)
Agree with most of this. Where is Trumper? I skimmed the Hill vs
Trumper thread, and it's certainly arguable that Hill was a
better player than Trumper. But also Lyons, Iredale, Ransford
(looking at just Australian players)? Who are these people? Duff
I've at least heard of. I'm still not quite sure how the PwC
ratings work, but, as Cardus wrote in selecting Hobbs for a World
XI based on names, I would be willing to go any lengths of
subterfuge to get Trumper into my pre-War top 5 list, leave alone
a top 20.
David North
2003-10-04 15:01:28 UTC
Permalink
Post by Uday Rajan
Post by John Hall
It looks as though for all-round you just add the batting a bowling
figures. This produces misleading results, in that an outstanding
batsman or bowler may be high in the list even when their other
discipline is mediocre. Tom Richardson is a good example. 761.5 for
bowling and presumably 153.2 for batting, giving a total of 914.7.
(Turner and Barnes are other examples.)
Apparently, PwC multiplies the two ratings and divides by 1000. I
won't attempt an example, because I have a nasty affliction to
taking the square root of the result instead.
I made my revision before reading this. Dividing by 1000 would give a highest rating of 373.7, so
I'll stick with square roots, which give ratings more in line with those for batting and bowling,
not that it makes any difference to the rankings.
Post by Uday Rajan
Post by John Hall
I'm surprised that Blythe and Trumble aren't higher amongst the bowlers.
Conversely, I'm surprised that Whitty and Blanckenberg are so high.
Amongst the batsmen, Lyons, Iredale and White are higher than I'd have
expected. Where do Hayward, Fry and Maclaren rate? What figure would you
have got for Grace if you'd ignored Tests after say 1890? (Of course,
his best years were over even by the time of his first Test in 1880.)
Agree with most of this. Where is Trumper?
I knew someone would ask that. He's 21st on 586.6. Let down by a poor start (apart from a century in
his second Tests) - his average was only 27 after 13 Tests, so his PwC rating at that point was
still under 400. He also had a fairly poor run in the 3 series from 1905 and 1909, when he averaged
26.96.
Post by Uday Rajan
I skimmed the Hill vs
Trumper thread, and it's certainly arguable that Hill was a
better player than Trumper. But also Lyons, Iredale, Ransford
(looking at just Australian players)? Who are these people?
Lyons and Iredale I've covered in another post. Ransford averaged 1.2 runs less than Trumper, but
against England, he averaged 38.82 (in 15 Tests) compared to Trumper's 32.79.
Post by Uday Rajan
Duff
I've at least heard of. I'm still not quite sure how the PwC
ratings work, but, as Cardus wrote in selecting Hobbs for a World
XI based on names, I would be willing to go any lengths of
subterfuge to get Trumper into my pre-War top 5 list, leave alone
a top 20.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-04 14:33:15 UTC
Permalink
Post by John Hall
Post by David North
Batting
1 796.6 Jack Hobbs (Eng)
2 761.6 John Lyons (Aus)
3 748 Ranji Ranjitsinhji (Eng)
4 738.8 Frank Iredale (Aus)
5 673.9 Andrew Stoddart (Eng)
6 673.5 Herbert Taylor (SA)
7 667.4 Clem Hill (Aus)
8 651.2 Robert Abel (Eng)
9 649 Allan Steel (Eng)
10 645 Stanley Jackson (Eng)
11 639.8 Walter Read (Eng)
12 631.7 Reggie Spooner (Eng)
13 628.5 WG Grace (Eng)
14 627.8 Warren Bardsley (Aus)
15 626 Vernon Ransford (Aus)
16 622.9 Arthur Shrewsbury (Eng)
17 622.4 Reggie Duff (Aus)
18 602.3 Gordon White (SA)
19 600.8 Johnny Tyldesley (Eng)
20 590.7 Aubrey Faulkner (SA)
Bowling
1 800.8 Charlie Turner (Aus)
2 784.2 Frank Foster (Eng)
3 761.5 Tom Richardson (Eng)
4 734.7 Sydney Barnes (Eng)
5 722.8 Bill Whitty (Aus)
6 715.2 George Lohmann (Eng)
7 713.3 Bobby Peel (Eng)
8 688.5 Frederick Spofforth (Aus)
9 667.5 Joey Palmer (Aus)
10 634.4 Jack Saunders (Aus)
11 634 Monty Noble (Aus)
12 608.2 Billy Bates (Eng)
13 593.8 Johnny Briggs (Eng)
14 587.6 Jack T Hearne (Eng)
15 576.7 Tibby Cotter (Aus)
16 566.5 Charlie Blythe (Eng)
17 553.2 Hugh Trumble (Aus)
18 551.2 Jimmy Blanckenberg (SA)
19 525.3 Bert Vogler (SA)
20 514 Wilfred Rhodes (Eng)
All-Round
1 1260.7 Frank Foster (Eng)
2 1163.4 Charlie Turner (Aus)
3 1090.8 Monty Noble (Aus)
4 1083.9 Aubrey Faulkner (SA)
5 1061.2 Billy Bates (Eng)
6 1060.5 Allan Steel (Eng)
7 982.7 George Lohmann (Eng)
8 957.4 George Ulyett (Eng)
9 922.3 Bobby Peel (Eng)
10 914.7 Tom Richardson (Eng)
11 910.2 Johnny Briggs (Eng)
12 900.1 Jimmy Sinclair (SA)
13 897.2 Frank Woolley (Eng)
14 896.8 Wilfred Rhodes (Eng)
15 875.5 John Lyons (Aus)
16 862.8 Bill Whitty (Aus)
17 858.7 Billy Barnes (Eng)
18 856.6 Sydney Barnes (Eng)
19 855.6 Warwick Armstrong (Aus)
20 844.7 Frederick Spofforth (Aus)
Minimum 10 Tests
<Snip explanation of methods.>
It looks as though for all-round you just add the batting a bowling
figures.
Correct.
Post by John Hall
This produces misleading results, in that an outstanding
batsman or bowler may be high in the list even when their other
discipline is mediocre. Tom Richardson is a good example. 761.5 for
bowling and presumably 153.2 for batting, giving a total of 914.7.
(Turner and Barnes are other examples.)
I take your point. I have made a quick change - instead of adding the two figures, I have now
multiplied them and then taken the square root. This should favour players with similar levels of
ability in both disciplines. The revised top 20 is as follows:

1 611.3 Frank Foster (Eng)
2 539.7 Aubrey Faulkner (SA)
3 538.9 Charlie Turner (Aus)
4 538.2 Monty Noble (Aus)
5 524.9 Billy Bates (Eng)
6 516.8 Allan Steel (Eng)
7 477.3 George Ulyett (Eng)
8 443.6 Wilfred Rhodes (Eng)
9 443 Jimmy Sinclair (SA)
10 437.4 George Lohmann (Eng)
11 433.5 Johnny Briggs (Eng)
12 432.4 Frank Woolley (Eng)
13 429.2 Billy Barnes (Eng)
14 410.5 Charles Macartney (Aus)
15 410.4 Johnny Douglas (Eng)
16 409.7 Warwick Armstrong (Aus)
17 405.9 Claude Carter (SA)
18 401.1 Charlie Llewellyn (SA)
19 399.8 Hugh Trumble (Aus)
20 391.4 Charles Kelleway (Aus)

Richardson now ranks 30th out of 80, and Sydney Barnes 43rd. Turner only drops to 3rd, as his
batting rating, at 362.6 (68th of 114), is much better than that of Richardson or Barnes.

Of those with less than 10 Tests, Albert Trott (744.8) would still be the only one to have a higher
all-round rating than Frank Foster.
Post by John Hall
If you averaged performances over the whole period, it explains why
Rhodes comes out comparatively poorly as a bowler. Great bowling figures
early on, but in the years just before WW1 he was playing primarily as a
batsman.
Yes. Rhodes could have been as high as 8th in the bowling if he had retired after the first 17 of
his 58 Tests, at the end of the 1905 season.

Mat Runs HS BatAv 100 50 W BB BowlAv 5w Ct St
Up to 1905 17 357 40* 25.50 0 0 76 8/68 19.46 5 16 0
After 1905 41 1968 179 31.23 2 11 51 5/83 38.15 1 44 0

Total 58 2325 179 30.19 2 11 127 8/68 26.96 6 60 0
Post by John Hall
I'm surprised that Blythe and Trumble aren't higher amongst the bowlers.
As I mentioned when I posted my pre-1900 ratings, my method tends to under-rate players who finish
their careers with a PwC rating well above their average rating, which is the case with both of
these, compared to those who continued for a while when their performances started to decline. I
would welcome any sensible suggestions for a method to compensate for this.

