The 'real value' of wickets: can anyone match McGrath?

A few months earlier I made an exhaustive analysis of the Real Value of Test runs. Here is the companion article: the Real Value of Test wickets.

Despite the nomenclature, the Real Value of Runs is really the Real Value of Innings, since the innings is the basic measuring unit. There is no sensible way to measure the Real Value of individual runs. A run is a run, and it is not correct to say that a winning hit for runs has much more value than previous runs in that innings.

The bowling is cheese to the batting chalk. Each ball bowled is a harmless one - maybe a dot ball or a few runs being conceded - unless it results in a dismissal. That dismissal could be that of Don Bradman on 0 or Bhagwath Chandrasekhar on 22, it could be a second-ball dismissal of a top batter or a late-order wicket in a collapsing innings. Hence, it is essential that we determine the Real Value of a wicket using various related factors. We must treat the bowler's spell as a collection of balls delivered, runs conceded and wickets taken - unlike an innings by a batter that is evaluated in its entirety.

There is another angle here. A bowler may bowl upwards of 200 balls in an innings, and only a few (at most ten) of these are successful, resulting in the capture of a wicket. What about the other deliveries? The bowler has bowled at top batters, bowled to a plan, kept one end under control, and helped their team in different ways. That means that there are several other factors involved, such as the bowling average for the innings, the quality of batting, the match status, the result. The wicket could have been taken in Hamilton in 2002 (Runs per Wicket 14.1) or in Hamilton in 2004 (RpW 53.0). It could have been in a narrow win or in a dead draw. These wide-footprint values cannot be directly assigned to specific wickets but to the complete bowling effort in the innings (hereafter referred to as "Innspell"), and then we go down to the individual wickets.

The bowling exercise is thus substantially more complicated than the batting exercise. I may not provide explanations in full here, and the readers must assume that due care has been taken while building the methodology wherever details are missing.

My Bowler Performance analysis - Red Cherry 25 or RC25 - took the Bowling Innspell as the base unit and worked on that, while here I take the wicket as the base point. There may be many similarities between the two, but the premise is fundamentally different. Let us now move onward to the overall methodology.

At match level (max 25 pts)

Pitch Quality Index for Match-half or Match, as appropriate
Wickets taken in low-PQI situations (bowler-friendly pitches) are valued lower than wickets taken on batter-friendly pitches. Using historical data will not work in this regard; Hamilton and Headingley are two venues, among many, that have had wild swings in pitch behaviour within a short period of time.

The opposition Batting Strength
Bowling against a team with a high Batting Strength index is far more difficult than bowling to a team whose highest batting average is, say, around 30.

Location/Result/Relative Team Strengths
I do not believe that winning is everything; however, winning should be rewarded. Similarly, drawing a match should get more credit than losing one. And bowlers win matches. Doing well away (winning or drawing) should be rewarded too. Three results, three locations and five strength-comparison indices lead to a complex matrix, and determine the points to be allotted to the team's bowlers.

Match Status
The Match Status at the start of the match is neutral, tweaked only by the Relative Team Strengths. Then, at the beginning of the second innings, the Match Status is wholly dependent on how many runs the team bowling second has behind them. The third innings presents more possibilities: the lead/deficit inherited and the sort of fourth-innings target to aim for play a significant part in determining the Status. In the fourth innings, bowling teams have defended anywhere between one run and 835 runs. Using the match-to-date RpW, the difficulty of the chasing task is calculated and Status points allotted. The points for each innings are determined after an analysis of all these conditions. The total is allotted to the bowlers in proportion to the wickets they took (80%) and the overs they bowled (20%). This is to give some recognition to those who bowled long spells, while keeping in mind the importance of taking wickets.

Courtney Walsh's wicket of Craig McDermott in Adelaide in 1993, denying Australia a tie by one run, is worth 30 points, the most any one fourth-innings wicket can be worth under this analysis © Getty Images

At Innspell level (max ten pts)

What is the bowler average for the innings? That plays an important part in determining this value. In the recent Karachi Test, Mitchell Starc took 3 for 29 (9.67), Pat Cummins 1 for 39 (39.0) and Mitchell Swepson 2 for 32 (16.0) in the first innings. It is logical to allot the most points on this measure to Starc, fewest to Cummins and somewhere in between to Swepson.

