June 28, 2014

Consistency of Test batsmen - Part 2

A new tool to analyse which batsmen have been the most consistent in Test cricket

A cumulative reading of Part 1 and Part 2 of this series could lead to the conclusion that Rohan Kanhai is the most consistent batsman in cricket history © Getty Images

In my previous article I had analysed the consistency of Test batsmen, from the innings point of view. I received a number of good comments and a few excellent ideas were sent by the readers. The one idea which appealed to me most was by Santosh Sequeira who suggested that the Consistency analysis will have much better value if done using a single Test as the basis. Once I got out of my self-created mental block that 100 and 0 in a single Test represented inconsistency, this made a lot of sense.

I could see the following benefits accruing if Test Consistency was measured by Test, and not by innings.

- A Test is the logical unit of delivery for a player since the result is driven by a Test.
- There was a well-justified concern that many batsmen, even the very best, were short-changed in the innings-based analysis. A 100 and 0 represented two innings out of the consistency zone. This will disappear if these two innings have been played within a single Test.
- Top batsmen rarely have double failures in a Test. They make up for failures in one innings with a good innings in the other. The Test-based analysis recognises this characteristic.
- The results bear out this improvement since many top batsmen who were languishing in the lower half of the table of the select group of batsmen, have moved up considerably.
- The upper limit of the consistency zone was fairly low and this meant that many a good innings went out of the consistency zone. This is partly alleviated in the Test-based analysis.
- The impact of not-outs is fairly negligible. Since there are two innings to combine, I could adopt a different approach.

Let me reassure the readers that the following ideas that went into my to-do list are still active candidates for inclusion in future articles.

- Consistency analysis using the Median (Q2) value. This will be an assumption-free analysis.
- Batsman low-score analysis.
- Consistency value at batsman peak.
- Analysis of batsman troughs.

At the end of this article I will check whether there is good correlation between the innings-based analysis and Test-based analysis. If there is good correlation, we could work with either of the methods. If there are many variations, we have to peg our hat on either of the analysis methods. The criteria could be many. Why cross a bridge which is a few kilometres away?

I have devised a simple concept of "Batsman-active Tests". If a batsmen batted in either innings, I consider that Test as an active one for him. Else I do not include it. Don Bradman batted in 50 Tests only, and Sachin Tendulkar in 197 Tests. Just to give a specific example, when South Africa scored 637 for 2 at The Oval and won, AB de Villiers did not bat at all. So this Test is excluded from this analysis. On the other hand Alviro Peterson scored a duck and this is certainly an "active Test" for him. This is eminently fair, simple to understand and easy to work out.

If a batsman is not out at 10 in the only innings he played in a Test, well, these would even out across a career. It is the same policy for all batsmen. I briefly considered, and discarded, the method of taking a fraction of a Test, derived from the scores, in the denominator. Quite confusing, and just not worth it. An analysis of 39.42 Tests? No way.