Trumble also had a slow start in Tests - 4 wickets at 55.25 in his first 4 Tests, 17 at 36.23 in his
first 8. Before taking 6 wickets in each innings at The Oval in1896, he had not taken more than
3 in an innings.
Post by John Hall
Conversely, I'm surprised that Whitty and Blanckenberg are so high.
Whitty's rating was possibly helped by playing a lot of his Tests against South Africa (although PwC
is supposed to allow for the strength of the opposition) - 50 wickets at 17.50 in 8 Tests, compared
with a moderate 15 at 33.20 in 6 Tests against England. I would agree that Blanckenberg's position
is difficult to explain.
Post by John Hall
Amongst the batsmen, Lyons, Iredale and White are higher than I'd have
expected.
Many of Lyons's matches were very low-scoring, so his performances were probably much better than
his stats would suggest. Iredale averaged 36.68; no other Australian who played in the same period
averaged over 31, except Hill. White is difficult to judge - he was excellent at home, but very poor
in England (as was Blanckenberg).
Post by John Hall
Where do Hayward, Fry and Maclaren rate?
Hayward 24th, 569.8
MacLaren 30th, 530.9
Fry 43rd, 464.9
Post by John Hall
What figure would you
have got for Grace if you'd ignored Tests after say 1890? (Of course,
his best years were over even by the time of his first Test in 1880.)
Up to 1890 (13 Tests), Grace would have had a rating of 651.9 (8th).
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
samarth harish shah
2003-10-04 18:11:21 UTC
Permalink
Post by David North
Batting
1 796.6 Jack Hobbs (Eng)
2 761.6 John Lyons (Aus)
3 748 Ranji Ranjitsinhji (Eng)
4 738.8 Frank Iredale (Aus)
5 673.9 Andrew Stoddart (Eng)
6 673.5 Herbert Taylor (SA)
7 667.4 Clem Hill (Aus)
8 651.2 Robert Abel (Eng)
9 649 Allan Steel (Eng)
10 645 Stanley Jackson (Eng)
11 639.8 Walter Read (Eng)
12 631.7 Reggie Spooner (Eng)
13 628.5 WG Grace (Eng)
14 627.8 Warren Bardsley (Aus)
15 626 Vernon Ransford (Aus)
16 622.9 Arthur Shrewsbury (Eng)
17 622.4 Reggie Duff (Aus)
18 602.3 Gordon White (SA)
19 600.8 Johnny Tyldesley (Eng)
20 590.7 Aubrey Faulkner (SA)
Bowling
1 800.8 Charlie Turner (Aus)
2 784.2 Frank Foster (Eng)
3 761.5 Tom Richardson (Eng)
4 734.7 Sydney Barnes (Eng)
5 722.8 Bill Whitty (Aus)
6 715.2 George Lohmann (Eng)
7 713.3 Bobby Peel (Eng)
8 688.5 Frederick Spofforth (Aus)
9 667.5 Joey Palmer (Aus)
10 634.4 Jack Saunders (Aus)
11 634 Monty Noble (Aus)
12 608.2 Billy Bates (Eng)
13 593.8 Johnny Briggs (Eng)
14 587.6 Jack T Hearne (Eng)
15 576.7 Tibby Cotter (Aus)
16 566.5 Charlie Blythe (Eng)
17 553.2 Hugh Trumble (Aus)
18 551.2 Jimmy Blanckenberg (SA)
19 525.3 Bert Vogler (SA)
20 514 Wilfred Rhodes (Eng)
All-Round
1 1260.7 Frank Foster (Eng)
2 1163.4 Charlie Turner (Aus)
3 1090.8 Monty Noble (Aus)
4 1083.9 Aubrey Faulkner (SA)
5 1061.2 Billy Bates (Eng)
6 1060.5 Allan Steel (Eng)
7 982.7 George Lohmann (Eng)
8 957.4 George Ulyett (Eng)
9 922.3 Bobby Peel (Eng)
10 914.7 Tom Richardson (Eng)
11 910.2 Johnny Briggs (Eng)
12 900.1 Jimmy Sinclair (SA)
13 897.2 Frank Woolley (Eng)
14 896.8 Wilfred Rhodes (Eng)
15 875.5 John Lyons (Aus)
16 862.8 Bill Whitty (Aus)
17 858.7 Billy Barnes (Eng)
18 856.6 Sydney Barnes (Eng)
19 855.6 Warwick Armstrong (Aus)
20 844.7 Frederick Spofforth (Aus)
Minimum 10 Tests
I have modified my rating system since I posted the top tens for 19th Century players.
The previous method was to take the average of each players PwC ratings from their 10th Test
onwards, on the basis that I had estimated that ratings tended to level off after 10 Tests.
On the data I have entered so far, bowling ratings tend to settle after 13 Tests (i.e. the average
rating for all players after their 13th Test is just higher than the average for all Tests from the
14th onwards). However, the batting ratings do not do the same until the 37th Test.
Apart from that, the method did not properly credit or penalise players for their performances in
their first few Tests, which is why, for instance, Clem Hill was rated well ahead in the batting,
despite having made only 30 runs in his first 6 Test innings.
The modified system uses a player's average rating after each of his Tests, including the first ten,
and then makes an adjustment to allow for lower initial ratings.
As an example, WG Grace has an average batting rating of 548.5 over his 22 Tests. The average of all
ratings for all players (up to 1914) up to their 22nd Test is 292.5, while the average including
Tests after the 22nd is 335.1. Therefore, Grace's average is multiplied by 335.1/292.5 to give an
overall rating of 628.5.
The all-round ratings are arrived at simply by adding together the players' overall batting and
bowling ratings.
Because the PwC ratings for all players are used in the calculation of each player's overall rating,
the ratings will change to some extent as more data is added, and players whose ratings are very
similar may change places on the list.
The lowest-rated batsman by far is Jack Saunders (Aus) with 23.6. No-one else has less than 80. The
lowest-rated bowler, of those who took at least one wicket, is Jack Hobbs (Eng) with 2.8.
A number players without the required 10 Tests would have higher ratings than those at the top of
the lists. In the batting, the highest would be Charles Bannerman (Aus) with 931.0 from 3 Tests, but
only Tip Foster (Eng, 818.0 from 8 Tests) would have a higher rating than Hobbs from more than 3
Tests. In the bowling, Frederick Martin (Eng) would score 1049.1 from 2 Tests; of those with more
than 2 Tests, Herbert Hordern (SA, 850.4 from 7 Tests), George Simpson-Hayward (Eng, 837.3 from 5
Tests) and Walter Lees (Eng, 813.5 from 5 Tests) would rating higher than Turner. The only
all-rounder who would rate higher than Frank Foster is Albert Trott (Aus/Eng) with 1492.7 from 5
Tests.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Did a word search for "Trumper" on this post on greatest cricketers to
have debuted before the great war. Trumper not found.

-Samarth.
Shripathi Kamath
2003-10-04 19:12:21 UTC
Permalink
"David North" <***@abbeymanor.fsbusiness.co.uk> wrote in message news:bllqee$dl7uu$***@ID-182826.news.uni-berlin.de...

<snip>
Post by David North
The modified system uses a player's average rating after each of his
Tests, including the first ten,
Post by David North
and then makes an adjustment to allow for lower initial ratings.
As an example, WG Grace has an average batting rating of 548.5 over his 22
Tests. The average of all
Post by David North
ratings for all players (up to 1914) up to their 22nd Test is 292.5, while
the average including
Post by David North
Tests after the 22nd is 335.1. Therefore, Grace's average is multiplied by
335.1/292.5 to give an
Post by David North
overall rating of 628.5.
I wonder if that does not undo some of the normalization that PwC performs
so painstakingly. Firstly, it is reasonable to use the PwC as it is, after
all, it is said to account for low or high initial ratings with appropriate
dampening factors. Secondly, multiplying Grace's rating by a factor of how
others in the same era did before and after Test n is arbitrary, and rather
redundant, irrelevant or both, imho. The ratings themselves take care of
that, don't they? Grace's rating, after all, is and should be based on his
efforts given the strength of the opposition, not with an additional
weightage given to how the others did--that has already been accounted for.

I think a simple average would be accurately reflective of a player based on
the PwC, if PwC is the way to go.

I understand if you have trouble reconciling their reputation as you recall
it and what PwC yields. For all but the best performers (and I do mean
'best' in an averages+reputation sort of a way), it produces, for me at
least, a fair bit of counter-intuitive results.

Trumper, by reputation should have made the top, or at least I'd lost my
house if I had bet on that. Similarly, I'd expected Lohmann and Barnes to
finish higher-- Barnes no lower than #1.

As another experiment, a little while back, I had asked the good folks of
rsc to rate a few of not-so-great bowlers of our era--Gough, Streak,
Caddick, Srinath, and Vaas. (I ensured that all contestants had played at
least 30 tests)

Nearly everybody (including I, who did not put forth a ranking) ranked Gough
as the best. Nearly everybody didn't think that highly of Vaas.

I then calculated average PwC ratings for them:

As of October 2, 2003:

Gough 615
Streak 655
Srinath 539
Vaas 588
Caddick 568

And then I did three more for fun:

Cairns 483
Gillespie 606
Kallis 466

According to the above, Streak is ahead of Gough and Gillespie by a bit.
I'd lost my house twice after betting on that.

To return to the thread, if it is not too much trouble, and you have the
data available in a spreadsheet, could you please repost just the average
PwC ratings for the players? Without any massaging, that is.