At Wicket level (Batter-related: max 60 pts)

A. Which batter was dismissed? (25 pts)
Here I use the Career-to-Date Location Average to get a clear indication of the quality of the wicket. I assign 25% weightage to Recent Form (RF10) - that is, the average in the batter's last ten innings. Kumar Sangakkara (RF10 Ave=118), Garry Sobers (112), Viv Richards (109) and Alastair Cook (110) returned Bradmanesque averages during their peaks, and bowlers' dismissals of such batters should be rewarded.

B. At what score was the batter dismissed? (25 pts)
Dismissing Bradman on 152 is an achievement and should help the bowling team. However, dismissing Bradman on 0 is magical, making a huge difference to the bowling team's fortunes. To qualify for these points, the batter must be dismissed at or below the CTD Location Ave-RF value, used in parameter A.

C. Partnership preventer (10 pts)
This is a special measure that recognises bowlers who did not allow a partnership to develop. Using the batting team's RpW as the base, I allocate a maximum of ten points, depending on the partnership that was formed. The allotted maximum value comes down as we move forward in the innings. By definition, this measure rewards all quick wickets in the opening spell. For example, the three wickets taken by Irfan Pathan in his opening-over hat-trick against Pakistan in Karachi in 2005-06 get 10.0 points each. An interesting fact emerges when we look at the summary of this measure: the average for fast bowlers is 3.24 per wicket, while for spinners it is 2.68.

At Innings Status level (Inns 1-3: 15 pts; Inns 4: 30 pts)

What is the target? What is the score? How many wickets are left? These are the questions used in the Innings Status, determining the points allotted to each wicket. The first three innings are treated differently from the fourth innings. It should be recognised that the moment before Courtney Walsh dismissed Craig McDermott in Adelaide in 1992-93, the score was 184 for 9 and Australia were one run away from tying the match. There is no other comparable situation in Test history. So this wicket is allotted 30 points, the maximum possible.

Let us now move on to a schematic of the process. The article is current up to and including the recent Trent Bridge Test between England and New Zealand.

The four match-level measures are determined, summed up and allotted to each wicket equally. The highest score on this measure is 23.8 out of 25.

The Innspell value is determined and then allotted to each wicket.

The three wicket-level values are determined and allotted to the wickets. The highest for the Batter Score measure is 22.7 points and for the Batter Dismissed measure it is 21.1 - both for the wicket of Bradman in separate instances. In the Partnership Preventer measure, wickets of batters dismissed at 0 and no partnership runs are allotted the maximum of 10.0 points.

Then the innings situations are analysed and status values allotted. The maximum values achieved for the four innings are 12.5, 11.8, 13.8 and 30.0 respectively.

A sum of all these values leads to a figure between 10.0 and 73.43 for the 73,644 Test wickets taken in cricket history. This is then mapped on to a value between 0.500 and 1.905, which becomes the Wicket Quality Index (WQI). The WQI average for the 73,644 wickets is 0.995, confirming that the spread is almost perfect. However, the average WQI for pace bowlers is about 2% above par and for spinners about 3% below par. The main reason for this is that pace bowlers tend to take early wickets and spinners tend to need to plan for their wickets, which takes time. Hence, while converting the interim points total to WQI, minor tweaks are applied for all the wickets. In the final reckoning, pace bowlers took 49,832 wickets at an average WQI of 1.001 and spinners 23,803 wickets at 0.982. The distribution of WQI values is below.

Note how symmetric the distribution of the WQI values is. Only the 1.20 to 1.30 range is slightly off. The highest frequency is at the 1.00-1.05 group, and the extreme groups have very low values. As mentioned, the mean is 0.995. The standard deviation is 0.198, and the coefficient of variation is a very acceptable 0.199.

