THE CORDON HOME

BLOGS ARCHIVES
SELECT BLOG
July 28, 2009

Tests - bowling

Comparing Test bowlers to their peers

Anantha Narayanan

I have done a lot of cricket analysis work over the past 20+ years. I love doing all this work. However once a while a new idea comes across which I consider as a watershed moment in my analytic efforts. The idea of comparing a player with peer players (the base idea of which was provided by Abdulla) is one such spark. I am very excited about this since it is one of the truest measures of a players' capabilities. I am posting this as an interim piece since I intend using some of the findings herein in the "Test Bowlers: follow-up" article.

The idea is to compare a player's performances with his peers. The comparisons with his own team is one limited step and is quite useful. However the real comparison is with all the peer players since it takes perfect care of the vexed question of a player playing in a very strong team. I had done this in a limited way for ODI strike rates. Now I have extended this to Test players in a much more extended manner as explained below.

My initial idea was to come out with the batting tables also in this article. However I have decided to that in a later article so that the analysis currently on hand, on Test bowlers, gets its due attention and does not get side-tracked.

1. For each player, I created a match subset of their career limits, in other words from their first to last Tests. For Muralitharan it is 1195(1992) to 1912 (2009), 717 Tests. For Tendulkar it is 1127(1989) to 1918(2009), a subset of 791 Tests, the longest span for any player.

2. For Bowling, sum the three main data elements, Balls Bowled, Runs Conceded, and Wickets Captured for all the players for these matches. These are quite high numbers.

3. For Batting, sum the three main data elements, Innings, Not Outs, Balls Faced (if available) and Runs Scored for all the players for these matches. This will be covered in depth in a later article.

4. Subtract the player's own career figures from the total for the match subset and post these figures as a database segment. Even though the players' own numbers are quite low compared to the match subsets (Muralitharan 770 out of 21281 wkts and Tendulkar 12773 out of 749558 runs) and the impact of this subtraction is minimal, it is done to get an exact peer segment.

I have not done a separation by bowler type nor by period. This is a pure peer comparison, cutting across all divisions. I wanted to see the place of a great spinner like Muralitharan across all bowlers, to understand his true value.

First let us look at the Bowler tables. There are three tables in all, one which compares the Bowling Average, the second, the Bowling Strike rate and the third, compares the RpO.

1. Bowler Peer comparisons - Bowling Average

SNo.Bowler            Cty    Own  <--Peer Bowlers-->
Avge    Runs  Wkts Avge Ratio

0.Lohmann G.A Eng 0022-0050( 29) 10.76 17664 847 20.85 1.94 0.Barnes S.F Eng 0065-0133( 69) 16.43 53823 2029 26.53 1.61 ... 1.Marshall M.D Win 0837-1175(339) 20.95 299245 9217 32.47 1.55 2.McGrath G.D Aus 1235-1826(592) 21.64 562481 17029 33.03 1.53 3.Muralitharan M Slk 1195-1912(718) 22.18 683748 20511 33.34 1.50 4.Garner J Win 0797-1072(276) 20.98 241822 7644 31.64 1.51 5.Ambrose C.E.L Win 1095-1509(415) 20.99 374642 11797 31.76 1.51 6.Wardle J.H Eng 0296-0440(145) 20.39 125187 4152 30.15 1.48 7.Hadlee R.J Nzl 0710-1147(438) 22.30 391665 12140 32.26 1.45 8.Steyn D.W Saf 1728-1916(189) 23.70 193060 5530 34.91 1.47 9.Pollock S.M Saf 1312-1860(549) 23.12 529531 15921 33.26 1.44 10.O'Reilly W.J Aus 0215-0275( 61) 22.60 52334 1617 32.36 1.43 ... 145.Boje N Saf 1484-1812(329) 42.65 325844 9701 33.59 0.79 146.Giffen G Aus 0005-0052( 48) 27.10 29298 1449 20.22 0.75 147.Hooper C.L Win 1085-1622(538) 49.43 496933 15592 31.87 0.64

The top two bowlers are from the "Wild west era" as Jeff calls it. A bowling average exceeding 20 was a poor one and this is borne out by the numbers of these two great bowlers, Lohmann and Barnes. Let us respect them and give them their top places and move on. I have also assigned them serial numbers of 0.

A number of readers are bound to be quite happy at seeing Marshall at the top. He was 55% ahead of his peers, including his illustrious team-mates. Probably this was the X-factor which many readers found in Marshall. Next is the incomparable McGrath who was 53% ahead of his peers. No surprise there. However there is a big surprise at the next placed bowler, Muralitharan. His figure of 50% over his peers should, once and for all, put to rest any doubts about his greatness. Those who say that he has succeeded only because he was in a weak team should stop and look at this figure. His figure of 50% is on all types of bowlers, pace included.