Thanks,
--
Shripathi Kamath
David North
2003-10-04 21:53:35 UTC
Permalink
Post by Shripathi Kamath
<snip>
Post by David North
The modified system uses a player's average rating after each of his
Tests, including the first ten,
Post by David North
and then makes an adjustment to allow for lower initial ratings.
As an example, WG Grace has an average batting rating of 548.5 over his 22
Tests. The average of all
Post by David North
ratings for all players (up to 1914) up to their 22nd Test is 292.5, while
the average including
Post by David North
Tests after the 22nd is 335.1. Therefore, Grace's average is multiplied by
335.1/292.5 to give an
Post by David North
overall rating of 628.5.
I wonder if that does not undo some of the normalization that PwC performs
so painstakingly. Firstly, it is reasonable to use the PwC as it is, after
all, it is said to account for low or high initial ratings with appropriate
dampening factors. Secondly, multiplying Grace's rating by a factor of how
others in the same era did before and after Test n is arbitrary, and rather
redundant, irrelevant or both, imho. The ratings themselves take care of
that, don't they? Grace's rating, after all, is and should be based on his
efforts given the strength of the opposition, not with an additional
weightage given to how the others did--that has already been accounted for.
I think a simple average would be accurately reflective of a player based on
the PwC, if PwC is the way to go.
I understand if you have trouble reconciling their reputation as you recall
it and what PwC yields. For all but the best performers (and I do mean
'best' in an averages+reputation sort of a way), it produces, for me at
least, a fair bit of counter-intuitive results.
--snip--

I think you misunderstand the reason for the adjustment.

Firstly, the adjustment will not be made according to how others *in the same era* did; it will be
made according to how other throughout Test history did. The only reason I specified "up to 1914"
was because so far I have only entered the data into the spreadsheet for players debuting up to
1914.

The reason for the adjustment is that PwC deliberately give players only a percentage of their
performance points until they have fully qualified, which is after 40 innings for batsmen and 100
wickets for bowlers. Therefore, players will generally have lower ratings up to this point. This
will have a greater effect on players who played fewer Tests. If two players perform consistently at
the same level, but player A plays 25 Tests and player B plays 50 Tests, player A will have a lower
average rating than player B, even though he is just as good. It is this effect that I am trying to
cancel out. If the PwC Web site gave the scale of percentages that they use during the qualifying
period, I might be able to use that to cancel out their adjustment (although it would be tricky, as
I am working from ratings after each Test, not by innings or by wickets), but it does not, so I use
another method to approximate this adjustment reversal.
Post by Shripathi Kamath
To return to the thread, if it is not too much trouble, and you have the
data available in a spreadsheet, could you please repost just the average
PwC ratings for the players? Without any massaging, that is.
OK:

Batting
1 796.3 Jack Hobbs (Eng)
2 660.2 Clem Hill (Aus)
3 655.8 Herbert Taylor (SA)
4 609.4 Warren Bardsley (Aus)
5 590.9 Ranji Ranjitsinhji (Eng)
6 589.6 John Lyons (Aus)
7 579.3 Victor Trumper (Aus)
8 572 Frank Iredale (Aus)
9 568 Frank Woolley (Eng)
10 560.4 Johnny Tyldesley (Eng)
11 550.3 Stanley Jackson (Eng)
12 549.2 Arthur Shrewsbury (Eng)
13 548.5 WG Grace (Eng)
14 548 Joe Darling (Aus)
15 545.7 Warwick Armstrong (Aus)
16 543.3 Reggie Duff (Aus)
17 542.3 Andrew Stoddart (Eng)
18 541.8 Tom Hayward (Eng)
19 534.1 Vernon Ransford (Aus)
20 531.1 Walter Read (Eng)

Bowling
1 735.9 Charlie Turner (Aus)
2 730.4 Sydney Barnes (Eng)
3 681.4 Bobby Peel (Eng)
4 668.4 George Lohmann (Eng)
5 656.7 Tom Richardson (Eng)
6 643.5 Frederick Spofforth (Aus)
7 636.6 Monty Noble (Aus)
8 623.3 Bill Whitty (Aus)
9 615.9 Frank Foster (Eng)
10 613.5 Joey Palmer (Aus)
11 597.4 Johnny Briggs (Eng)
12 556.4 Hugh Trumble (Aus)
13 555.1 Tibby Cotter (Aus)
14 547.1 Jack Saunders (Aus)
15 536.4 Billy Bates (Eng)
16 536.1 Charlie Blythe (Eng)
17 515.2 Jimmy Blanckenberg (SA)
18 514.5 Wilfred Rhodes (Eng)
19 486.2 Aubrey Faulkner (SA)
20 477.6 Jack T Hearne (Eng)

All-Round
1 532.1 Monty Noble (Aus)
2 507.7 Aubrey Faulkner (SA)
3 467.2 Charlie Turner (Aus)
4 458.2 Frank Foster (Eng)
5 448.9 George Ulyett (Eng)
6 443.3 Wilfred Rhodes (Eng)
7 438.1 Billy Bates (Eng)
8 432.4 Frank Woolley (Eng)
9 421.9 Johnny Briggs (Eng)
10 416.7 Jimmy Sinclair (SA)
11 411.5 Allan Steel (Eng)
12 408.4 Warwick Armstrong (Aus)
13 401.2 Charles Macartney (Aus)
14 391.2 Billy Barnes (Eng)
15 388.2 Hugh Trumble (Aus)
16 385.3 George Lohmann (Eng)
17 380.4 Johnny Douglas (Eng)
18 370.3 Charles Kelleway (Aus)
19 362.7 Dave Nourse (SA)
20 358.6 George Giffen (Aus)

Using this method, we have Clem Hill at no. 2 in the batting, partly because he played a lot of
Tests. Was he really worth a rating nearly 70 points higher than Ranji? I think 80 points behind
Ranji is more like it.

I do think my method gives too little credit to performances later in a player's career, but not
because of the adjustment.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Shripathi Kamath
2003-10-04 22:39:40 UTC
Permalink
<snip>
Post by David North
I think you misunderstand the reason for the adjustment.
Firstly, the adjustment will not be made according to how others *in the
same era* did; it will be
Post by David North
made according to how other throughout Test history did. The only reason I
specified "up to 1914"
Post by David North
was because so far I have only entered the data into the spreadsheet for
players debuting up to
Post by David North
1914.
But why should how others did in their before and after Test n have an
impact on Grace?
Post by David North
The reason for the adjustment is that PwC deliberately give players only a
percentage of their
Post by David North
performance points until they have fully qualified, which is after 40
innings for batsmen and 100
Post by David North
wickets for bowlers. Therefore, players will generally have lower ratings
up to this point.

Which is exactly what it is meant to do, dampen early ratings so that flash
in the pan careers are accounted for appropriately. A Bob Massie or a Jimmy
Adams will otherwise have an unbelievable impact than, say, McGrath, or a
Waugh.
Post by David North
This
will have a greater effect on players who played fewer Tests. If two
players perform consistently at
Post by David North
the same level, but player A plays 25 Tests and player B plays 50 Tests,
player A will have a lower
Post by David North
average rating than player B, even though he is just as good.
Actually, it is harder for a player to maintain such form. A player who
averages, say, 50 runs per innings over 50 tests, should be better regarded
than one who averages the same over only 25 though. (Other things like
opposition etc., being in the same vicinity, of course.)
Post by David North
It is this effect that I am trying to
cancel out. If the PwC Web site gave the scale of percentages that they
use during the qualifying
Post by David North
period, I might be able to use that to cancel out their adjustment
(although it would be tricky, as
Post by David North
I am working from ratings after each Test, not by innings or by wickets),
but it does not, so I use
Post by David North
another method to approximate this adjustment reversal.
Which is why I said that it undoes some of the model assumptions of the PwC
Post by David North
Post by Shripathi Kamath
To return to the thread, if it is not too much trouble, and you have the
data available in a spreadsheet, could you please repost just the average
PwC ratings for the players? Without any massaging, that is.
Thanks for this
Post by David North
Batting
1 796.3 Jack Hobbs (Eng)
2 660.2 Clem Hill (Aus)
3 655.8 Herbert Taylor (SA)
4 609.4 Warren Bardsley (Aus)
5 590.9 Ranji Ranjitsinhji (Eng)
6 589.6 John Lyons (Aus)
7 579.3 Victor Trumper (Aus)
Now that is more like it: Trumper does get noticed now, whereas he was not
even a top 20 before.
Post by David North
8 572 Frank Iredale (Aus)
9 568 Frank Woolley (Eng)
10 560.4 Johnny Tyldesley (Eng)
11 550.3 Stanley Jackson (Eng)
12 549.2 Arthur Shrewsbury (Eng)
13 548.5 WG Grace (Eng)
14 548 Joe Darling (Aus)
15 545.7 Warwick Armstrong (Aus)
16 543.3 Reggie Duff (Aus)
17 542.3 Andrew Stoddart (Eng)
18 541.8 Tom Hayward (Eng)
19 534.1 Vernon Ransford (Aus)
20 531.1 Walter Read (Eng)
Bowling
1 735.9 Charlie Turner (Aus)
2 730.4 Sydney Barnes (Eng)
Darn, still only #2 :-)
Post by David North
3 681.4 Bobby Peel (Eng)
4 668.4 George Lohmann (Eng)
5 656.7 Tom Richardson (Eng)
6 643.5 Frederick Spofforth (Aus)
7 636.6 Monty Noble (Aus)
8 623.3 Bill Whitty (Aus)
9 615.9 Frank Foster (Eng)
10 613.5 Joey Palmer (Aus)
11 597.4 Johnny Briggs (Eng)
12 556.4 Hugh Trumble (Aus)
13 555.1 Tibby Cotter (Aus)
14 547.1 Jack Saunders (Aus)
15 536.4 Billy Bates (Eng)
16 536.1 Charlie Blythe (Eng)
17 515.2 Jimmy Blanckenberg (SA)
18 514.5 Wilfred Rhodes (Eng)
19 486.2 Aubrey Faulkner (SA)
20 477.6 Jack T Hearne (Eng)
All-Round
1 532.1 Monty Noble (Aus)
2 507.7 Aubrey Faulkner (SA)
3 467.2 Charlie Turner (Aus)
4 458.2 Frank Foster (Eng)
5 448.9 George Ulyett (Eng)
6 443.3 Wilfred Rhodes (Eng)
7 438.1 Billy Bates (Eng)
8 432.4 Frank Woolley (Eng)
9 421.9 Johnny Briggs (Eng)
10 416.7 Jimmy Sinclair (SA)
11 411.5 Allan Steel (Eng)
12 408.4 Warwick Armstrong (Aus)
13 401.2 Charles Macartney (Aus)
14 391.2 Billy Barnes (Eng)
15 388.2 Hugh Trumble (Aus)
16 385.3 George Lohmann (Eng)
17 380.4 Johnny Douglas (Eng)
18 370.3 Charles Kelleway (Aus)
19 362.7 Dave Nourse (SA)
20 358.6 George Giffen (Aus)
Using this method, we have Clem Hill at no. 2 in the batting, partly
because he played a lot of
Post by David North
Tests. Was he really worth a rating nearly 70 points higher than Ranji? I
think 80 points behind
Post by David North
Ranji is more like it.
I honestly dunno, but I do know that I like where Trumper ends up now :-)