The table above gives us an idea of how the WQI values have been distributed across the four innings. The WQI average is slightly below 1.00 in the first two innings and just above 1.00 in the latter two. This is understandable, since wickets get to be more valuable heading towards the end of the match. Third-innings wickets are the most valuable, since the match is still in the balance then. In the fourth innings, quite a few wickets are inconsequential, in wins by big runs or by upwards of five wickets.

The first table below is arguably the most important. It lists the top wickets of all time by WQI value. Readers might expect the table to be dominated by Bradman dismissals. But Bradman too had his lows in CTD values, and he went gone through an RF average of around 40 just after the Bodyline series. This means that there is good representation all around.

The top wicket of all time is the dismissal of Bradman by Hedley Verity at Lord's in the 1934 Ashes. Australia, after following on, were 43 for 1 when Bradman walked in and was subsequently dismissed for 13. Australia were in deficit by only 113 runs at the time, making the dismissal a crucial one. A target of 150 would have been tough to chase. Close behind comes another Bradman dismissal, this time by Bill Voce in Australia's first innings in Sydney in 1936. Coming in at 1 for 1, Bradman was dismissed first ball. England won by an innings.

Third is a modern dismissal, from Edgbaston in the 2005 Ashes. Australia, chasing 282, had Adam Gilchrist arriving at 134 for 5 and Ashley Giles dismissing him for 1. Though Australia recovered, they lost the match by two runs. Quite close to this comes the opening-day dismissal of Everton Weekes in Kingston in 1953-54. Weekes was on a roll, with CTD and RF values well over 70. Trevor Bailey's dismissal of him for 0 eventually led to a comfortable win for England.

Fifty years later, in Kingston, after two 300-plus innings totals, Brian Lara walked in at 15 for 3 and was dismissed immediately by Matthew Hoggard. West Indies were bushwhacked for 47 and England won comfortably. The next four dismissals are those of Bradman by Dick Pollard for 7 in a drawn match, of Bradman by Vijay Hazare for 13 in a draw at the SCG, of Sobers, with a high CTD average, for 0 by Fred Trueman, and of Bradman by Neville Quinn for 2 on the opening day in Melbourne in 1931 (albeit in a match Quinn's team, South Africa, lost).

The best fourth-innings wicket was Giles' dismissal of Gilchrist, as we have already seen. However, in the table above, I look at those fourth-innings wickets that were primarily responsible for close wins. This table is ordered on the fourth-innings Status points. Of course, if a top batter were to be dismissed at 190 for 9 with a couple of runs needed to win, that wicket would have topped the previous chart. But that has never happened. Most of these wickets are of late-order batters.

The best wicket at the end of the fourth innings was that of McDermott, who was dismissed when a single was needed to tie. It is the only wicket to score a maximum 30 points on this parameter. The second wicket is that of Michael Kasprowicz by Steve Harmison (was it out at all?) in the 2005 Ashes classic at Edgbaston. This is followed by three wickets taken when three runs were needed to tie in each case: the first is the wicket of Jeff Thomson by Ian Botham in Melbourne in 1982, and the next two are from the famous 1902 Old Trafford Ashes Test.

In sixth place comes the first top-order wicket - No. 3 Azhar Ali's dismissal by Ajaz Patel. New Zealand won that pulsating Abu Dhabi Test against Pakistan in 2019; Ajaz also features on the table with the previous wicket of the innings, Hasan Ali's.

Above is the table of top Innspells - from an aggregate WQI point of view. The list does not feature Anil Kumble's and Ajaz's ten-wicket hauls, but it does contain quite a few eight-wicket hauls. This is owing to the combination of factors that go towards constituting the WQI: the quality of batters, the timing of the dismissals, the innings patterns, the result, the margin, the location.

The two Jim Laker performances at Old Trafford in 1956 bookend the 1994 Devon Malcolm masterclass, when he tore through a strong South African line-up and secured a great win at The Oval. Then comes Stuart Broad's Trent Bridge pre-lunch blitz in 2015 against Australia, followed by Richard Hadlee's famous effort in Brisbane in 1985 - the top-rated RC25 performance. No surprises among the performances that follow. The only interesting point is that the top 15 contains only one effort in a drawn Test - Glenn McGrath's at Lord's in 1997.

