Imran Khan bowls

Thirty-three's the charm: Imran Khan took 171 wickets in his best 33 Tests

© PA Photos
35

The Jury's Out

The best stats measure

Four number crunchers present four above-average options

Peak-33
By Andy Zaltzman

How good was Ian Botham? Overall he averaged 28.40 with the ball, 33.54 with the bat. In the first 25 Tests of his 102-Test career, those figures are 18.52 and 40.48; in the final 25, 42.00 bowling, 23.45 batting. Overall-Botham was very good. Late-slump-Botham, scuttled by injury and time, did not merit selection. Peak-Botham was one of the greatest Test cricketers of all time. Peak-Waqar took 19 five-fors in his first 31 Tests; Increasing-Back-Trouble-Waqar took only three more in his final 56 games.

Cricket needs a measure of how good a player was in his best years. There are, evidently, greater priorities on this often-malfunctioning planet, and in this often-malfunctioning sport, but the career average, at best, needs considerable prodding to reveal its truths, and, at worst, is wilfully misleading. I therefore unveil: Peak-33 - a player's numbers in the best 33-Test phase of their careers.

Peak-33 is based on the 33 matches in which batsmen scored most runs and bowlers took most wickets, rather than the 33 in which they returned the best average.

I could have chosen another mathematically convenient number: Peak-50, for example, the length of the high plateau period of Donald Bradman's Test career; in other words, his entire career excluding his debut (18 and 1), and his final-Test duck. I could have chosen Peak-25, the number of Tests in which SF Barnes bowled a significant number of overs during his legendary 189-wickets-at-16 England career. While no one has come close to Barnes' career average since then - an average itself skewed by a deluge of late-career wickets against a relatively underpowered South Africa - Imran Khan's Peak-25 (from 1981 to 1986, excluding two Tests in Australia in which he did not bowl due to injury) produced 154 wickets at 14.85. Or I could have chosen Peak-200, to appeal to the Tendulkar fans.

I chose Peak-33, however, for the following reasons:

1. It is long enough to require prolonged consistency, even in the modern age of hyper-hectic golden-goose-squeezing schedules.
2. It is short enough to encompass the careers of far more of the pre-war titans of cricket than Peak-50.
3. It sounds better.
4. It could be the sequel to Catch-22.
5. It is the length of the Test career of naughty, naughty Salman Butt, who, with an average of 30.46, does not emerge well from the statistic.
6. It is the atomic number of arsenic. (Thank you, Wikipedia.)
7. It is the number of vertebrae in the normal human spine (coccyx included). (Ditto.)
8. It was the average number of different mystery balls announced by Shane Warne before an Ashes series. If I remember correctly.
9. It is the number of tracer bullets fired by Ravi Shastri in his special commentary research laboratory, in order to ascertain the average speed of a tracer bullet, against which to compare the speed of cricket balls.

I admit that Peak-33 is, at best, in the pre-alpha stage of its development as a statistical measure. It needs to take into account comparative contemporary scoring trends - early 21st-century batsmen predominate - as well as opposition and context. Some interesting candidates emerge, however, in particular Peak-Imran-Khan, whose best outdoes everyone by a considerable margin, and Peak-Jack-Hobbs, whose 1910-1925 average of 65.22 was 20 runs an innings ahead of any Peak-33 achieved at the time.

Perhaps a peak measurement needs to involve some combination of time period, number of matches, and proportion of a player's career, as well as factoring in the impact of a player's performances, rather than the mere quantity of his runs and wickets.

Here, nonetheless, to start the conversation - a conversation which I will in all likelihood be having only with myself - are the top ten Peak-33 players, based on their averages in the matches in which they scored most runs or took most wickets.

Peak-33 top ten batsmen
Player Team Start date Runs Average 100s 50s
Don Bradman Aus 1928-12-29 4835 102.87 19 8
Viv Richards WI 1976-01-03 3483 72.56 12 15
Rahul Dravid Ind 2002-04-11 3329 72.37 11 12
Steve Waugh Aus 1993-11-26 2876 71.90 9 16
Mahela Jayawardene SL 2006-07-27 3785 71.42 14 9
AB de Villiers SA 2010-06-10 3209 71.31 10 14
Wally Hammond Eng 1928-06-23 3410 71.04 13 9
Garry Sobers WI 1957-08-22 3481 71.04 14 9
Sachin Tendulkar Ind 1997-12-03 3650 70.19 15 12
Jacques Kallis SA 2003-09-04 3438 70.16 13 15

Peak-33 top ten bowlers
Player Team Start date Wickets Average 5wi 10wm
Imran Khan Pak 1980-01-29 171 15.90 12 3
Muttiah Muralitharan SL 2003-06-20 250 17.32 23 9
Johnny Briggs Eng 1884-12-12 118 17.75 9 4
Malcolm Marshall WI 1984-03-30 196 17.65 15 3
Jim Laker Eng 1951-07-05 146 17.73 7 3
Richard Hadlee NZ 1982-03-12 196 18.47 18 4
Curtly Ambrose WI 1990-03-23 164 18.84 10 3
Shaun Pollock SA 1997-10-24 153 19.31 9 0
Alan Davidson Aus 1956-10-26 171 19.39 14 2
Waqar Younis Pak 1990-10-18 191 19.56 19 4

Andy Zaltzman is a stand-up comedian, a regular on BBC Radio 4, and a writer

****

Weighted Wicket Probability
By Phil Oliver

Wahab went wicketless in a testing five-over spell in the Lord's Test, but how dangerous was he and how did it impact the match result?