I do think though that sustaining the same standard over a larger number of
tests should get more recognition.

<snip>

Thanks again for the info
--
Shripathi Kamath
Shripathi Kamath
2003-10-05 00:40:15 UTC
Permalink
wrote in message news:I_Hfb.5534$***@fed1read04...

<snip>
Post by Shripathi Kamath
But why should how others did in their before and after Test n have an
impact on Grace?
OK, my previous answer was not accurate. It is not others' performances
that I am taking into
account, but the way they are rated by PwC.
Post by Shripathi Kamath
Post by David North
The reason for the adjustment is that PwC deliberately give players only a
percentage of their
Post by David North
performance points until they have fully qualified, which is after 40
innings for batsmen and 100
Post by David North
wickets for bowlers. Therefore, players will generally have lower ratings
up to this point.
Which is exactly what it is meant to do, dampen early ratings so that flash
in the pan careers are accounted for appropriately. A Bob Massie or a Jimmy
Adams will otherwise have an unbelievable impact than, say, McGrath, or a
Waugh.
But it also means that Waugh's (or anyone's) first Test is deemed to be
less important than his
last. Why it that valid when rating his career as a whole?
The way I understand it is that, initially, a player's initial performances
do not count fully , but they do count fully after n innings. So Waugh's
initial failures do come into picture. So the answer to your question is
no, Waugh's latter performances are more important, but his initial ones do
have an impact.
Post by Shripathi Kamath
Post by David North
This
will have a greater effect on players who played fewer Tests. If two
players perform consistently at
Post by David North
the same level, but player A plays 25 Tests and player B plays 50 Tests,
player A will have a lower
Post by David North
average rating than player B, even though he is just as good.
Actually, it is harder for a player to maintain such form. A player who
averages, say, 50 runs per innings over 50 tests, should be better regarded
than one who averages the same over only 25 though. (Other things like
opposition etc., being in the same vicinity, of course.)
Let's see. Steve Waugh first achieved his current average at the end of
the 1998/9 series in the
West Indies. If the plane has crashed on the way back to Australia, should
he then have been rated
lower than he actually is?
Not lower than he actually is, but I'd rate him higher today than if he had
stopped playing at the same average back then. I am assuming of course that
since 1998/99 he has played about the same quality of opposition.
To put it another way, how many Tests do you think Herbert Sutcliffe
should have had to play,
averaging 60.73, to be rated as highly as Bradman?
I dunno for sure, but a lot. And just playing more wouldn't do it either,
he would have to have scored a lot more too. Besides, after 20 innings or
so, a batsman's early performances are no longer discounted
--
Shripathi Kamath

<snip>
Rahul Tyagi
2003-10-05 07:10:28 UTC
Permalink
Post by Shripathi Kamath
The way I understand it is that, initially, a player's initial performances
do not count fully , but they do count fully after n innings. So Waugh's
initial failures do come into picture. So the answer to your question is
no, Waugh's latter performances are more important, but his initial ones do
have an impact.
I think you are right in the sense that those early performances are
no longer discounted after you have qualified (40 innings/100
wickets). But, when you start talking about taking the average Pwc
rating over the career of a player you use the discounted ratings of
the early part of his career which, in a sense, is an injustice to
him... I hope I'm clear enough... let me try and take an example to
illustrate what I am trying to say....

Assume there exists a "perfect" batsman - Mr. SRT - who is a 1000
rating batsman. This means that he actually is a perfect batsman from
his first innings to his last, say his 80th innings. So, his
non-discounted rating remains 1000 throughout his career. But before
his 40th innings he will be rated lower. Will probably start with 200
after his first inning and then will keep rising till he is rated 1000
after his 40th. His rating from this point on suggest that he has been
a perfect batsman *over his career* as each rating at each point is
affected by every inning played till that point. Any average rating
calculation should ideally give us a figure of 1000 for Mr. SRT. Now,
if you just take the average of his Pwc ratings it will come out to
something like 900, or maybe even less than that, over his 80
innings.... this seems injustice to a perfect batsman...

I think a simple way of doing this might be to just take the
non-discounted ratings and then take the average (once the player has
become fully qualified).. IIRC Pwc site gives the complete break-up of
discounting factor after each innings/wicket. so one just has to
compensate for that factor and then take the average. this gives us
1000 for Mr. SRT.

I am not so sure about your "longer career at comparable rating means
better player" theory... when we are looking at players who have
played more than a certain number of matches - the qualifying mark -
we should probably assume that their level of performance has become
steady and they would have performed at a similar level had they got a
chance to play as many matches as other players from a different era.

maybe length of a career can be used as a tie-breaker of sorts in
cases with equal (or close) ratings.. but I think a higher average
rating over 60 innings should be deemed to be better than a lower
average rating over 100 innings... Probably multiplying the total
number of innings raised to a reasonably low quotient (close to
zero... 0.10 maybe... or your favorite fraction less than 0.5) to the
average rating should be more reasonable..

Rahul Tyagi
David North
2003-10-05 07:49:02 UTC
Permalink
Post by Rahul Tyagi
Post by Shripathi Kamath
The way I understand it is that, initially, a player's initial performances
do not count fully , but they do count fully after n innings. So Waugh's
initial failures do come into picture. So the answer to your question is
no, Waugh's latter performances are more important, but his initial ones do
have an impact.
I think you are right in the sense that those early performances are
no longer discounted after you have qualified (40 innings/100
wickets). But, when you start talking about taking the average Pwc
rating over the career of a player you use the discounted ratings of
the early part of his career which, in a sense, is an injustice to
him... I hope I'm clear enough... let me try and take an example to
illustrate what I am trying to say....
Assume there exists a "perfect" batsman - Mr. SRT - who is a 1000
rating batsman. This means that he actually is a perfect batsman from
his first innings to his last, say his 80th innings. So, his
non-discounted rating remains 1000 throughout his career. But before
his 40th innings he will be rated lower. Will probably start with 200
after his first inning and then will keep rising till he is rated 1000
after his 40th.
(Nit-picking) It would be more than 200 after the first innings. There seems to be a limit of about
450 after the first Test. Of the players that I have looked at, four had 447 points after their
first Test. Only Tip Foster had 448, after making 287 and 19. The highest first-Test batting rating
from one innings is 417 - Harry Graham (Aus), who made 107 out of 269 at Lord's in 1893.
Post by Rahul Tyagi
His rating from this point on suggest that he has been
a perfect batsman *over his career* as each rating at each point is
affected by every inning played till that point. Any average rating
calculation should ideally give us a figure of 1000 for Mr. SRT. Now,
if you just take the average of his Pwc ratings it will come out to
something like 900, or maybe even less than that, over his 80
innings.... this seems injustice to a perfect batsman...
Yes, that is my point.
Post by Rahul Tyagi
I think a simple way of doing this might be to just take the
non-discounted ratings and then take the average (once the player has
become fully qualified).. IIRC Pwc site gives the complete break-up of
discounting factor after each innings/wicket. so one just has to
compensate for that factor and then take the average. this gives us
1000 for Mr. SRT.
I don't think it does, but if you find it, please post it or let me know where it is.