Test Batsmen Consistency analysis: Top 30 batsmen
No Batsman LHB Ctry Tests Runs RpT Inactive-Tests Active-Tests Real RpT Cons-Zone Range Below CZ Below CZ % Cons-Zone Tests Cons-Index
1Saeed AhmedPak 41 2991 73.0 1 40 74.837.4-112.2 1025.0%2562.5%
2RB KanhaiWin 79 6227 78.8 0 79 78.839.4-118.2 1822.8%4658.2%
3RC FredericksLWin 59 4334 73.5 0 59 73.536.7-110.2 1322.0%3457.6%
4CH LloydLWin110 7515 68.3 1109 68.934.5-103.4 2522.9%6256.9%
5KD WaltersAus 74 5357 72.4 0 74 72.436.2-108.6 1824.3%4256.8%
6AH JonesNzl 39 2922 74.9 0 39 74.937.5-112.4 923.1%2256.4%
7GM TurnerNzl 41 2991 73.0 0 41 73.036.5-109.4 1126.8%2356.1%
8NC O'NeillAus 42 2779 66.2 1 41 67.833.9-101.7 922.0%2356.1%
9SR WatsonAus 52 3408 65.5 0 52 65.532.8- 98.3 1325.0%2955.8%
10Misbah-ul-HaqPak 46 3218 70.0 1 45 71.535.8-107.3 1022.2%2555.6%
11MH RichardsonLNzl 38 2776 73.1 0 38 73.136.5-109.6 923.7%2155.3%
12IJL TrottEng 49 3763 76.8 0 49 76.838.4-115.2 1326.5%2755.1%
13FE WoolleyLEng 64 3283 51.3 2 62 53.026.5- 79.4 1829.0%3454.8%
14A RanatungaLSlk 93 5105 54.9 2 91 56.128.0- 84.1 2527.5%4953.8%
15ND McKenzieSaf 58 3253 56.1 2 56 58.129.0- 87.1 1628.6%3053.6%
16AL HassettAus 43 3073 71.5 0 43 71.535.7-107.2 1023.3%2353.5%
17RB RichardsonWin 86 5949 69.2 0 86 69.234.6-103.8 2124.4%4552.3%
18JH EdrichLEng 77 5138 66.7 2 75 68.534.3-102.8 2026.7%3952.0%
19GR MarshAus 50 2854 57.1 0 50 57.128.5- 85.6 1428.0%2652.0%
20L HuttonEng 79 6971 88.2 0 79 88.244.1-132.4 2329.1%4151.9%
21BJ HaddinAus 57 3033 53.2 1 56 54.227.1- 81.2 1730.4%2951.8%
22DPMD JayawardeneSlk14411392 79.1 1143 79.739.8-119.5 3725.9%7451.7%
23ER DexterEng 62 4502 72.6 0 62 72.636.3-108.9 1524.2%3251.6%
24WJ CronjeSaf 68 3714 54.6 2 66 56.328.1- 84.4 1928.8%3451.5%
25ME TrescothickLEng 76 5820 76.6 0 76 76.638.3-114.9 2330.3%3951.3%
26SP FlemingLNzl111 7172 64.6 3108 66.433.2- 99.6 3431.5%5550.9%
27Asif IqbalPak 58 3575 61.6 1 57 62.731.4- 94.1 1526.3%2950.9%
28A FlowerLZim 63 4794 76.1 0 63 76.138.0-114.1 2031.7%3250.8%
29KJ HughesAus 70 4415 63.1 1 69 64.032.0- 96.0 2029.0%3550.7%
30WR HammondEng 85 7249 85.3 0 85 85.342.6-127.9 2428.2%4350.6%
31AD NourseSaf 34 2960 87.1 0 34 87.143.5-130.6 1029.4%1750.0%

The top position in the Test-based Consistency table is taken by Saeed Ahmed, the attacking Pakistani batsmen whose Test average of 40.41 belied his value to his team. Out of the 40 Tests he batted in, he was in the ConZone an amazing 25 times, leading to an outstanding index value of 62.5%: That is 5 out of 8 Tests. The other telling statistic is the fact that he failed to reach the ConZone in only ten Tests out of these 40. That is a very low failure rate of 25%. In 1965, he made an unforgettable 172, out of a score of 307 for 8, against New Zealand, saving Pakistan from possible defeat. That was his highest Test score.

Then come four more attacking batsmen: Rohan Kanhai, Roy Fredericks, Clive Lloyd and Doug Walters. There are three West Indians and one Australian. This re-emphasises my belief that the attacking batsmen are as likely to be as consistent as the staid batsmen. With their more aggressive attitude, they are more likely to be able to make up for failures in one innings with good showings in the other. All these batsmen have Consistency indices above 56%.

There are a number of lovely batsmen in the top-20. Norm O'Neill is in the top-10. He might not have been the "next Bradman" but was a terrific batsman. The under-rated Misbah-ul-Haq rounds off the top-10. We will look at Jonathan Trott later on. I am very happy to see John Edrich in 18th place. Frank Woolley was a classical left-hander who is deservedly in 13th position. They are both favourites of mine. The 20th ranked batsman is Len Hutton, possibly the best in this lot. He himself clocks in at 51.9%. There are 36 batsmen who have index values of 50% or higher. The last batsman featured is Dudley Nourse who was placed in first position in the other table.