The two great West Indian fast bowlers, Garner and Ambrose come in next, again a vindication of their position among their contemporaries. Wardle (a surprise), Hadlee, Steyn, Shaun Pollock (a recognition of this modern great) and O'Reilly complete the top-10. Maybe that is why O'Reilly was chosen ahead of Grimmett in the Cricinfo all-time Australian XI.

The top-10 consists of 7 fast bowlers and 3 spinners, one from each era. There are three great West Indian fast bowlers, 2 South African speedsters and two Australian bowlers in this group.

The table is propped up by two average modern spinners and Giffen from the pre-WW1 era.

To view the complete list, please click here.

2. Bowler Peer comparisons - Bowling Strike rate

SNo.Bowler            Cty   Own  <-Peer Bowlers-->
S/R   Overs  Wkts S/R Ratio

1.Steyn D.W Saf 1728-1916(189) 39.3 60370 5530 65.5 1.67 2.Trueman F.S Eng 0351-0592(242) 49.4 92110 6759 81.8 1.65 3.Waqar Younis Pak 1127-1637(511) 43.5 167408 14587 68.9 1.58 4.Lohmann G.A Eng 0022-0050( 29) 34.1 7478 847 53.0 1.55 5.Marshall M.D Win 0837-1175(339) 46.8 110126 9217 71.7 1.53 6.Hall W.W Win 0459-0648(190) 54.3 73998 5449 81.5 1.50 7.Donald A.A Saf 1188-1590(403) 47.0 132130 11470 69.1 1.47 8.Shoaib Akhtar Pak 1389-1852(464) 45.7 151393 13672 66.4 1.45 9.Hadlee R.J Nzl 0710-1147(438) 50.9 146757 12140 72.5 1.43 10.Pollock P.M Saf 0515-0673(159) 56.2 62434 4672 80.2 1.43 ... 147.Shastri R.J Ind 0897-1206(310) 104 101002 8600 70.5 0.68 148.Emburey J.E Eng 0830-1301(472) 104 156168 13341 70.2 0.67 149.Hooper C.L Win 1085-1622(538) 121 178031 15592 68.5 0.57

The Strike Rate is dominated by fast bowlers who occupy all 10 places. Steyn's attacking skills are evidenced by his top position. He is followed by Trueman and the Pakistani giant, Waqar Younis, the WW1 great Lohmann and the top West Indian bowler of all time, Marshall. Five other great fast bowlers complete the top-10 table. The highest placed spinner is Laker, who is in 26th place.

The table is propped by three very average modern spinners.

To view the complete list, please click here.

3. Bowler Peer comparisons - Bowling RpO

SNo.Bowler            Cty   Own  <--Peer Bowlers-->
RpO   Overs  Runs  RpO Ratio

1.Goddard T.L Saf 0407-0672(266) 1.65 102848 240647 2.34 1.42 2.Verity H Eng 0210-0272( 63) 1.88 20504 53897 2.63 1.39 3.O'Reilly W.J Aus 0215-0275( 61) 1.95 19804 52334 2.64 1.36 4.Tate M.W Eng 0153-0245( 93) 1.94 31583 80403 2.55 1.31 5.Edmonds P.H Eng 0762-1079(318) 2.13 105373 282754 2.68 1.26 6.Pollock S.M Saf 1312-1860(549) 2.40 176869 529531 2.99 1.25 7.Illingworth R Eng 0457-0727(271) 1.91 105842 253356 2.39 1.25 8.Lohmann G.A Eng 0022-0050( 29) 1.89 7478 17664 2.36 1.25 9.Emburey J.E Eng 0830-1301(472) 2.20 156168 425350 2.72 1.24 10.Gibbs L.R Win 0448-0770(323) 1.99 122295 297389 2.43 1.22 ... 147.Hall W.W Win 0459-0648(190) 2.92 73998 176672 2.39 0.82 148.Edwards F.H Win 1649-1920(272) 3.98 88839 281972 3.17 0.80 149.Wright D.V.P Eng 0263-0333( 71) 3.12 26891 65859 2.45 0.79

Trevor Goddard, the most accurate bowler of all time, is on top. As expected, the RpO table is dominated by spinners, headed by Verity and O'Reilly. Then comes the doyen of fast-medium bowlers, Tate. Edmonds, average otherwise, follows next. The real surprise is the placement of Shaun Pollock in the 5th position indicating how accurately he has bowled during these batsmen-dominated period. The other surprise is Emburey who occupies a top-10 placement here even though he is in the last 3 in the Strike Rate list indicating that he was of great value to the English team. Nadkarni who would have been right at the top does not qualify. Steyn and Lee, incidentally, are as low as 135th and 136th respectively indicating that they have been very expensive.