Wahab went wicketless in a testing five-over spell in the Lord's Test, but how dangerous was he and how did it impact the match result? © Getty Images

Cricket is a game of statistics. Every delivery is an event in itself, full of information that can be used to measure player performance - so many variables, so much data. It is for this reason that providing the best possible context and finding the most significant statistics is not easy. We look at scorecards, series averages and career records, wondering who has produced the best performances. Traditional measures can go some way in making these comparisons, but the application of analytics adds the necessary context.

Cricket fans know that the top scorers and wicket-takers in a match have not necessarily performed the best. CricViz's analysis of ball-tracking data allows the quality of each delivery to be measured, enabling more precise evaluation and comparison. That enables the calculation of a Weighted Wicket Probability (WWP) for every ball.

From live ball-tracking feeds, CricViz evaluates each delivery based on six criteria: line, length, bounce, speed, movement in the air and movement off the pitch. We conduct a nearest-neighbour analysis, examining the runs and wickets associated with the 1000 most similar balls in our database based on these criteria.

If the 1000 near-identical deliveries in the database took ten wickets, then the ball in question would take a wicket 1% of the time. This allows the measurement of the most threatening bowlers, spells and sessions. For example, James Anderson's superiority in England's innings-and-88-run victory over Sri Lanka at Headingley this summer was highlighted by his WWP. Anderson's mastery of seam and swing saw him take ten wickets in the match, with an average WWP of 2.13% per ball (meaning his average delivery would take a wicket 2.13% of the time he bowled it). For context, Sri Lanka's bowlers had an average WWP of 1.38% per delivery.

However, the significance of this statistic is that it can offer deeper insight than scorecards and other traditional measures can: the instances where bowlers go unrewarded, the wicketless spells that deserved more or built pressure for team-mates. Wahab Riaz's second spell on day four of Pakistan's win at Lord's in July was one such passage of play.

England were edging towards their target when Wahab returned for a five-over burst. It was a riveting 45 minutes. Jonny Bairstow and Chris Woakes battled hard to repel Wahab's pace and reverse swing. That they succeeded was testament to their skill and concentration, but it was also due to a healthy dose of luck. Wahab's spell read 5-1-8-0, but his repeated beating of the bat meant his WWP was 2.14% per ball in those five overs. In other words he was more threatening than Anderson was in his ten-wicket Headingley masterclass.

In England's second innings as a whole, Wahab's WWP was 1.88%, compared with Yasir Shah's 1.35%, Mohammad Amir's 1.22% and Rahat Ali's 1.05%. He took just one wicket but that was not the significant statistic. WWP is made even more powerful when combined with weighted runs. If the 1000 near-identical balls in the analysis went for 400 runs and took ten wickets, the ball in question would average 40. This would be an easier ball to face than we would expect, based on the career averages of most front-line bowlers.

Weighted wicket probability for England v Pakistan 2016
Bowler Lord's Old Trafford Edgbaston The Oval
Mohammad Amir 1.18 1.38 1.18 1.57
Wahab Riaz 1.67 1.53   1.51
Yasir Shah 1.34 1.22 1.10 1.07
Sohail Khan     1.37 1.28
Rahat Ali 1.19 1.30 1.34  
Chris Woakes 1.66 1.46 1.64 1.63
Stuart Broad 1.42 1.56 1.45 1.39
James Anderson   1.64 1.76 1.54
Ben Stokes   1.66    
Steven Finn 1.33   1.47 1.59
Jake Ball 1.33      
Moeen Ali 1.20 1.07 1.02 0.98

This analysis of ball-tracking data allows for deeper analysis of batting. We can categorise how hard the bowling actually was to face, and not just make judgements based on career records and reputation. Significant statistics inform, prompt debate and contextualise performances. They are often used by fans to make a point, so next time you are discussing unlucky bowlers, perhaps reference to WWP can help.

Phil Oliver is managing editor and co-founder of the cricket analytics app CricViz

****

Batting/Bowling Score based on adjusted average and adjusted strike rate/economy rate
By S Rajesh

Viv Richards was streets ahead of his time in one-day cricket

Viv Richards was streets ahead of his time in one-day cricket © PA Photos

When evaluating batting or bowling performances in limited-overs cricket, the rate of scoring or conceding runs plays as big a role as the number of runs scored or wickets taken. Conventional cricket stats measure the two separately, by way of averages and strike/economy rates. Ideally you'd want the batsmen to have high averages and strike rates, and bowlers to have low averages and economy rates, but it becomes difficult to directly compare a Virender Sehwag (ODI average 35.05, strike rate 104.33) with a Michael Bevan (average 53.58, strike rate 74.16).

A relatively easy and intuitive way to combine the two factors is to multiply the average by the number of runs scored per ball (strike rate divided by 100). For bowlers, multiply the average by the runs conceded per ball (economy rate divided by six). The higher the product, the better for batsmen, while the opposite holds true for bowlers. Doing this for Sehwag and Bevan, we get 36.57 for Sehwag, and 39.73 for Bevan, which is probably a fair estimate of their abilities.