Test Batsmen Consistency analysis: Bottom 10 batsmen
No Batsman LHB Ctry Tests Runs RpT Inactive-Tests Active-Tests Real RpT Cons-Zone Range Below CZ Below CZ % Cons-Zone Tests Cons-Index
191Shoaib MohammadPak 45 2705 60.1 1 44 61.530.7- 92.2 1840.9%1534.1%
192Aamer SohailLPak 47 2823 60.1 0 47 60.130.0- 90.1 1940.4%1634.0%
193VT TrumperAus 48 3163 65.9 1 47 67.333.6-100.9 1940.4%1634.0%
194DL AmissEng 50 3612 72.2 0 50 72.236.1-108.4 2040.0%1734.0%
195HW TaylorSaf 42 2936 69.9 0 42 69.935.0-104.9 1535.7%1433.3%
196GA HickEng 65 3383 52.0 0 65 52.026.0- 78.1 2640.0%2132.3%
197SK WarneAus145 3154 21.8 8137 23.011.5- 34.5 5943.1%4331.4%
198Ijaz AhmedPak 60 3315 55.2 2 58 57.228.6- 85.7 2441.4%1831.0%
199AC ParoreNzl 78 2865 36.7 3 75 38.219.1- 57.3 3040.0%2330.7%
200MS AtapattuSlk 90 5502 61.1 2 88 62.531.3- 93.8 4045.5%2629.5%

Let us look at the table proppers. Marvan Atapattu takes possession of the 200th position. What makes Atapattu so inconsistent? He batted in 88 Tests. He reached the ConZone mark of 31-94 only 26 times. That is a meagre 29.5%: not even a third of his Tests. This index is less than a half of Saeed Ahmed, the table topper. And Atapattu's inconsistency is emphasised by the 40 Tests below the ConZone. A look at his series of scores indicates that there were 30 Tests in which he scored 20 runs or less, and 11 Tests in which he scored 150 or higher.

Ijaz Ahmad has been a revelation. He was 200th in the innings-based table and 198th in this one: a very firm indicator that he was the embodiment of inconsistency. His index is a very low 31.0%. The inscrutable Graeme Hick is in the last-5 with a low index value of 32.3%. Dennis Amiss also confirms that his inconsistency moves on from the innings level to Test level, with an index of 34%. He was 193rd in the earlier table and he has moved one place below to 194th in this one. Michael Clarke has moved away from 192nd to 167th, still way down the table, but at least some distance away from the bottom.

The graphs are self-explanatory. The first one plots the top five batsmen and three from the bottom-10. I have used a modified box plot to do this visual depiction. The benefit is that I can easily show ten batsmen in one graph. The values for all batsmen are scaled, with 100 being taken to represent the number of Active Tests. This makes comparisons easier. The wider the rectangle, the higher will be the Consistency Index. The more the rectangle is to the right, the more will be the sub-CZ numbers, indicating a greater number of failures.

In this chart, I have picked ten notable batsmen and plotted their numbers. Nourse has been added to this list since he was No. 1 in the earlier table. It can be seen that most of these batsmen now have Consistency index values between 40 and 50%. In the earlier table, they had values between 30 and 40%. This clearly confirms that the top batsmen tend to make up for a failure in one innings with a good innings in the other. Double failures are that much rarer. The first number is also interesting. This indicates the real failures. Garry Sobers and Ricky Ponting lead in this measure with failures in a third of their Tests. The top-of-the-table values for this measure are around 25% and below.

Let us now study the ranks in the two tables. Not one batsman features in the top-10 positions in the two tables. However there is a very close contender for this honour. Trott is in 12th position in both tables. This could very well qualify him to be a contender for the most consistent batsman ever. Shane Watson is very close to Trott with ranks of 14 and 9 respectively in the two tables. Neil McKenzie is in positions 17 and 16. Richie Richardson is in creditable sub-20 positions of 13 and 18. Kanhai is very well placed at 19 and 2. Fredericks is at 16 and 3. Now we come to Dudley Nourse. He was first in the earlier table but has moved down to 31st here. Saeed Ahmed was 57th in the innings-based table and has moved to top position here.

The three batsmen whose ranks zoomed upwards are Walters who moved from 169 to 5, Thilan Samaraweera who moved from 194 to 38 and Bruce Mitchell who saw his position skyrocket from 186 to 35.