If you take an average of the ratings after qualification, performances would be given unequal
weightings. Since the rating at any time gives most weight to the most recent innings, you would end
up with the innings when the player becomes fully qualified having the highest weighting, and his
first and last having much lower weightings.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Rahul Tyagi
2003-10-05 17:06:59 UTC
Permalink
Post by David North
(Nit-picking) It would be more than 200 after the first innings. There seems to be a limit of about
450 after the first Test. Of the players that I have looked at, four had 447 points after their
first Test. Only Tip Foster had 448, after making 287 and 19. The highest first-Test batting rating
from one innings is 417 - Harry Graham (Aus), who made 107 out of 269 at Lord's in 1893.
ya ya... it was just an example.. :))
Post by David North
I don't think it does, but if you find it, please post it or let me know where it is.
yes it does, but only for the batsmen. I didnt find the damping
factors for bowlers anywhere on the site.

for batsmen, this is the method

n = 1 to 10 : 40 + 3n
n = 11 to 20 : 55 + 1.5n
n = 21 to 40 : 70 + 0.75n
n = 41 to inf : 100

where n is the number of innings and damping factor is in percentage
Post by David North
If you take an average of the ratings after qualification, performances would be given unequal
weightings. Since the rating at any time gives most weight to the most recent innings, you would end
up with the innings when the player becomes fully qualified having the highest weighting, and his
first and last having much lower weightings.
I hadnt thought of it before, but now that you point it out it really
does seem to be a big problem.

Rahul
David North
2003-10-05 18:50:12 UTC
Permalink
Post by Rahul Tyagi
Post by David North
(Nit-picking) It would be more than 200 after the first innings. There seems to be a limit of
about
Post by Rahul Tyagi
Post by David North
450 after the first Test. Of the players that I have looked at, four had 447 points after their
first Test. Only Tip Foster had 448, after making 287 and 19. The highest first-Test batting
rating
Post by Rahul Tyagi
Post by David North
from one innings is 417 - Harry Graham (Aus), who made 107 out of 269 at Lord's in 1893.
ya ya... it was just an example.. :))
Post by David North
I don't think it does, but if you find it, please post it or let me know where it is.
yes it does, but only for the batsmen. I didnt find the damping
factors for bowlers anywhere on the site.
for batsmen, this is the method
n = 1 to 10 : 40 + 3n
n = 11 to 20 : 55 + 1.5n
n = 21 to 40 : 70 + 0.75n
n = 41 to inf : 100
where n is the number of innings and damping factor is in percentage
Thanks for that. So Harry Graham's 417 was out of a maximum of 430, and Tip Foster's 448 from a
possible 460.

I guess the bowling scale is similar, but from 1 to 100 wickets, so it would reach 70% after 25
wickets and 85% after 50.
Post by Rahul Tyagi
Post by David North
If you take an average of the ratings after qualification, performances would be given unequal
weightings. Since the rating at any time gives most weight to the most recent innings, you would
end
Post by Rahul Tyagi
Post by David North
up with the innings when the player becomes fully qualified having the highest weighting, and
his
Post by Rahul Tyagi
Post by David North
first and last having much lower weightings.
I hadnt thought of it before, but now that you point it out it really
does seem to be a big problem.
Yes, trying to work backwards from the published ratings to get the points for each Test or innings
is pretty mind-boggling.

Using an assumed decay factor per Test for each batsman simplifies things a bit, e.g. WG Grace
played 36 innings in 22 Tests, so we can assume a decay factor per Test of 0.96^(36/22)=0.935, which
should give a fairly close approximation.

Even with this assumption, and ignoring the 1% penalty for missed matches, the algebra is quite
complicated.

If d is the decay factor, p[n] is the number of points for Test n, r[n] is the rating after Test n,
and f is the damping factor, for which we could also assume a number of innings as above.

p[1] = r[1]/f

p[2] = (r[2](1+d)-dp[1])/f

p[3]= (r[3](1+d+d^2)-dp[2]-(d^2)p[1])/f

As you can see, by the time we get to 20, 50 or 100 Tests, the formula will be getting pretty long!
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-05 22:18:04 UTC
Permalink
Post by David North
Post by Rahul Tyagi
Post by David North
(Nit-picking) It would be more than 200 after the first innings. There seems to be a limit of
about
Post by Rahul Tyagi
Post by David North
450 after the first Test. Of the players that I have looked at, four had 447 points after
their
Post by David North
Post by Rahul Tyagi
Post by David North
first Test. Only Tip Foster had 448, after making 287 and 19. The highest first-Test batting
rating
Post by Rahul Tyagi
Post by David North
from one innings is 417 - Harry Graham (Aus), who made 107 out of 269 at Lord's in 1893.
ya ya... it was just an example.. :))
Post by David North
I don't think it does, but if you find it, please post it or let me know where it is.
yes it does, but only for the batsmen. I didnt find the damping
factors for bowlers anywhere on the site.
for batsmen, this is the method
n = 1 to 10 : 40 + 3n
n = 11 to 20 : 55 + 1.5n
n = 21 to 40 : 70 + 0.75n
n = 41 to inf : 100
where n is the number of innings and damping factor is in percentage
Thanks for that. So Harry Graham's 417 was out of a maximum of 430, and Tip Foster's 448 from a
possible 460.
I guess the bowling scale is similar, but from 1 to 100 wickets, so it would reach 70% after 25
wickets and 85% after 50.
Post by Rahul Tyagi
Post by David North
If you take an average of the ratings after qualification, performances would be given unequal
weightings. Since the rating at any time gives most weight to the most recent innings, you
would
Post by David North
end
Post by Rahul Tyagi
Post by David North
up with the innings when the player becomes fully qualified having the highest weighting, and
his
Post by Rahul Tyagi
Post by David North
first and last having much lower weightings.
I hadnt thought of it before, but now that you point it out it really
does seem to be a big problem.
Yes, trying to work backwards from the published ratings to get the points for each Test or
innings
Post by David North
is pretty mind-boggling.
Using an assumed decay factor per Test for each batsman simplifies things a bit, e.g. WG Grace
played 36 innings in 22 Tests, so we can assume a decay factor per Test of 0.96^(36/22)=0.935,
which
Post by David North
should give a fairly close approximation.
Even with this assumption, and ignoring the 1% penalty for missed matches, the algebra is quite
complicated.
If d is the decay factor, p[n] is the number of points for Test n, r[n] is the rating after Test
n,
Post by David North
and f is the damping factor, for which we could also assume a number of innings as above.
p[1] = r[1]/f
p[2] = (r[2](1+d)-dp[1])/f
p[3]= (r[3](1+d+d^2)-dp[2]-(d^2)p[1])/f
As you can see, by the time we get to 20, 50 or 100 Tests, the formula will be getting pretty
long!
Correction to the formulae:

p[2] = r[2](1+d)/f-dp[1]

p[3]= r[3](1+d+d^2)/f-dp[2]-(d^2)p[1]

It turns out that this is easier than I thought, using Excel array formulae, so I will give it a go.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-06 06:42:14 UTC
Permalink
Post by David North
Post by David North
Using an assumed decay factor per Test for each batsman simplifies
things a bit, e.g. WG Grace
Post by David North
played 36 innings in 22 Tests, so we can assume a decay factor per Test
of 0.96^(36/22)=0.935,
which
Post by David North
should give a fairly close approximation.
<nit picking :)> doesnt this mean that the performances in a "1 innings
test" will be underrated?
It seems not. I tried it on Hobbs. The first time he batted once was his 12th Test, when he made 187
against South Africa. This method gives him 1714 points for that Test, but it also gives 1253, 1438
and 1472 for his 8th, 9th and 10th, when he batted twice. This might possibly be partly due to him
having missed two Tests between his 7th and 8th, but must also be because, having batted twice in
every Test up to then, the assumed decay factor is getting too far away from reality. For his 18th
Test, when he batted once and made 4, he gets 872 points! He does get an overall average of 823.1,
which sounds plausible, but I don't think it's going to work.
Post by David North
most of the
matches won by an innings will have underrated performaces by the batsmen of
the winning team. So players playing
in teams that have dominated an era will have many of their performances
underrated.
Australian tail-enders (who probably bat less often than most) would, of course, have a higher decay
factor than Bangladeshi openers.
Post by David North
but yes, I do understand that your averaging out the decay factor makes
things much easier and gives
a good approximation. :)
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-06 19:02:32 UTC
Permalink
Post by David North
Post by David North
Post by David North
Using an assumed decay factor per Test for each batsman simplifies
things a bit, e.g. WG Grace
Post by David North
played 36 innings in 22 Tests, so we can assume a decay factor per Test
of 0.96^(36/22)=0.935,
which
Post by David North
should give a fairly close approximation.
<nit picking :)> doesnt this mean that the performances in a "1 innings
test" will be underrated?
It seems not. I tried it on Hobbs. The first time he batted once was his 12th Test, when he made
187
Post by David North
against South Africa. This method gives him 1714 points for that Test, but it also gives 1253,
1438
Post by David North
and 1472 for his 8th, 9th and 10th, when he batted twice. This might possibly be partly due to him
having missed two Tests between his 7th and 8th, but must also be because, having batted twice in
every Test up to then, the assumed decay factor is getting too far away from reality. For his 18th
Test, when he batted once and made 4, he gets 872 points! He does get an overall average of 823.1,
which sounds plausible, but I don't think it's going to work.
I applied it to the top 30 batsmen from my list, and the top 20 of those come out as follows:

1 823.1 Jack Hobbs (Eng)
2 768.1 Stanley Jackson (Eng)
3 754.3 Frank Iredale (Aus)
4 739.4 Allan Steel (Eng)
5 737.4 Herbert Taylor (SA)
6 737.3 John Lyons (Aus)
7 717.6 Arthur Shrewsbury (Eng)
8 716 Clem Hill (Aus)
9 691.7 Ranji Ranjitsinhji (Eng)
10 690.9 Reggie Duff (Aus)
11 688.7 Vernon Ransford (Aus)
12 673.3 Victor Trumper (Aus)
13 671.4 William Bruce (Aus)
14 667.8 Warren Bardsley (Aus)
15 663.7 Andrew Stoddart (Eng)
16 653.6 Aubrey Faulkner (SA)
17 648.7 Robert Abel (Eng)
18 645.2 Reggie Spooner (Eng)
19 644.8 WG Grace (Eng)
20 637.3 Johnny Tyldesley (Eng)

Some on the scores for individual Tests are absurd, ranging from 2715 for one of Warwick Armstrong's
to -1923 for one of George Gunn's, but amazingly the overall results seem mostly plausible.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-05 07:55:58 UTC
Permalink
Post by David North
Post by Rahul Tyagi
Post by Shripathi Kamath
The way I understand it is that, initially, a player's initial
performances
Post by Rahul Tyagi
Post by Shripathi Kamath
do not count fully , but they do count fully after n innings. So
Waugh's
Post by Rahul Tyagi
Post by Shripathi Kamath
initial failures do come into picture. So the answer to your question
is
Post by Rahul Tyagi
Post by Shripathi Kamath
no, Waugh's latter performances are more important, but his initial ones
do
Post by Rahul Tyagi
Post by Shripathi Kamath
have an impact.
I think you are right in the sense that those early performances are
no longer discounted after you have qualified (40 innings/100
wickets). But, when you start talking about taking the average Pwc
rating over the career of a player you use the discounted ratings of
the early part of his career which, in a sense, is an injustice to
him... I hope I'm clear enough... let me try and take an example to
illustrate what I am trying to say....
Assume there exists a "perfect" batsman - Mr. SRT - who is a 1000
rating batsman. This means that he actually is a perfect batsman from
his first innings to his last, say his 80th innings. So, his
non-discounted rating remains 1000 throughout his career. But before
his 40th innings he will be rated lower. Will probably start with 200
after his first inning and then will keep rising till he is rated 1000
after his 40th. His rating from this point on suggest that he has been
a perfect batsman *over his career* as each rating at each point is
affected by every inning played till that point. Any average rating
calculation should ideally give us a figure of 1000 for Mr. SRT. Now,
if you just take the average of his Pwc ratings it will come out to
something like 900, or maybe even less than that, over his 80
innings.... this seems injustice to a perfect batsman...
Nope, that is not how it works, or at least I do not believe so. The
discounting stops as soon as he reaches his 40th innings. And the
discounting is continually reduced on the way to 40 as well.
So it is more like (example, not exact formula)
After 1 innings, only 10% of his innings thus far count (90% discounted)
After 2 innings, only 11% of his innings thus far count
...
After 38 innings, only 85% of his innings thus far count (15% discounted)
After 39 innings, only 95% of his innings thus far count
After 40 innings, all 100% of his innings thus far count
So unless someone plays far less than 40 innings, early performances count
just as much, but only after that 40 figure is reached.
That's fine, as long as you don't use the discounted figures in a calculation. If you take the
average of a player's ratings over his career, you are doing that.
Post by David North
<snip>
Post by Rahul Tyagi
I am not so sure about your "longer career at comparable rating means
better player" theory... when we are looking at players who have
played more than a certain number of matches - the qualifying mark -
we should probably assume that their level of performance has become
steady and they would have performed at a similar level had they got a
chance to play as many matches as other players from a different era.
a. Tendulkar scores 6000 runs in 75 tests at 50 per and retires.
b. Tendulkar scores 10000 runs in 125 tests at 50 per and retires
Is b. as good a career as a. or is it better?
I am saying b. without a doubt.
Yes, but not by very much.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Rahul Tyagi
2003-10-05 16:55:19 UTC
Permalink
Post by David North
Post by Rahul Tyagi
Assume there exists a "perfect" batsman - Mr. SRT - who is a 1000
rating batsman. This means that he actually is a perfect batsman from
his first innings to his last, say his 80th innings. So, his
non-discounted rating remains 1000 throughout his career. But before
his 40th innings he will be rated lower. Will probably start with 200
after his first inning and then will keep rising till he is rated 1000
after his 40th. His rating from this point on suggest that he has been
a perfect batsman *over his career* as each rating at each point is
affected by every inning played till that point. Any average rating
calculation should ideally give us a figure of 1000 for Mr. SRT. Now,
if you just take the average of his Pwc ratings it will come out to
something like 900, or maybe even less than that, over his 80
innings.... this seems injustice to a perfect batsman...
Nope, that is not how it works, or at least I do not believe so. The
discounting stops as soon as he reaches his 40th innings. And the
discounting is continually reduced on the way to 40 as well.
that is exactly what I said. The discounting does stop after 40th
innings but that means all his ratings after that time will will be
based on all his innings till that point without any dampening. That
is why Mr SRT in my example had a rating of 1000 from 40th innings
onwards. However, my point is that his earlier ratings are still based
on the dampening factor and hence any calculation which uses these
ratings gives a much lower weight to his earlier innings than later
innings. Even though dampening makes sense for calculating a "current
rating", one consequence of this is that different innings of the
batsman's career will carry very different weights in any calculation
of average rating over a career.
Post by David North
So it is more like (example, not exact formula)
After 1 innings, only 10% of his innings thus far count (90% discounted)
After 2 innings, only 11% of his innings thus far count
...
After 38 innings, only 85% of his innings thus far count (15% discounted)
After 39 innings, only 95% of his innings thus far count
After 40 innings, all 100% of his innings thus far count
I dont understand what you mean by "10% of his innings" after 1
innings. did you mean 10% of his rating? because that is really how it
works. a player's rating is calculated by the normal formula
(including exponential weight decay of previous innings) and then a
damping factor is multiplied to it depending on the total number of
innings that he has played.

another example: lets assume the qualifying mark is 3 innings. with
the damping factor being 0.6 (after 1 inning), 0.85 (2), 1.0 (3 or
greater than 3)

now Mr SRT - our perfect batsman - will have following ratings
assuming he plays 6 innings in his life (proportional to 80 career
innings with qualifying mark set at 40).

1 - 600
2 - 850
3 - 1000
4 - 1000
5 - 1000
6 - 1000

if one takes an average of this, it will be 908.33
why do you think his rating decreased? because we took average of
ratings that themselves were discounted.
Post by David North
So unless someone plays far less than 40 innings, early performances count
just as much, but only after that 40 figure is reached.
well, not just as much. think about his first inning. It didnt carry a
weight of 1.0 when it was played because of the dampening factor
(which is 0.43 really). After that it could never have carried a
weight of 1.0, even after the 40th innings, because of exponential
decay thing. when you calculate a career-wide average rating, you end
up giving too low weightage to this inning.
Post by David North
a. Tendulkar scores 6000 runs in 75 tests at 50 per and retires.
b. Tendulkar scores 10000 runs in 125 tests at 50 per and retires
Is b. as good a career as a. or is it better?
I am saying b. without a doubt.
(Note, I am not saying 10000 at 45 vs. 6000 at 50)
that is what I meant by saying that it can be used as a tie-breaker.
If you have to choose between average 50 over 75 tests and average 50
aver 125 tests, you'll choose the latter, no doubt. But, just as you
yourself said, this doesnt mean that 50 over 75 is not as good as 45
over 125. So, how do we reconcile this thing? I think - just as I
suggested in the previous post - that multiplying with innings raised
to a low exponent is a good idea.

example: lets take the exponent to be 0.2, then

a. gives 50 * 75^ 0.2 = 118.6
b. gives 50 * 125 ^ 0.2 = 131.32

so a longer career at same average is rated higher.

however, for 45 over 125 vs 50 over 75, we have

a. 50 * 75 ^ 0.2 = 118.6
b. 45 * 125 ^ 0.2 = 118.2

so, depending on the exponent, we get a compensation for longer
innings.

Rahul
Shripathi Kamath
2003-10-05 17:13:56 UTC
Permalink
Post by Rahul Tyagi
Post by David North
Post by Rahul Tyagi
Assume there exists a "perfect" batsman - Mr. SRT - who is a 1000
rating batsman. This means that he actually is a perfect batsman from
his first innings to his last, say his 80th innings. So, his
non-discounted rating remains 1000 throughout his career. But before
his 40th innings he will be rated lower. Will probably start with 200
after his first inning and then will keep rising till he is rated 1000
after his 40th. His rating from this point on suggest that he has been
a perfect batsman *over his career* as each rating at each point is
affected by every inning played till that point. Any average rating
calculation should ideally give us a figure of 1000 for Mr. SRT. Now,
if you just take the average of his Pwc ratings it will come out to
something like 900, or maybe even less than that, over his 80
innings.... this seems injustice to a perfect batsman...
Nope, that is not how it works, or at least I do not believe so. The
discounting stops as soon as he reaches his 40th innings. And the
discounting is continually reduced on the way to 40 as well.
that is exactly what I said. The discounting does stop after 40th
innings but that means all his ratings after that time will will be
based on all his innings till that point without any dampening. That
is why Mr SRT in my example had a rating of 1000 from 40th innings
onwards. However, my point is that his earlier ratings are still based
on the dampening factor and hence any calculation which uses these
ratings gives a much lower weight to his earlier innings than later
innings. Even though dampening makes sense for calculating a "current
rating", one consequence of this is that different innings of the
batsman's career will carry very different weights in any calculation
of average rating over a career.
Post by David North
So it is more like (example, not exact formula)
After 1 innings, only 10% of his innings thus far count (90% discounted)
After 2 innings, only 11% of his innings thus far count
...
After 38 innings, only 85% of his innings thus far count (15% discounted)
After 39 innings, only 95% of his innings thus far count
After 40 innings, all 100% of his innings thus far count
I dont understand what you mean by "10% of his innings" after 1
innings. did you mean 10% of his rating?
Yes. In other words, calculate his rating according to the normal PwC
exponential decay calculations.