However, there is also the small matter of the changing tempo of the ODI game, and of different scoring rates in different parts of the world. In the 1980s, the average strike rate was 66.52; in the 2010s it is 81.39: an increase of 22%. Since 2000, the average ODI strike rate in India has been 84.34, while in the West Indies it is 74.50. The runs-per-dismissal numbers have changed too, with an 11% increase in the 2010s from the 1980s. To compare batsmen and bowlers meaningfully across eras in limited-overs cricket, it is necessary to take as a benchmark the par performances of that era and, to eliminate the difference in conditions between, say, England and India within the same era, the par performances in the matches the player played.

The Batting Score is the product of the average and strike-rate factors, both of which are adjusted to reflect how much better (or worse) the player is than the mean, in the matches he played in. Sachin Tendulkar, for example, had a career average and strike rate of 44.83 and 86.23 in ODIs, and in the matches he played, the overall batting average and strike rate were 29.93 and 76.45. This means Tendulkar's average was better by a factor of 1.50, and his strike rate by a factor of 1.13. The product of the two, multiplied by 100, gives the Batting Score. (To make the method even more accurate, you could exclude the player's stats when calculating the overall batting and bowling averages.)

This adjustment allows us to compare batsmen and bowlers across eras. The table below shows just how far ahead of everyone else Viv Richards is: his Batting Score is 248.19, while AB de Villiers is the only other batsman with a 200-plus score. Richards' greatness was that he produced numbers that are exceptional even by 2010s standards, despite playing in an era when the benchmarks were far lower: his average was 1.78 times the norm, and his strike rate 1.40 times better. The table also illustrates why Dean Jones was so highly rated as an ODI batsman.

Batsman Average Strike rate Average factor Strike rate factor Batting score
Viv Richards 47 90.20 1.78 1.40 248.19
AB de Villiers 53.63 99.87 1.76 1.20 212.42
Michael Bevan 53.58 74.16 1.87 1.01 188.72
Dean Jones 44.61 72.56 1.69 1.09 183.64
MS Dhoni 51.25 89.27 1.61 1.06 169.89
Sachin Tendulkar 44.83 86.23 1.50 1.13 168.94
Adam Gilchrist 35.89 96.94 1.23 1.26 155.69
Kumar Sangakkara 41.98 78.86 1.51 1.02 154.48
Ricky Ponting 42.03 80.39 1.46 1.04 151.70
Virender Sehwag 35.05 104.33 1.15 1.29 148.30

A similar exercise works for bowlers as well, replacing strike rate with economy rate. In the 98 ODIs that Joel Garner played, for instance, the overall bowling average was 29.66, and the overall economy rate 3.91, compared to his own average of 18.84 and economy rate of 3.09. His average was thus 1.57 times better, and his economy rate 1.27 times better than the overall mean. The product of those two factors, multiplied by 100, gives him a Bowling Score of 199.21. The table below has the corresponding numbers for some of the other leading ODI bowlers, and Garner is clearly the best of the lot, followed by Glenn McGrath, Shaun Pollock and Muttiah Muralitharan.

An added advantage is the ability to compare batsmen with bowlers, through their overall Batting/Bowling Scores, which tells us how much better they were than the average performers in the matches they played in.

Bowler Average Economy rate Average factor Econ rate factor Bowling score
Joel Garner 18.84 3.09 1.57 1.27 199.21
Glenn McGrath 22.02 3.88 1.50 1.21 181.31
Shaun Pollock 24.5 3.67 1.35 1.27 171.3
Muttiah Muralitharan 23.08 3.93 1.39 1.18 163.45
Michael Holding 21.36 3.32 1.37 1.18 162.18
Shane Bond 20.88 4.28 1.42 1.09 154.84
Richard Hadlee 21.56 3.3 1.29 1.18 152.60
Curtly Ambrose 24.12 3.48 1.25 1.21 51.52
Wasim Akram 23.52 3.89 1.31 1.12 146.87
Waqar Younis 23.84 4.68 1.28 0.95 122.4

S Rajesh is stats editor of ESPNcricinfo

****

Strike Rate Ratio
By Kartikeya Date

How do you compare the bowling careers of Andy Roberts and Jason Gillespie, both of whom averaged 26 with a strike rate of 55?

How do you compare the bowling careers of Andy Roberts and Jason Gillespie, both of whom averaged 26 with a strike rate of 55? © Getty Images

Only three things are counted in cricket: deliveries, runs and wickets. The two most common measurements - batting average and bowling average - measure the runs scored per dismissal and the runs conceded per wicket. The limited-overs game has brought with it strike-rate and economy-rate measures - runs scored per 100 deliveries, and runs conceded per over. These measures have endured and even describe thresholds of greatness in some cases - a batting average above 50, or a bowling average below 25. The strike rate for bowlers is calculated as the number of balls bowled per wicket.

All these conventional measures consider players individually. To consider players against specific opponents, one has to calculate these measures for games against those opponents. Here I propose a measure which considers a player relative to team-mates. For example, while Jason Gillespie and Andy Roberts ended their careers with near-identical bowling averages and strike rates (average 26, strike rate 55), what does this tell us about their roles in their respective sides?

The Strike Rate Ratio (SRR) is a measure of the bowler's strike rate to that of all the other bowlers in the team. Over a career, this measure adds richness to a bowler's record by comparing it to an average team-mate in identical bowling conditions.

A few intuitions about the SRR are possible. For specialist fast bowlers, the ratio should be well below 1. For specialist spinners, it is likely to be above 1. Fast bowlers get wickets more frequently than spinners. Further, for strong bowling attacks, the ratio for all bowlers should be close to 1 on either side.