The last three is a motley collection of a West Indian great, West Indian journeyman and an outstanding but extravagant leg spinner.

To view the complete list, please click here.

Test Bowlers Analysis: Follow-up

Based on the comments received, both in public and personal mails, I have decided to make the following tweaks to the Test bowlers analysis. Interested readers may send in their comments at the earliest.

1. Have a cut-off of 200 wickets for the current era, reducing the number from 89 to 44. We will lose Shoaib Akhtar, Steyn, Alderman, Bishop et al. But it cannot be helped.
2. Increase the Wickets weight from 5 points to 7.5 points. Within this, do a 5% on either side (105% & 95%) valuation for Away and Home wickets.
3. Correspondingly reduce the Wickets per Innspell weight from 5 points to 2.5 points.
4. Remove the Performance Ratio measure, the last column in the table.
5. Instead introduce the Peer Comparison ratios. This time I have allotted an equal weight for Strike Rate and Accuracy (Yash will be happy to note).
6. Introduce a simple 5-Test slice based Consistency index using wickets captured as the indicator.
7. In the Match performance Ratings, halve the balls bowled base points (a wicket equivalent for about 45 overs).
8. In the Match performance Ratings, introduce the bowler strike rate, in relation to Team strike rate as a new base measure, at a relatively lower weight.
9. In the Match performance Ratings, minor changes to the batsman dismissed base point calculation, to be based on recent form. This will lower the value of wickets of top batsmen while going through a poor patch and increase the weight of capturing in-form batsman.

The revised allocations of the Career points are given below. The points have gone up to 45 and there is a slight increase in the Match performance points because of changes in Base points calculation.

- Career wickets captured (7.5 points)
- Career wickets per innspell (2.5 points)
- Bowling Strike rate-BpW (9 points)
- Bowling accuracy-RpO (6 points)
- Consistency (4) points
- Average Quality of batsmen dismissed - based on CtD bat avge (4 points)
- Type of wickets captured - Top/Middle order/Late order (4 points)
- Peer ratio: Strike rate (4 points)
- Peer ratio: Accuracy (RpO) (4 points).

My thanks to Arjun Hemnany, Shankar Krishnan, Kartik, Alex, Ed, Yash Rungta et al.

The Batting Peer tables will follow the Test Bowlers follow-up article.

Anantha Narayanan has written for ESPNcricinfo and CastrolCricket and worked with a number of companies on their cricket performance ratings-related systems

RSS Feeds: Anantha Narayanan

© ESPN Sports Media Ltd.

Posted by Ravi on (August 11, 2009, 13:50 GMT)

A little off the point. But on the Federer/cricket angle. The thing (problem?) with Tennis is that it is a “winner takes all” sport. i.e. if Federer is say a ½ % better than his rivals he will win and so gets the Titles. It is entirely different in team sports where the impact of the other 21 players has a major impact on proceedings. So, we have the gold standard of “batting average” in cricket. We can relatively safely assume that a batsman who averages 60 over a long stretch of matches is better than a batsman who averages 50 over a similarly long stretch…(all things being equal i.e. varied opponents ,roughly similar number of ODIs and tests played, injuries etc)… But in Tennis, you either win or lose. Which is why the sheer number of titles/slams won by Federer does not automatically make him the “best ever”(perhaps the most successful ever)…because the exact same skills may or may not have delivered results in another era with other players. Because the same ½ % by which he is better than others now may not hold good.

Posted by Abhi on (August 4, 2009, 13:18 GMT)

Ananth, This belatedly struck me:

“If Kumble took 8 for 141, with the 705 behind him, that spell, wonderful it was, will be undervalued. On the other hand the same Kumble's spell of 5 for 90, defending a total of 105 will be increased in value.” I am in two minds whether this should be so…because if the wicket was good, shouldn’t the value increase? I seem to remember on some previous blog you mentioned that in taking into account pitch quality you have taken total runs( both teams)whereas clearly It should be separated…perhaps for bowlers we can correct this inaccuracy? [[ Abhi I think this is a Test Bowler analysis and we should stop the Batsmen related discussions at this point. I also think you have not read the Test Bowler/Batsmen analysis articles carefully. The sum of the top-10 scores in the match determine how much of a batting pitch it is. The said match is one of the best batting pitches ever (1110 runs) and Kumble will get credit for that. Ananth: ]]