This table, along with the next, represents the most devastating bowling performances in history. They look at the top Innspells, but from an average - rather than aggregate - WQI point of view. The first table has a four-wickets-minimum cut-off. Let us go back to the SCG in 1936. After a good England score, a strong Australia side started their innings on a pitch transformed by a thunderstorm. Voce dismissed Leo O'Brien for 0, Bradman for 0 (a second consecutive duck), Stan McCabe for 0, and Jack Fingleton for 12, and the score was 16 for 4. For some reason Voce bowled only eight overs, and finished with 4 for 10. This Innspell averaged 1.605 on WQI.

In the first tied Test, in Brisbane in 1960, Wes Hall dismissed Bob Simpson for 0, Neil Harvey for 5, Norman O'Neill for 26, Les Favell for 7, and finally took the key wicket of Richie Benaud. In Kingston in 2009, England took the field trailing by only 74 runs. A few hours later they went down to an innings loss, having been dismissed for 51. Jerome Taylor took five top-order wickets for single figures, averaging 1.55 on the WQI.

Mohammad Asif's four top wickets in Kandy in 2005-06 and Cummins' effort in the Indian capitulation of 36 in Adelaide in 2020-21 follow next - separated by 0.003 on the WQI.

This table recognises top Innspells of six wickets or more, ordered, again, by the average WQI. The list is topped by Broad's demolition of South Africa at the Wanderers in 2015-16. He dismissed the top six batters for 17 and averaged 1.45 on the WQI. Broad also has the sixth-placed performance on this table, with his opening-day show against Australia in Nottingham in 2015 - his 8 for 15 averaged 1.38 on the WQI.

In second place is Ebadot Hossain's very recent 6 for 46, which resulted in one of the most amazing upsets ever - Bangladesh over New Zealand, away. In third place is Fazal Mahmood's famous six wickets against West Indies in Dacca in 1958-59. Next comes Lance Gibbs' marathon effort against England in Georgetown in 1967-68, which almost secured a win for West Indies, followed by Muthiah Muralidaran's third-innings destruction of India (in the company of Ajantha Mendis) in 2008.

Five of these performances are within a range of 0.2 of each other on the WQI. The last performance on the table, Hugh Tayfield's valiant 7 for 23 in Durban in 1949-50, is the only one by a bowler ending on the losing side; Neil Harvey's magnificent unbeaten 151 carried Australia through in a memorable chase there.

This table considers the leading performances by average WQI performance - at a match level. The cut-off criteria are a minimum of five match wickets, and at least two in an innings. Neil Hawke took six wickets in the match in Port-of-Spain in 1964-65, all of top-order batters (Conrad Hunte, Bryan Davis twice, Rohan Kanhai, Basil Butcher and Seymour Nurse) for single figures. His WQI average was a stunning 1.44. Suranga Lakmal's five wickets, away in Bridgetown in 2018, were similar. Among Sydney Barnes' eight victims in Melbourne in 1911-12 six were top-order batters. All five of Danny Morrison's wickets against Sri Lanka in Wellington in 1990-91 were from the top seven. Voce's first-innings effort in Sydney has already been chronicled; in the second innings, he took two other top wickets out of three.

Now for career performances. I have organised this data in two tables: the first one for bowlers who took between 100 and 200 career wickets, the second one for those who took over 200.

England's early 20th-century left-arm spinner Colin Blythe averaged nearly 1.1 on the WQI for his 100 wickets and leads the table. Pathan, who too just about reached 100 wickets, is next, with an average WQI value of 1.08: across his career he took many top-order wickets. Mohammad Asif, despite having a short career, had an excellent one as far as quality of wickets was concerned, and he is just a decimal point away from Pathan. Alan Davidson, known for his incisive and economical spells at the beginning of innings, comes in next.

There is a decent representation of spinners here, with five in the top 15, led by Laker in fifth place.

The value in the last column is the percentage of high-WQI-value wickets in the bowler's overall tally. Fazal Mahmood is the only bowler who has more than 10% such wickets. Asif is close with 9.4%.