Of the 67 bowlers* who have taken at least 200 Test wickets, 49 are pace bowlers, 17 are spinners and one is Garry Sobers. Five out of the 17 spinners have an SRR below 1. Seven out of the 48 pacers have an SRR above 1.

Bowlers ranked by strike rate ratio (qual: at least 200 Test wickets)
Player Career wickets Strike rate Strike rate at other end Strike rate ratio
Richard Hadlee 431 50.9 79.8 0.64
Dale Steyn 416 41.4 60.6 0.68
Fred Trueman 307 49.4 69.2 0.71
Alan Donald 330 47.0 65.7 0.72
Chris Cairns 218 53.7 73.8 0.73
Darren Gough 229 51.6 70.7 0.73
Waqar Younis 373 43.5 59.0 0.74
Imran Khan 362 53.8 70.7 0.76
Bob Willis 325 53.4 69.4 0.77
Heath Streak 216 62.8 79.8 0.79
Kapil Dev 434 63.9 81.2 0.79
Jeff Thomson 200 52.7 81.2 0.79
Muttiah Muralitharan 800 55.0 69.2 0.80
Malcolm Marshall 376 46.8 58.7 0.80
John Snow 202 59.5 74.6 0.80
Bhagwath Chandrasekhar 242 66.0 82.6 0.80
Ian Botham 383 57.0 70.9 0.80
Dennis Lillee 355 52.0 64.1 0.81
Craig McDermott 291 57.0 69.5 0.82
Andy Caddick 234 57.9 69.2 0.84
Clarrie Grimmett 216 66.9 79.5 0.84
Michael Holding 249 50.9 60.5 0.84
Mitchell Johnson 313 51.1 60.0 0.85
Chris Martin 233 60.2 70.4 0.86
Andy Roberts 202 55.1 63.8 0.86
Graham McKenzie 246 71.9 82.9 0.87
Ray Lindwall 228 59.9 69.0 0.87
Alec Bedser 236 67.4 77.6 0.87
Makhaya Ntini 390 53.4 60.9 0.88
Merv Hughes 212 57.9 66.0 0.88
Javagal Srinath 236 64.0 72.4 0.88
Curtly Ambrose 405 54.6 60.7 0.90
Zaheer Khan 311 60.4 67.0 0.90
Joel Garner 259 50.8 55.9 0.91
Stuart MacGill 208 54.0 59.4 0.91
Wasim Akram 414 54.7 59.2 0.92
Matthew Hoggard 248 56.1 60.2 0.93
Stuart Broad 358 56.8 60.9 0.93
Brian Statham 252 63.7 68.3 0.93
James Anderson 463 56.9 60.9 0.93
Anil Kumble 619 66.0 69.8 0.95
Courtney Walsh 519 57.8 61.0 0.95
Shaun Pollock 421 57.8 61.0 0.95
Brett Lee 310 53.3 55.8 0.96
Glenn McGrath 563 52.0 53.7 0.97
Rangana Herath 332 62.1 62.4 1.00
Steve Harmison 226 59.2 59.2 1.00
Morne Morkel 242 56.4 56.0 1.01
Graeme Swann 255 60.2 59.1 1.02
Richie Benaud 248 77.0 75.2 1.02
Ishant Sharma 209 66.6 64.7 1.03
Chaminda Vaas 355 66.0 64.0 1.03
Peter Siddle 208 61.1 59.3 1.03
Shane Warne 708 57.5 55.7 1.03
Jason Gillespie 259 55.0 53.0 1.04
Andrew Flintoff 226 66.2 62.9 1.05
Harbhajan Singh 417 68.5 63.6 1.08
Bishan Bedi 266 80.3 74.1 1.08
Derek Underwood 297 73.6 67.0 1.10
Abdul Qadir 236 72.6 65.0 1.12
Saqlain Mushtaq 208 67.6 59.5 1.14
Danish Kaneria 261 67.8 59.3 1.14
Nathan Lyon 211 62.4 53.9 1.16
Garry Sobers 235 91.9 78.9 1.17
Lance Gibbs 309 87.8 73.4 1.20
Jacques Kallis 292 69.3 56.9 1.22
Daniel Vettori 362 79.6 64.0 1.24

Morne Morkel, Ishant Sharma, Chaminda Vaas, Peter Siddle, Jason Gillespie, Andrew Flintoff and Jacques Kallis all have an SRR above 1. Gillespie's SRR is 1.04, while Roberts has an SRR of 0.86 over his career. The bowlers at the other end for Australia in Tests that Gillespie played took their wickets at the rate of one every 53 balls. Over the course of his career, Roberts' team-mates managed a wicket every 64 balls. This gives you an overall sense of the teams that Gillespie and Roberts played in. Seen alongside a bowler's individual strike rate, SRR presents a rich picture of what opponents were up against.

Richard Hadlee ended his career with a strike rate ratio of 0.64. In other words, he bowled only 64% of the deliveries needed at the other end for each wicket. His superb career strike rate of a wicket every 51 balls came in a New Zealand side in which bowlers at the other end required 80 balls for each wicket. Hadlee, Dale Steyn, Fred Trueman, Allan Donald, Chris Cairns and Waqar Younis all have a career SRR under 0.75. In Cairns' case it reflected the overall modesty of the bowling of the New Zealand teams he played in. The others were classical strike bowlers in relatively strong attacks.