Posted by Abhi on (August 4, 2009, 10:45 GMT)

Ok Ananth, too longwinded? Will try to compress. If you take “peer comparison” then Tendulkar comes out ok out of the top batsmen at that time. We can assume that the various batsmen then faced more or less the same bowlers in more or less the same conditions. Again, as we have seen, not exactly but approximately. For eg Azhar avg 23 in 12inns!!(Never mind the rest of the batsmen also on the 91 tour). So, we know how well “X” has done against his peers world wide in that particular time. But then in some other period (perhaps generally much easier for batting) dozens of batsmen avg say 50+, then how can we compare these 2 periods to each other? How correct is it to say that some 100 batsmen have done “better” than Tendulkar against SA in SA. I don’t know if I am quite getting my point through. But this “peer comparison” measure gives us much greater insight as to what REALLY went on in a definite particular period/era, instead of blindly comparing stats across eras with wildly different conditions. And for this reason it should form the BEDROCK of such analyses. [[ After the Test Bowlers follow-up article, I am coming out with the peer analysis of Test Batsmen. The only word I object to is "blindly". There is a lot of insight which goes into the comparisons. If a bowler has bowled at 2.67 in an era where the average is 2.89 and the all-time average is 2.78, his rpo will be adjusted suitably. Similarly if a bowler has a strike rate of 35 in an era of great strike rates (avge 40 against all time 50), his strike rate will be suitably adjusted. If Kumble took 8 for 141, with the 705 behind him, that spell, wonderful it was, will be undervalued. On the other hand the same Kumble's spell of 5 for 90, defending a total of 105 will be increased in value. And so on and so on. Man-years of work has gone into it. I am happy that a new metric, the peer comparisons, great in its intrinsic value has come through. Ananth: ]]

Posted by Abhi on (August 4, 2009, 6:26 GMT)

“Peer comparison” is probably the holy grail of all player analysis. It is the only solid stat (or as close to solid as you can get with stats) which really indicates anything. All else, including how X would have done in Ys era...is pure conjecture. A simple eg would be Murali….a couple of decades back he would almost certainly have had his career terminated since with the “naked eye” his action certainly does look dubious. But the variables are impossible to factor in in different eras…the best that can be said of any player (with any certainty) is how good he was in his own era. Many think that Federer is the greatest ever now. But a simple point pricks this argument- Then how come he has such a lopsided record against Nadal and Murray? If he is the “greatest of all time”…why can’t he beat his own contemporaries in his PRIME. This last part is the KEY…we are not talking about a hobbled, injured, past it Federer. Federer, “in his Prime” loses more to these guys than wins. So, how on earth can we say with any conviction that he would have beaten Laver, Borg, Mac, Sampras…who have you? So, the correct title would be the “Most successful player of his era”- that is all.

Similar, to the title we can confer on any bowler. [[ Abhi Peer comparison is a very strong indicator of a player's status within his playing career. However we cannot forget overall achievements. It would be unfair and churlish to deny someone like Federer his place at the top, after he finishes his career with, say 18 titles, just because he had a net minus record against, say, Nadal. That way, I can guarantee, every Tennis player, barring none, would have had a nemesis player. For that matter all top batsmen, barring (n)one, have had their waterloos. Richards against Pakistan, Tendulkar against South Africa, Lara against India, Ponting against India, one can go on. In no way does this diminish their overall greatness. Ananth: ]]

Posted by Jeff on (August 3, 2009, 11:03 GMT)

@ Engle - Ananth summed it up perfectly.

The two best "independent" variables to use in analysing bowling are SR (wkts/ball) and EconR (runs/ball).

The question is how much weight to give either.

Bowling Average gives a simple 50/50 weighting (ER/SR)

I believe that the weighting depends on a 3rd variable - resources (balls) available.

The lower the balls available, the more valuable Economy Rate becomes. To illustrate, take the extreme example of one of those rain affected 20/20 games that gets reduced to a 5 over slog. Here you have little chance of bowling the oppostion out, so EconR is the most valuable attribute.

At the other extreme are the Timeless Tests of old - here, it doesn't matter how slow the batsmen score, they will win eventually unless you take wickets - therefor SR is most important.