The table for bowlers with over 200 wickets is led by McGrath. Vernon Philander is only a fraction away. Then come Malcolm Marshall, Ray Lindwall, and a surprise in the form of Andrew Caddick. The top ten is a who's who of fast bowlers.

To provide additional information, I have listed five high wicket-takers who missed out on the top ten. Shane Warne, James Anderson and Broad have WQI averages higher than 1.02, while Courtney Walsh and Murali just about break even. Murali took a lot of late-order wickets. Kumble has an average WQI value of around 0.99. Overall, spinners find it difficult to get on to the table. Ravindra Jadeja is the highest-placed spinner, in 14th place, with an average WQI of 1.037.

McGrath took 563 wickets in his Test career. Of these, 142 (25%) were late-order wickets. Most of these wickets would have had WQIs between 0.7 and 0.9. He made up for this with high WQI values for the top- and middle-order wickets he took, to achieve the relatively high value of 7.5% for wickets with WQI values above 1.4. Lindwall tops on this count, with 8.3% of his wickets being such high-value ones.

This is a special table that lists teams that did well on the WQI front in a particular innings. At the top is South Africa's dismissal of Australia for 47 in 2011-12 in Cape Town. They secured 14.53 points there. The first eight Australia batters scored in single figures, totalling 21 runs between them. The second place is secured by West Indies in the 51-run demolition of England that we encountered earlier. Also here is the Broad match at Trent Bridge: England aggregated 13.86 in that steamrolling of Australia. The one relatively high score in in the top five is Australia's dismissal for 169 of the strong Indian team in Bangalore in 1998. The two 400-plus first-innings totals certainly helped to increase the WQI values in the match.

Finally, a table of the best ten-Test streaks from the point of view of average WQI. England left-arm spinner Tony Lock leads here. In a ten-Test streak in the mid-1950s, he took 42 wickets at an average WQI of 1.17. He looked more the English destroyer than Laker during this period, although Old Trafford 1956 changed all that. In second place is McGrath, with 35 wickets at an average of 1.166, though he did not playing in quite a few Tests in that period in the early 2000s. He returned a low number of wickets but most were key wickets.

Pakistan left-arm spinner Iqbal Qasim follows close behind, and his streak includes the famous 1987 Bangalore Test, won narrowly by Pakistan. A hair's breadth behind is Broad, whose streak in 2015-16 fetched him 39 wickets. All his matches were against strong teams, and included the 6 for 17 already discussed. Barnes took an astonishing 69 wickets in a ten-Test streak in 1911-12.

Conclusion
Viewed from one angle, Walsh's dismissal of McDermott in Adelaide could be termed the greatest wicket of all - one run needed to tie the match. However, the fact that McDermott was a lower-order batter prevents me from saying this. If the batter had been Allan Border, I would probably anoint that wicket as the greatest of all. As such, I would say that any Bradman dismissal for a low score - such as Verity's and Voce's - resulting in a win for the other team should qualify for this distinction. Note that a few other low-score dismissals of Bradman - such as Quinn's, Alec Bedser's and Eric Hollies' - did not stop Australia from winning comfortably. Hollies made sure that generations of cricket analysts, myself included, did not need to set a six-character limit for career batting average, and that the number 99.94 attained mythical prominence - but that match was virtually won by Australia before Bradman even padded up.

In general, spinners have found it a little difficult to gain high WQI values. They take their time and bowl more balls per wicket, whereas fast bowlers have the first bite of the cherry - their early strikes are necessary to keep the batting team on the back foot, and so they gain their marginal benefit in the values.

What about Warne's "ball of the century" (although I do not think it was)? That dismissal of Mike Gatting scored quite highly: 1.36 WQI points. However, as far as the analysis goes, the next wicket, of Robin Smith, received 1.47 WQI points. The reason is simple. Both were quick wickets to follow earlier wickets. But at the time Smith was a far better batter than Gatting - Career-to-Date average of 47 vs 35, and Recent Form average of 43 vs 29. Unfortunately, the WQI cannot capture the turn imparted - according to some pundits, close to three feet in the case of the Gatting delivery.

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

Stats feature

The 'real value' of wickets: can anyone match McGrath?

RELATED ARTICLES