Unsurprisingly, Hadlee is followed in the all-time list by strike bowlers. Steyn, Trueman, Waqar all have a career SRR below 0.75. Kapil Dev took his wickets at the rate of one every 64 balls, but his career SRR is 0.79. Imran Khan's career strike rate of 53.7 is accompanied by an SRR of 0.76.

Among the spinners, Bhagwath Chandrasekhar, Clarrie Grimmett, Muttiah Muralitharan, Stuart MacGill and Anil Kumble, all have an SRR below 1. With the exception of MacGill, this should not be surprising, since the other four were the dominant wicket-takers in their teams and were not supported by a noted fast bowler for the most part. MacGill played 16 of his 44 Tests with Shane Warne. In these Tests, he took 82 wickets at the rate of one every 42 balls. In Tests where Warne was not present, MacGill's strike rate is 62. Given Australia's fast-bowling strength in this era, it is possible to conclude that MacGill's career SRR is due to the fact that he played an unusually high percentage of his Tests in conditions that favoured spinners.

At their best, statistics suggest explanations and improve judgement. They are not, as is often believed, the alternative to exercising judgement. With this in mind, statistics that measure players relative to their team would be significant additions to the current set of elementary cricket statistics.

*This was written before R Ashwin got to his 200th Test wicket

Kartikeya Date writes at A Cricketing View and tweets here

 

RELATED ARTICLES

 

LOGIN TO POST YOUR COMMENTS

  • POSTED BY Harsh on | October 31, 2016, 9:18 GMT

    The single most important criteria is the ability to turn the complexion of games single-handedly which Don Bradman,Viv Richards,Ian Botham ,Gary Sobers and Imran Khan did more than anyone.Infact in his peak from 1977-82 Botham arguablycreated more dramatic turnabouts than any cricketer including Bradman and Sobers if you remember the 1981 Ashes and 1980 Jubilee test in Mumbai.It reminded one of a dramatic reversal of plot in a Hollywood epic like lightning and thunder coming from nowhere.He literally ressurected his team from the grave.Still it was Gary Sobers who was the more consistent performer.

  • POSTED BY Harsh on | October 31, 2016, 9:09 GMT

    One of the most complex question that within peak period should greater emphasis be given to best performances or to consistency.Even in peak period Lara had a greater string of low scores than Tendulkar but at his best turned the complexion of matches more.Similarly Curtly Ambrose in his peak period turned in more stunning performances even than Malcolm Marshall in term sof turning games.There are disadvantages and advantages of playing for a champion team like Viv Richards and Malcolm Marshal illustrated.Playing for weaker teams arguably facilitated higher scoring for Brian Lara and better wicket hauls for Ambrose and Hadlee.Statistics did injustice to some true greats who never made such lists like Rohan Kanhai or Andy Roberts who were as good as any great on their day.Sadly world series cricket figures are excluded which would have put Viv Richards on an even higher stature or Greg Chappell or Dennis Lillee for that matter.

  • POSTED BY Harsh on | October 31, 2016, 8:59 GMT

    The yardstick in peak era should not be merely of statistics but assess the sum package of percentage score or bowling haul out of team's total,strength of opposition,situation of the game,nature of the pitch and effect on the complexion of the game.No doubt Don Bradman ,Viv Richards and Imran Khan are well deserved candidates to be at the top of the pedestal.However never forget the great performances of Brian Lara ,Curtly Ambrose and Richard Hadlee that single-handedly turned the course of games for weak teams in their peak era.In peak era consistency and best performances have to be assesed.Weighing all factors in peak era Hadlee was best while after Bradman considering his phenomenal domination Viv Richards was 2nd.If you scale the strength of batting side then Lara may almost draw level with Viv.

  • POSTED BY babu on | October 23, 2016, 19:09 GMT

    But when it comes to bowling , its not the same.. imran khan at his peak 33 is the only threatening bowler for opposition..Batsmen would have tried to see him off - so his average is high..I saw imran bowl , he is absolutely pleasure to watch.. But there are more bowlers in his era who impressed me a Malcolm marshall and Richard hadlee - who took 196 wickets @ an average of 17+ and 18+ were better bowlers than Imran khan..bcoz there are other bowlers in theor respecrive team who has good reputation I cant even comment on sidney barnes ...leaving averages behind, Murali's 250 wickets in 33 matches also outstanding and unmatched ...its like Don's 99.8 average when i count no of wickets per match ..its 7.5..; instead of again going for averages - if i go with "FEEL" factor in my illustrated career of 35 years as a spectator ... I have rate few cricketers as unbeleivably great... 1. VIV RICHARDS . 2. IAN BOTHAM 3.Lara 4.Murali 5.ABDevilliers

  • POSTED BY Daniel on | October 20, 2016, 8:48 GMT

    It is very interesting to see AB De Villiers' name appear in the Peak-33 and Batting Score (S Rajesh) tables. His name appears in such illustrious company that you wonder why he isn't considered a great of the game already... The fact that he is the only player, other than Viv Richards, to have a 200+ score is such a revealing stat. It shows his greatness and the gulf in talent between him and every other current batsmen. The Peak-33 and Batting Score stats are the most insightful and revealing for me. Lost the strike rate ratio one, its a very obscure stat measure. I think these "non-standard" stat measures should be easy to explain so that readers can place them in context and get a good picture of what they capture. I would like to see a stat measure of which are the most closely matched test teams in terms of win/loss ratios and close finishes.