There were quite a few batsmen who saw drastic drops in their positions. Not many were top batsmen. The highest drop was also the top-most one: Sobers dropped 133 places from 22 to 155. Aamer Sohail moved from 63 to 192. Vijay Manjrekar dropped like a stone from an exalted 5 to 130. De Villiers drops from a respectable 18 to a mid-table 89. Why did they drop? I do not really have one single explanation. An explanation has to consider many factors including scoring patterns.

Bradman moved 99 places up from 159 to 60. Sunil Gavaskar moved up 79 places from 166 to 87. Tendulkar jumped 71 places from 153 to 82. Kevin Pietersen moved from 113 to 49. Brian Lara improved his position by 63 places: from 127 to 64. Andy Flower, Gordon Greenidge, Mahela Jayawardene, VVS Laxman, Neil Harvey et al., all moved up by around 50 places.

A final point on the ranks secured by the table. I added the two ranks, rather elementary step this is, just to get the batsmen who were top or bottom in both. Fredericks leads this ad-hoc combined table with a combined rank of 19. Kanhai follows with 21. Watson is in third position with 23. Trott, with an even 12 in both, is next with 24. Glenn Turner and Richie Richardson round off the top-5 with a combined rank of 31. Dudley Nourse, McKenzie, Dexter and Hammond round off the top-10.

Ijaz Ahmed is at the bottom with a combined rank of 398. Herbie Taylor is next at 390. Amiss confirms his high level of inconsistency with 386. John Reid and MAK Pataudi clock in at 386. Michael Clarke is tenth from last with a combined rank value of 359.

A few technical details on the comparisons of player ranks derived from the two methods.
- The Pearson's Correlation Coefficient between the two sets of ranks is 0.431, indicating that the two sets of values have virtually no correlation at all.
- The average of the absolute rank difference is 50.1 indicating that there are a number of very significant changes in ranks.
- A mere 28 players have rank differences in single digits. This is only 14% of the population.

Thus it is easy to conclude that the two methods create two entirely different sets of tables. We have to bite the bullet and select one of these two as the final defining table. I am sure the readers would agree with me that the Test-based Consistency analysis has a lot more going for it and that would be my choice. The reasons are summarised below.
- This method allows for players to compensate for one poor innings with a good innings during the course of a single match.
- Many of the top players have achieved this and have moved up in the Test-based table.
- Many of the top scoring and high performance batsmen are in higher positions in the Test-based table.
- The unit of a Test as the contest which produces the result is the final clincher.

What do we conclude? With the Test-based consistency analysis as the guiding factor and the innings-based analysis as a supporting entity, five batsmen stand out. Saeed Ahmed, Nourse, Trott, Kanhai and Fredericks. Considering the number of runs scored, the quality of opposition faced, the very high second position in the important table and the length of the career, I would place Rohan Kanhai at the top of the list.

To me, the type of attacking batsman that he was, the Consistency index values of 58.2% and 40.6% and the very low level of below-ConZone innings/Tests, Kanhai has been absolutely wonderful in the consistency stakes. Kanhai has played only eight Tests in which he scored 20 or fewer runs. He has 46 Tests in the ConZone range of 40-118. Rohan Bholalall Kanhai is an under-rated batsman who lived right through his career under the shadow of Sobers. However, he was a world-class batsman in his own right and this analysis substantiates that.

I have uploaded a single composite Excel chart containing the data for all 200 qualifying players. This contains the Test-based table, Innings-based table and Rank comparisons. To download/view this file, please CLICK HERE.

"A tale of two penultimate balls" could very well be the Dickensian title of the two glorious Tests over the past fortnight. Sitting on the outside it is easy to say that there were no losers and cricket was the winner. But England would feel gutted. They drew a match they should have won and lost a match they should have drawn match. But they could take a lot of positives from the series. The four centurions were all new players: not Cook or Bell. Plunkett's 11 wickets in the series was impressive. The fight England's lower order batsmen showed on the last day was magnificent.

"The private club" is however another thing. The lopsided scheduling in favour of the troika is already being seen. Why could the Sri Lankans not have been given three Tests and India, four Tests? This will happen in the future. This series cried out for a third Test. Next time it could be a two-Test Australia-South Africa series and a five-Test Australia-India series. This sort of imbalance will be the order of the day.

Amazing, but true. If Sangakkara and Jayawardene were to retire tomorrow, they would have shared 22986 career Test runs between them, split right down the middle.

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