And it's a sliding scale between, from 20/20 through ODIs and Tests. Given the balls available for tests, I think SR is more important,I just don't know by how much [[ Jeff My current weight of 9(+4) & 6(+4) for strike rate and accuracy looks good. Ananth: ]]

Posted by Arjun on (August 3, 2009, 8:44 GMT)

Ananth,

Thanks, I was specially interested in Bond, Clark's numbers since i knew they were right up there. They were averaging 20-22 in the ERA where overall avg. has shoot-up to 35.0 per wkt. why australia are not picking Clark in Ashes is mystery to me. I want to know your views on constant increase in Avg. runs per wicket. will it ever stop or it will cross over 40.0 and keep on increasing?

Posted by Alex on (August 3, 2009, 8:41 GMT)

Ananth - I didn't dispute Kapil's ranking. Just that some people put Kapil down a bit in their comments, and hence I defended him. He was a great player and a simple human being. Sure, there have been better bowlers who are deservedly placed higher in these lists. For example, Marshall and Holding outperformed Kapil in the 1983 series India, which saw Kapil at his near best, Marshall coming into his own, and Holding in his decline. [[ Alex My comments were not addressed you. I just used your comment as a vehicle to pass a message. Another example which comes to my mind is Vettori, by a few miles, the best New Zealand spinner ever. However he is placed quite low, 62nd, in the bowler table. It is a wonderful achievement of Vettori what he has done on the seamer-friendly/recently flat pitches of New Zealand. However the concerned readers are aware that, in the world arena, Vettori is competing with the spin giants and is not going to do very well. Ananth: ]]

Posted by Ananth on (August 3, 2009, 4:00 GMT)

Arjun I have not forgotten your query on Johnson et al. He completed his 100 wkts after this analysis was sent to press. I have since almost concluded the follow-up article and I waited for that, to do a special run since the cut-off is now 200 wkts. I did a special run with 79 wickets as cut-off. The following are the positions. 19.Nzl Bond S.E RFM 47.61 ... 38.Aus Clark S.R RFM 45.16 ... 42.Aus Johnson M.G LFM 44.92 ... Please remember that a lot more bowlers qualify because of this specific lowering of the bar. I hope that Bond is able to play a few more tests for New Zealand since I consider him to be no less than Dale Steyn. Note how highly he is placed despite taking only 79 wickets. Regards Ananth

Posted by Alex on (August 2, 2009, 4:53 GMT)

Great analysis Ananth! The only problem, if any, with the peer comparison is that the playing conditions are not uniformly the same world-wide, and that goes against some players. For example, a Binny or a Kapil would have been more effective as a bowler for England/Australia ... instead, they mostly toiled on adversarial pitches in India. If we view your lists as "ranking of effectiveness as per this measure over their career as it stands?" (and not as "who was the best, 2nd best, etc" ... which is more of an intrinsic quality), even this objection vanishes.

In other words, it is not necessarily true that Kapil or Chandra were that bad as bowlers. Just that if BCCI had insisted on preparing bowler friendly wickets, they might have fared better over their careers. Kapil was a great cricketer, and what he managed as a bowler is amazing ... before him, no fast bowler for India took more than 120 wickets while he ended up with over 430, averaging better than them. [[ Alex All Indian supporters have every right to think that Kapil should be rated no.1 by a comfortable margin in India. However I do not know why people should think that Kapil should be rated high in the all-time list when there have been giants, true giants, playing over the past 40 years. And do not forget that Kapil's wickets in the sub-continent have been given higher weight. Otherwise why should Imran be rated so high. Ananth: ]]

Posted by Ananth on (August 1, 2009, 4:26 GMT)

Arjun The batsman's Ctd average measure you have referred to, "Wicket average" should be seen in conjunction with the next one "Wicket quality" which recognizes capturing top and middle order batsmen. In any test match it is essential to take top order wickets to win the match and this factor is recognized. Currently these two measures have weight of 4 points each. Without disturbing the balance of the analysis, I can change this weight to 5 for "Wicket average" and 3 for "Wicket quality" respectively. Ananth

Comments have now been closed for this article

ABOUT THE AUTHOR

Anantha Narayanan
Anantha spent the first half of his four-decade working career with corporates like IBM, Shaw Wallace, NCR, Sime Darby and the Spinneys group in IT-related positions. In the second half, he has worked on cricket simulation, ratings, data mining, analysis and writing, amongst other things. He was the creator of the Wisden 100 lists, released in 2001. He has written for ESPNcricinfo and CastrolCricket, and worked extensively with Maruti Motors, Idea Cellular and Castrol on their performance ratings-related systems. He is an armchair connoisseur of most sports. His other passion is tennis, and he thinks Roger Federer is the greatest sportsman to have walked on earth.

All articles by this writer