  • POSTED BY Rohit on | October 19, 2016, 23:58 GMT

    I have always wondered about measuring value of player to team. A bowler who gets top order wickets is of much value than a bowler who gets tail enders. Is there a way to measure effectiveness of bowler based on value of wicket he has taken. The other aspect is bowler who is good at breaking long partnerships. Again a bowler who consistently can provide breakthrough when required is more valuable. Similarly for batsmen we should be looking at players who build partnerships and scores runs when team most needs them. For example, 50 runs by player when whole team collapses below 150 is much more value than player scoring 100 when team scores 500+ score.

  • POSTED BY Ranojoy on | October 19, 2016, 20:07 GMT

    Strike rate*average list is bloated by not-outs of middle order batsmen. Dhoni and Bevan probably do not get on that list without those (or am I incorrect?).

  • POSTED BY Cricinfouser on | October 19, 2016, 12:42 GMT

    I'd love to see Carl Hooper's Peak whatever. First part of his career he was pretty but ineffective and the second half he was gold. A lot of the proposed stats are ways of comparing players entire careers across eras. I like the Bowling and Batting Scores because I think it allows you to compare batsmen vs bowlers and/or the contribution of an allrounder's batting vs his bowling.

    I agree with many that a fielding score is needed. When Ian Bell was suffering the dropsies a couple years ago you could argue that the runs he dropped more than negated the runs he scored.

  • POSTED BY Saurabh on | October 18, 2016, 15:14 GMT

    M.S.Dhoni is missing from Peak-33. His Peak 33 avg is 76.79 which is 2nd best.

  • POSTED BY Jim on | October 18, 2016, 13:31 GMT

    Peak-33 has definite mileage. Let's make it stick. There have been many so-called star players over the years whose stats seem ordinary. The Peak-33 could give the answer. Neil Harvey comes to mind - always admired by old-timers, but an average of 48 doesn't befit the title of the "greatest left-hander of all time". Well he averaged over 60 for most of his career. The supremely talented, but ultimately disappointing, Graeme Hick gets up nearly to 44, which was quite decent in his era, and much more representative of his true ability, I believe, than his final paltry 31. Not sure for an all-rounder whether you need the same 33 games for each discipline, but Freddie Flintoff seemed to be sometimes batsman, sometimes bowler, but both only for one glorious Ashes triumph in 2005. Could Peak-33 save him from statistical unspectacularity?

  • POSTED BY michael on | October 18, 2016, 13:05 GMT

    What about the "cost" of a dropped catch and attributed to the culprit? If y=the dropped batsman is the next out, the cost would be the extra runs the partnership added. If the other batsman is the next out, the cost would be the extra runs the partnership put on plus any additional runs beyond that point the dropped batsman made.

  • POSTED BY Vinay Kolhatkar on | October 18, 2016, 7:12 GMT

    Excellent. Why is Virat Kohli missing from the ODI table with his approx 50 Avg and approx 90 SR?

    Also, I proposed another measure for subcontinental batsmen and bowlers ---Anglosphere batting and bowling averages--meaning average against Aus, Eng, SAf, and NZ ONLY counting matches played in Aus, Eng, SAf, and NZ.

  • POSTED BY harshit on | October 17, 2016, 23:34 GMT

    A brilliant piece, but too bad it excludes R Ashwin, if possible please update the article with Ashiwn's stats, would be very interested to see how he fares. Kohli's absence from Best batsman based on adjusted average/SR is also very surprising. Would like to know his count even though he does not makes the list

  • POSTED BY Alex on | October 17, 2016, 19:34 GMT

    Of all the proposed new measures, I find Peak33 most appealing - but with the arbitrary 33 marker removed. Instead, it could be termed "PeakN" and a curve plotted, from 1 to the number of years they played. This means Peak1 is the best year of their career, Peak 2 is the best two consecutive years of their career and so on until their career length is spanned. This standardises the measurement to each player's best year, but shows those with plots that are "higher" and "higher for longer" had much more of an impact than someone with one or two exemplary flash seasons. It also standardises for players who miss lengths of time through major injury or non-selection. But I don't think "PeakN" is snappy enough a name for this measure - I'd probably call the plot a "ZaltPeak Ski Run" as a nod to its creator.

  • POSTED BY Alex on | October 17, 2016, 19:21 GMT

    There is a line in statistics that says "correlation does not infer causality". Evaluating a delivery "based on six criteria: line, length, bounce, speed, movement in the air and movement off the pitch" - is not what creates the wicket. Instead it is the factors that are not (and cannot) be measured which affect the batsman's playing style and concentration - for example, cumulative batsman on-field concentration minutes, player comfort levels (heat, humidity, rain etc), distractions (crowd, opposition sledging, imminent roast at the lunch break), rate of pitch deterioration (seemingly less of a factor in recent times!), "scoreboard pressure" and so on. It's a nice attempt but there is always a danger that in analysing the analysis of one's analysis, one can disappear up one's own backside.

  • POSTED BY Bala on | October 17, 2016, 18:59 GMT

    Thanks for the various proposals. The common thread running through these proposals is a need to be able to compare performances across eras and/or geographies. As is natural with such an exercise, various approximations have been made.