Then dampen it by considering only 10% of it for his first, 11 for his
second and so on.
Post by Rahul Tyagi
because that is really how it
works. a player's rating is calculated by the normal formula
(including exponential weight decay of previous innings) and then a
damping factor is multiplied to it depending on the total number of
innings that he has played.
another example: lets assume the qualifying mark is 3 innings. with
the damping factor being 0.6 (after 1 inning), 0.85 (2), 1.0 (3 or
greater than 3)
now Mr SRT - our perfect batsman - will have following ratings
assuming he plays 6 innings in his life (proportional to 80 career
innings with qualifying mark set at 40).
1 - 600
2 - 850
3 - 1000
4 - 1000
5 - 1000
6 - 1000
if one takes an average of this, it will be 908.33
why do you think his rating decreased? because we took average of
ratings that themselves were discounted.
Er, 1000 is not a prize winning number. All that matters is that the 908.33
is more than any player who has performed less.

The number decreases for *everybody*, not just SRT. So in that sense it is
fair.
Post by Rahul Tyagi
Post by David North
So unless someone plays far less than 40 innings, early performances count
just as much, but only after that 40 figure is reached.
well, not just as much. think about his first inning. It didnt carry a
weight of 1.0 when it was played because of the dampening factor
(which is 0.43 really). After that it could never have carried a
weight of 1.0, even after the 40th innings, because of exponential
decay thing. when you calculate a career-wide average rating, you end
up giving too low weightage to this inning.
No because, a player doing poorly will have a harder time raising his
rating, SRT will be there after his 40th innings anyway.

Think of it in relative terms to other players, not absolute.
Post by Rahul Tyagi
Post by David North
a. Tendulkar scores 6000 runs in 75 tests at 50 per and retires.
b. Tendulkar scores 10000 runs in 125 tests at 50 per and retires
Is b. as good a career as a. or is it better?
I am saying b. without a doubt.
(Note, I am not saying 10000 at 45 vs. 6000 at 50)
that is what I meant by saying that it can be used as a tie-breaker.
If you have to choose between average 50 over 75 tests and average 50
aver 125 tests, you'll choose the latter, no doubt. But, just as you
yourself said, this doesnt mean that 50 over 75 is not as good as 45
over 125. So, how do we reconcile this thing? I think - just as I
suggested in the previous post - that multiplying with innings raised
to a low exponent is a good idea.
example: lets take the exponent to be 0.2, then
a. gives 50 * 75^ 0.2 = 118.6
b. gives 50 * 125 ^ 0.2 = 131.32
so a longer career at same average is rated higher.
however, for 45 over 125 vs 50 over 75, we have
a. 50 * 75 ^ 0.2 = 118.6
b. 45 * 125 ^ 0.2 = 118.2
so, depending on the exponent, we get a compensation for longer
innings.
Yes, I think you should submit that to the PwC. And watch out for the DN
dude :-)
--
Shripathi Kamath
David North
2003-10-05 17:52:28 UTC
Permalink
Post by Shripathi Kamath
Post by Rahul Tyagi
another example: lets assume the qualifying mark is 3 innings. with
the damping factor being 0.6 (after 1 inning), 0.85 (2), 1.0 (3 or
greater than 3)
now Mr SRT - our perfect batsman - will have following ratings
assuming he plays 6 innings in his life (proportional to 80 career
innings with qualifying mark set at 40).
1 - 600
2 - 850
3 - 1000
4 - 1000
5 - 1000
6 - 1000
if one takes an average of this, it will be 908.33
why do you think his rating decreased? because we took average of
ratings that themselves were discounted.
Er, 1000 is not a prize winning number. All that matters is that the 908.33
is more than any player who has performed less.
The number decreases for *everybody*, not just SRT. So in that sense it is
fair.
It affects everybody, but not equally. It has a greater effect on players with fewer Tests, and also
on those who did better at the start of their careers. In the above scenario, Mr 97% will have a
higher average than Mr SRT if he plays 9 innings (equivalent to 120).

--snip--
Post by Shripathi Kamath
And watch out for the DN
dude :-)
Don't call me dude - I might mistake you for Larry. ;)
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-05 07:14:12 UTC
Permalink
--snip--
Post by David North
Post by David North
Post by David North
The reason for the adjustment is that PwC deliberately give players
only a
Post by David North
percentage of their
Post by David North
performance points until they have fully qualified, which is after 40
innings for batsmen and 100
Post by David North
wickets for bowlers. Therefore, players will generally have lower
ratings
Post by David North
up to this point.
Which is exactly what it is meant to do, dampen early ratings so that
flash
Post by David North
in the pan careers are accounted for appropriately. A Bob Massie or a
Jimmy
Post by David North
Adams will otherwise have an unbelievable impact than, say, McGrath, or
a
Post by David North
Waugh.
But it also means that Waugh's (or anyone's) first Test is deemed to be
less important than his
last. Why it that valid when rating his career as a whole?
The way I understand it is that, initially, a player's initial performances
do not count fully , but they do count fully after n innings.
So the "damping" is removed after the qualifying period - that's fine, but if you start using the
earlier damped ratings in producing an overall career rating, you are giving a lower weight to
earlier performances, which is not correct.
Post by David North
So Waugh's
initial failures do come into picture. So the answer to your question is
no, Waugh's latter performances are more important, but his initial ones do
have an impact.
The latter performances are more important in a current rating, but not in an overall career rating.
Post by David North
Post by David North
Post by David North
This
will have a greater effect on players who played fewer Tests. If two
players perform consistently at
Post by David North
the same level, but player A plays 25 Tests and player B plays 50
Tests,
Post by David North
player A will have a lower
Post by David North
average rating than player B, even though he is just as good.
Actually, it is harder for a player to maintain such form. A player who
averages, say, 50 runs per innings over 50 tests, should be better
regarded
Post by David North
than one who averages the same over only 25 though. (Other things like
opposition etc., being in the same vicinity, of course.)
Let's see. Steve Waugh first achieved his current average at the end of
the 1998/9 series in the
West Indies. If the plane has crashed on the way back to Australia, should
he then have been rated
lower than he actually is?
Not lower than he actually is, but I'd rate him higher today than if he had
stopped playing at the same average back then.
OK, but the difference would have to be slight. As the number of Tests increases, the effects of
bursts of form (good or bad) on a player's averages decrease exponentially.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Shripathi Kamath
2003-10-05 07:48:02 UTC
Permalink
wrote in message news:LLJfb.5559$***@fed1read04...
<snip>
Post by David North
Post by Shripathi Kamath
The way I understand it is that, initially, a player's initial performances
do not count fully , but they do count fully after n innings.
So the "damping" is removed after the qualifying period - that's fine, but
if you start using the
Post by David North
earlier damped ratings in producing an overall career rating, you are
giving a lower weight to
Post by David North
earlier performances, which is not correct.
No, the way I believe it works is that earlier ratings gain their full
effect after n (20 or 40) innings. Only for the first n innings do they
remain damped. (please see my response to Tyagi for a crude example)

To be clear, even if a player has all hundreds till that figure of n
innings, they do not contribute a 100% to his ratings, but do fully as soon
as the n figure is crossed. With exponential decay and all.

I believe that this is done so that no Graeme Smith or Kambli meanders into
the du jour top ten very easily early on.

I do believe though that if a player plays less than n innings for his
career, he is out of luck in that his performances do not get their full
weight.

Again, I think this is fair enough, but I can see if you and others don't.
--
Shripathi Kamath
John Hall
2003-10-05 09:19:51 UTC
Permalink
Post by David North
Post by David North
The reason for the adjustment is that PwC deliberately give players only a
percentage of their
Post by David North
performance points until they have fully qualified, which is after 40
innings for batsmen and 100
Post by David North
wickets for bowlers. Therefore, players will generally have lower ratings
up to this point.
Which is exactly what it is meant to do, dampen early ratings so that flash
in the pan careers are accounted for appropriately. A Bob Massie or a Jimmy
Adams will otherwise have an unbelievable impact than, say, McGrath, or a
Waugh.
The problem with that, though, is that far fewer Tests were played a
hundred years ago, so if you don't give full weighting for early
performances then players from that era will be under a handicap when
their ratings are compared with more recent players.

Having said that, though, for whatever reason the "raw" PwC rankings do
seem to more closely correspond to what I would have expected - though
naturally there are some exceptions.
--
John Hall
Johnson: "Well, we had a good talk."
Boswell: "Yes, Sir, you tossed and gored several persons."
Dr Samuel Johnson (1709-84); James Boswell (1740-95)
Shripathi Kamath
2003-10-05 17:08:01 UTC
Permalink
<snip>
Post by John Hall
The problem with that, though, is that far fewer Tests were played a
hundred years ago, so if you don't give full weighting for early
performances then players from that era will be under a handicap when
their ratings are compared with more recent players.
If they played less than the carefully selected 40 innings mark, yes. But
they do that evenly across the generations. I do see your point that there
are players who might have played only 15 tests, and get short-changed,
however, I don't think that they have played enough that their 50 run
average is comparable to that of one who has played 78 tests. DN clearly
disagrees.


<snip>
Also, I am guessing that most people would expect the players who they are
more familiar with to be
ranked near the top. This will tend to be those who played more Tests, and
it is those who are
favoured more by the raw ranking method. Most people are not very familiar
with the likes of John
Lyons and Frank Iredale, who played only 14 Tests each, so are surprised
if they are rated highly,
despite the fact that their performances do stand out from those of their
team-mates.
I believe that this is a cogent point when it comes to 'rating' greatness by
numbers. Either we accept what a set of numbers dole out, or we simply use
them as a crutch to support a main point, presumably a better crutch than
just plain averages.

(The usual way seems to be only use it as a crutch when they conveniently
support my POV, and as a stick to beat people with when they disagree :-))

It is not, however, limited to the Lyons and the Iredales. Even for
well-known players who have played a sufficient number of tests, some
comparisons do not exactly reveal what one expects to see.

The PwC ratings are just another set of numbers, they do ostensibly account
for things like strength of opposition, consistency, etc., beyond a simple
career average or a recent slice of the career average. For the very best,
they seem to work out fine, for the others, I have found more than a few
that run contrary to expectations.

They promise to publish their formulas someday, maybe that will illuminate
something more.
--
Shripathi Kamath
John Hall
2003-10-05 17:57:50 UTC
Permalink
Post by Shripathi Kamath
<snip>
Post by John Hall
The problem with that, though, is that far fewer Tests were played a
hundred years ago, so if you don't give full weighting for early
performances then players from that era will be under a handicap when
their ratings are compared with more recent players.
If they played less than the carefully selected 40 innings mark, yes. But
they do that evenly across the generations.
Not evenly. 40 innings represents something like 25 Tests. Pre WW1, the
frequency of Tests was low enough that a player might take something
like a decade to play that many Tests. Nowadays an ever-present might do
it in under 3 years.
Post by Shripathi Kamath
I do see your point that there
are players who might have played only 15 tests, and get short-changed,
however, I don't think that they have played enough that their 50 run
average is comparable to that of one who has played 78 tests.
I would agree with you here, though. Frankly I'm not sure that there is
a wholly satisfactory solution.
--
John Hall

"Whenever people agree with me I always feel I must be wrong."
Oscar Wilde
David North
2003-10-05 19:08:27 UTC
Permalink
Post by John Hall
Post by Shripathi Kamath
<snip>
Post by John Hall
The problem with that, though, is that far fewer Tests were played a
hundred years ago, so if you don't give full weighting for early
performances then players from that era will be under a handicap when
their ratings are compared with more recent players.
If they played less than the carefully selected 40 innings mark, yes. But
they do that evenly across the generations.
Not evenly. 40 innings represents something like 25 Tests. Pre WW1, the
frequency of Tests was low enough that a player might take something
like a decade to play that many Tests. Nowadays an ever-present might do
it in under 3 years.
Less than 2 in some cases. By the end of the tour of West Indies, England will have played 16 Tests
in a year.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-05 19:05:43 UTC
Permalink
Post by Shripathi Kamath
<snip>
Post by John Hall
The problem with that, though, is that far fewer Tests were played a
hundred years ago, so if you don't give full weighting for early
performances then players from that era will be under a handicap when
their ratings are compared with more recent players.
If they played less than the carefully selected 40 innings mark, yes. But
they do that evenly across the generations. I do see your point that there
are players who might have played only 15 tests, and get short-changed,
however, I don't think that they have played enough that their 50 run
average is comparable to that of one who has played 78 tests. DN clearly
disagrees.
I don't really disagree. I was playing devil's advocate earlier. However, if you want to add on a
bonus for sustained performance, it should be done as an adjustment at the end of the calculation,
not by devaluing everyone's early performances arbitrarily.
Post by Shripathi Kamath
<snip>
Also, I am guessing that most people would expect the players who they are
more familiar with to be
ranked near the top. This will tend to be those who played more Tests, and
it is those who are
favoured more by the raw ranking method. Most people are not very familiar
with the likes of John
Lyons and Frank Iredale, who played only 14 Tests each, so are surprised
if they are rated highly,
despite the fact that their performances do stand out from those of their
team-mates.
I believe that this is a cogent point when it comes to 'rating' greatness by
numbers. Either we accept what a set of numbers dole out, or we simply use
them as a crutch to support a main point, presumably a better crutch than
just plain averages.
(The usual way seems to be only use it as a crutch when they conveniently
support my POV, and as a stick to beat people with when they disagree :-))
It is a test of one's objectivity to go into any exercise of statistical research with an open mind
and publish the results whether you like them or not. I'm not very good at it either. :-)
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
David North
2003-10-09 23:46:10 UTC
Permalink
I have now amended the rating system by removing (approximately) the scaling-down that PwC apply to
their ratings for batsmen who have played fewer than 40 innings and bowlers who have taken fewer
than 100 wickets. This means that the adjustments I made under the previous system are no longer
required, and the ratings are a straight average of the players' PwC ratings throughout their
career, after the scaling-down has been removed. Therefore, the ratings will not now change as
further data is added.

I have assumed that the scale that PwC use for bowlers is equivalent to that for batsmen, but I have
submitted a question on their Web site asking for the bowling scale, and if it is forthcoming, I
will post a further amendment to the bowling and all-round lists if necessary.

In order to avoid have to enter the number of innings played and wickets taken by each player in
each Test (in order to remove the scaling-down), I have also assumed a constant rate for innings
played over each player's career, and for wickets taken up to the match in which the 100th wicket
was taken, by dividing the total number of innings and wickets by the number of matches played.
Obviously this assumption is much more likely to deviate from reality for bowlers, and bowlers whose
wickets-per-match rate slowed down during their career (Wilfred Rhodes is the obvious example) will
have somewhat higher ratings because of this assumption.

Here are the amended top 20s:

Batting
1 856.5 Jack Hobbs (Eng)
2 833.7 Ranji Ranjitsinhji (Eng)
3 816.8 Frank Iredale (Aus)
4 789 John Lyons (Aus)
5 744.7 Walter Read (Eng)
6 719.6 Andrew Stoddart (Eng)
7 717.5 Stanley Jackson (Eng)
8 707.7 WG Grace (Eng)
9 706 Herbert Taylor (SA)
10 702.9 Allan Steel (Eng)
11 701 Clem Hill (Aus)
12 698.1 Robert Abel (Eng)
13 696.5 Reggie Spooner (Eng)
14 689.7 Reggie Duff (Aus)
15 679 Warren Bardsley (Aus)
16 672.1 Gordon White (SA)
17 662.3 Arthur Shrewsbury (Eng)
18 650.7 Vernon Ransford (Aus)
19 650.3 Johnny Tyldesley (Eng)
20 629.7 Joe Darling (Aus)

Bowling
1 914.3 Frank Foster (Eng)
2 905.8 Charlie Turner (Aus)
3 829.4 Tom Richardson (Eng)
4 829.3 Bobby Peel (Eng)
5 823.7 Bill Whitty (Aus)
6 812.7 Sydney Barnes (Eng)
7 790.8 Frederick Spofforth (Aus)
8 774.7 Joey Palmer (Aus)
9 768.7 Billy Bates (Eng)
10 757.9 George Lohmann (Eng)
11 736 Monty Noble (Aus)
12 707.3 Jack Saunders (Aus)
13 704.4 Jimmy Blanckenberg (SA)
14 695.5 Tibby Cotter (Aus)
15 666.2 Jack T Hearne (Eng)
16 647.7 Johnny Briggs (Eng)
17 646.2 Wilfred Rhodes (Eng)
18 643.2 Charlie Blythe (Eng)
19 639.8 Reggie Schwarz (SA)
20 628.3 Aubrey Faulkner (SA)

All-Round
1 716.4 Frank Foster (Eng)
2 634.2 Allan Steel (Eng)
3 612.5 Billy Bates (Eng)
4 608.9 Aubrey Faulkner (SA)
5 604.6 George Ulyett (Eng)
6 600.7 Monty Noble (Aus)
7 586.2 Charlie Turner (Aus)
8 535 Charles Macartney (Aus)
9 533.5 WG Grace (Eng)
10 530.4 Jimmy Sinclair (SA)
11 526.7 Johnny Douglas (Eng)
12 509.1 Frank Woolley (Eng)
13 506.3 Billy Barnes (Eng)
14 505.8 Wilfred Rhodes (Eng)
15 496.5 George Lohmann (Eng)
16 491.8 Tip Snooke (SA)
17 478.7 Len Braund (Eng)
18 478.3 Johnny Briggs (Eng)
19 474.7 Claude Carter (SA)
20 473.8 Charles Kelleway (Aus)

Sorry Trumper fans - he's still 21st. One notable change is that WG Grace is now in the top 10 in
both the batting and all-round lists.
--
David North
Email to this address will be deleted as spam
Use usenetATlaneHYPHENfarm.fsnet.co.uk
Continue reading on narkive:
Loading...