    One important measure is a valuation of the wickets taken or runs scored depending on the opposition. It is clear that dismissing a top-order batsman is more difficult than dismissing a tail-ender. Similarly, it is easier to score runs off a pie-chucker who is filling in overs before the new ball than it is off a spearhead. Could the opposition's batting/bowling averages prior to the match be taken to weight the wicket taken or the runs scored (maybe even the weight on the wicket lost when one is out)?. As we can see from the Peak-33 measure, the performance of a player is variable over his career. Maybe we need a running measure (like a Current-10) to weight the wickets or runs. Any suggestions, gentlemen?

  • POSTED BY Malikshahid on | October 17, 2016, 18:08 GMT

    Speaking of new stats to keep, I think in T20 cricket (or even all formats) the stats of fielding errors should be kept - just like they do in baseball. The dropped catches, number of fielding errors (miss-fielding) and number of runs conceded due to dropped catches and miss-fielding would a great start!

  • POSTED BY Nilesh on | October 17, 2016, 14:18 GMT

    Interesting. Zaltzman's proposal was most interesting (and funny to read). I think rather than stick to a Peak-33, it should be extended to a Peak-10, Peak-25, Peak-50, Peak-75 kind of analysis and players who have played atleast twice that many number of tests should qualify for that metric (I have not included Peak-100 because only 1 player - Tendulkar would qualify).

    S.Rajesh and Kartikeya Date's proposals were interesting although they may need some work to make them relevant for each game. I think the Weighted Wicket Probability was perhaps the most radical (least interesting to me).

    I am surprised that no one mentioned something as simple as tracking number of dropped catches or some kind of a metric which weighs the final score of a batsman with the number of missed opportunities. I am sure something similar can be done for the bowlers of whom the opportunities were missed and also for fielders/WK (how many catches did a fielder take versus how many did they drop).

  • POSTED BY B on | October 17, 2016, 12:25 GMT

    I like Zaltzman's Peak33 analysis. Since we are constantly comparing batsmen of different eras, how about one where batsmen's career stats are adjusted and rank-ordered in a grid or matrix form: 1. One axis: Batsmen with no protective gear moving all the way up to today's hper-protective gear, 2. Bat weights ranging from 2 lb 5 oz. (Bradman era) to today's 3 lb.+, 3. Covered vs. uncovered wickets, and 4. Bouncer rule (unlimited to today's coddled and limited). If we did this and calibrated all batsmen to today's standards, Bradman would have a career average of 130+, Sobers 85-90+, etc.

  • POSTED BY Deepak on | October 17, 2016, 12:17 GMT

    Phil Oliver- "We conduct a nearest-neighbor analysis, examining the runs and wickets associated with the 1000 most similar balls in our database based on these criteria." This is Sheer Genius- hats off guys for such an ingenious approach to cricket stats. You guys made my day. Fellow Insights & Analytics Professional

  • POSTED BY Gopalakrishna on | October 17, 2016, 9:49 GMT

    How about this one Indian batsmen batting at number four and five scored 150 plus runs in the first innings of the test {In India's first innings}, while New Zealand batsman batting at number four and five scored ducks in the second innings of the test {New Zealand's first innings} to provide the first such instance in the annals of test cricket

    V Kohli batting at number four scored 211 and AM Rahane batting at number five scored 188 in India's first innings {first innings of the test}, while LRPL Taylor batting at number four scored a duck and L Ronchi batting at number five scored a duck in New Zealand's first innings {second innings of the test}

    Such stats are found in each game whether it be test, odi and t20. it is for the statistician to unearth such interesting tit bits. Hope the writers who have produced tables for the article concur with my views HR Gopala krishna - Cricket Statistician - Bengaluru

  • POSTED BY Gopalakrishna on | October 17, 2016, 9:46 GMT

    All statistical tables are comparative figures according to me. Descriptive stats are the best stats of cricket. Look at this - One more coincidence that is noted with interest is that when these two Pakistan opening batsmen scored triple hundreds, their opening partner was dismissed in nineties. When Hanif Mohammad scored 337, his opening partner Imtiaz Ahmed was dismissed for 91 against West Indies at Bridgetown in Jan 1958. When Azhar Ali scored 302 not out, his opening partner Sami Aslam was dismissed for 90 against West Indies in the ongoing test at Dubai-DSC in Oct 2016 Such stats always make an interesting stats on cricket HR Gopala Krishna - Cricket Statistician - Bengaluru

  • POSTED BY Will on | October 17, 2016, 7:54 GMT

    I've always thought there needs to be a weighted wicket value. Getting Dravid out on a road with a jaffa in the 1st innings after losing the toss, when india are 0-1, is worth a lot more than getting out Chris Martin with a pie going for the big heave-ho when NZ are 500 in front in their second innings. It's a tricky one to model and depends on pitch and atmospheric conditions, the status of the match as well the quality of the batsman, but it would be very revealing.

  • POSTED BY Vivek on | October 17, 2016, 7:51 GMT

    Nice article! In addition to individual-level statistics, a couple of new team stats might be worth considering too: 1. Today count of wins and win % are the metrics primarily used to quantify teams' success. Large & narrow wins are treated alike. Instead if we can award teams points proportional to the magnitude of the win, then we would have a much better metric to truly assess how dominant a team was in an era. Key challenge here is normalizing across different measures (runs diff, wickets left, overs to spare, etc.) - but we can possibly use DL table to convert chasing wins to an equivalent runs diff. 2. Win probabilities have become quite ubiquitous these days. If these models are accurate and the probabilities are archived during each match (like end of each innings), it could provide an interesting way to identify and analyze "improbable victories" (i.e. a team winning a match despite very low win prob at the start of their innings) - such as Kolkata 2001 & Wanderers 2006 ODI.

  • POSTED BY birfda8845345 on | October 17, 2016, 7:24 GMT

    Dismissed batsman's score per wicket taken.

    I want to see who gets out the highest scoring batsmen.

    Sure, there are other factors at play, batsman may have dominated bowler before being dismissed. Part-timers getting lucky vs fatigue batsmen, etc.

    But I'm sure a long-term average would give a better indication of a bowler's strike rate vs the best batsmen.

  • POSTED BY John on | October 17, 2016, 4:49 GMT

    Thanks to all 4 of you. Andy, I am happy to converse with you and not leave you all alone. I may have misunderstood the process but I have Matt Hayden in 33 Tests from Feb 2001 scoring 3501 runs at 72.93 with 15 centuries and 9 fifties - maybe I have chosen an incorrect timespan (is there another 33-test phase in which he scored more runs but at a lower average?) but these stats would seem to make him second on the list behind Bradman. Please let me know where I have gone wrong or, if I have it right, have you missed others?

  • POSTED BY Cricinfouser on | October 17, 2016, 3:49 GMT

    The first thing that came to mind when see the article and as others have commented, is, there is a need for more fielding stats. It's still the lonely sibling of batting and bowling and is so oft forgotten. Think of why a guy like Andrew Symonds was so valuable to Australia, or why Australia's keeping standards have dropped. These things can't be measured at the moment. Where is catch percentage, or drop percentage? For those stats listed here, Rajesh's stand out as being the most accurate and believable. I like where we can adjust averages according to era.

  • POSTED BY Gavin on | October 17, 2016, 2:45 GMT

    Hm, is there any adjustment made on the WWP for spinners as they don't bowl at the same speed as pace bowlers, and are more likely to pitch the ball on a different line (i.e outside leg) to spin it back to the batsman? ALSO WE NEED FIELDING STATS!!!!! And I'd like someone to look at a batsman's median score instead of a mean average.

  • POSTED BY a on | October 17, 2016, 2:29 GMT

    The peak-33 is brilliant and tells us who was the better bowler at his sustained prime. It seems from the results that Imran has been under-rated, compared to Lillee, Wasim and Marshall, as the greatest fast bowler in test match history.

  • POSTED BY Stephen on | October 17, 2016, 1:31 GMT

    SRR is deeply flawed as an objective measure as it just measures you relative to your own team.

    If McGrath, Warne and Gillespe are all nearly equally awesome (by some objective measure) then they will all get an SRR around 1. If Hadlee (Or Cairns) plays with a Team of numpties they gets awesome SRR.

    At the very least you need to weight the SR by comparison with the opposition SR in the same match to obtain a more average skill level. Any stat which ranks Andy Caddick much higher than McGrath is clearly a non-starter. McGrath's Strike rate in games played with Andy Caddick in the opposition was 39.8. Andy Caddick's was 59.23.

  • POSTED BY Alex on | October 17, 2016, 0:31 GMT

    Strike Rate Ratio is a waste of time. All it does is promotes you if you are a good bowler stuck with a bowling attack of alsorans (sorry Sir Paddles and NZ) while it penalises you if you are in a damn good bowling attack with others hunting and capable of taking wickets (e.g. Chris Cairns outgunning both Courtney Walsh and Glen McGrath?!). I know who I'd rather have in my team! Change the denominator to a more comparable yardstick - try Strike Rate Ratio (SRR) as a measure of the bowler's strike rate to that of ALL the other bowlers in the WORLD (Test, ODI and T20) playing over the duration of that bowler's international career. This will tell you how good Paddles was when compared with all other bowlers in his era - not just the his own team members who found it much harder to take wickets than Sir Richard.

  • POSTED BY michael on | October 16, 2016, 21:59 GMT

    There is need for fielding stats re % catches / stumping / run outs made vs missed. This is particularly important when looking at the performance of keepers and setting standards for good keeping. For instance I think that keepers must consistently take 90 + % of chances that go their way or they put stress on the rest of the team. Also a batsman who cannot catch hurts the bowlers who create chances but the stats do not show wickets taken. The bowlers get dropped but the same batsmen are invariably kept on even though their contribution to the team is negative..

  • POSTED BY John on | October 16, 2016, 21:23 GMT

    We needs to know number of batsman's errors. Commentators go on and on (and on) about dropped catches but soon forget play and misses or edges/mishits that do not go to hand. All are equally relevant in assessing how well a batsman is playing. Also, I would love to know how many errors, or average a batsman gets away with per innings - about 5 I should think. It would help spectators understand how much work a bowler has to do to get a dismissal.

  • POSTED BY nigelk6489614 on | October 16, 2016, 19:40 GMT

    None of this is reliable unless it is calculated after making adjustments for pitch conditions. A run scored at Antigua or Adelaide doesn't have the same value as a run scored at Headingley or Eden Park.

  • POSTED BY 2rv64 on | October 16, 2016, 18:49 GMT

    What a joke some very good ideas but I got in contact with Cricinfo about the third idea weeks ago and I was ignored and now you take my idea.