April 16, 2014

A measure for batting and bowling effectiveness in T20

Strike rates and economy rates do not quite tell the whole story. Introducing a new, two-dimensional method of assessing players in T20

Franchise-owner's dream: Chris Gayle and Virat Kohli have high Power and Certainty scores © AFP

The data used in this post is from all the games in the six seasons of the IPL. A player v player chart is available for each IPL game, and so is the number of dot balls, singles, twos, threes, fours, sixes and dismissals for each batsman and each bowler, and for each batsman-bowler pair is available. My thanks to S Rajesh for help with this data set.

Traditionally, batting in cricket occurs in innings, while bowling occurs in spells. Innings and spells have their own rhythms, their own specific phases. The disfiguring compression brought about by the 20-over contest renders spells and innings moot. Every ball counts. Dot balls are considered vital. The batting average, which tells us the number of runs a batsman makes per dismissal, is not necessarily important in T20; neither is the bowling average. Batting strike rates are deceptive, given the small number of deliveries in most single innings.

For example, a batsman could hit three sixes in an innings and still score 21 off 20 balls. Another batsman could score 21 off 20 with a single boundary. The overall effect of either on a team's fortunes could be very different from the other. In the latter innings, the strike would keep being rotated, and runs would also be scored at the other end. In the former, the batsman would probably play a lot of scoreless deliveries and use up overs valuable.

As Sidharth Monga argued persuasively, a batsman trying to fight off a tough period hurts his team terribly in a short 20-over innings. Much better to take chances. If the runs come off the middle or off the edge, like they did for Suresh Raina in the World T20 semi-final against South Africa, it's all good. Players who take a few balls to get going are a liability compared to players who are willing to take chances from the word go.

So how might we measure performance in T20? Most basically, three things matter in T20:

1. The proportion of the total available deliveries that are scored off.
2. The nature of these runs.
3. The cost of these runs in terms of dismissals.

The ideal T20 batsman would hit every single ball for runs, preferably to the boundary, and never be dismissed. The ideal T20 bowler would not concede a single run.

I propose a measure along two dimensions - "Power" and "Certainty" - which can be commonly applied to batsmen and bowlers.

Power is the number of runs scored per dot ball. For a batsman, the higher the Power the better. For a bowler, the lower the Power the better. Certainty is the number of balls faced per dismissal. For a batsman, the higher the Certainty, the better. For a bowler, the lower the Certainty the better.

A batsman can achieve a high Power score by scoring off a large percentage of the deliveries he faces, or by maximising the share of boundaries among his scoring shots. Here is a chart showing how strike rates of some individual batsmen in T20 compare to their Power score.

Most readers will be familiar with the strike rate being measured per 100 balls. I have simply reduced this to number of runs per ball (along the horizontal axis). The number of runs per dot ball is given along the Y axis.

© Kartikeya Date

Chris Gayle is arguably the classic batsman of the T20 era. In the IPL, he has scored 2512 runs off 1568 balls for a career strike rate of 160. This is the best strike rate for any batsman with more than 300 career IPL runs. MS Dhoni has scored 2243 runs off 1589 balls in the league, a strike rate of 141. However, for every dot ball he faces, Dhoni scores 1 run more than Gayle. Gayle has a Power score of 4.32, while Dhoni has a Power score of 5.35. Virender Sehwag, an opening batsman like Gayle, has a career strike rate of 160 and a Power score of 5.21.

Strike rate tells us less about a player's ability to score quickly and consistently than the Power measure. Failures hurt a player's strike rate less than they do his Power score. This is because failures tend to include a higher share of dot balls and lower share of boundaries, while set batsmen tend to score off a higher percentage of deliveries than new batsmen.

For example, consider the following two sets of five innings. Each produced 206 runs in 136 balls. (I'm assuming that all deliveries were legal, and all innings were not-outs.)

© Kartikeya Date

ESPNcricinfo's player-v-player data provides dot balls (which include deliveries off which extras are conceded and deliveries on which wickets fell). For the purpose of this calculation, these two types of deliveries are excluded from the calculation of the number of dot balls. A dot ball is defined as a delivery on which zero runs are scored and the batsman is not dismissed. I have chosen to do this because I use Certainty as a separate metric to factor in dismissal rate.

The best T20 batsmen achieve high Power along with high Certainty. These are batsmen who score quickly from the word go while being dismissed as infrequently as possible. Typically, these tend to be players who are willing to take risks from the outset. Virender Sehwag and Suresh Raina come to mind as examples.

The Power and Certainty figures over six seasons of the league as a whole are 3.49 and 21.8. This means that 3.49 runs are scored per dot ball, and a batsman is dismissed by a bowler once every 21.8 balls.

© Kartikeya Date

The same measures can be used to depict the effectiveness of bowlers. The best bowlers have the lowest Power and Certainty scores.

© Kartikeya Date

T20 games are recorded in great detail. Eventually, it should be possible to measure Power and Certainty scores by over, by batting position, by bowling position or bowling style. I will conclude this post with lists of players who fall into each category over the six seasons of the league. It is apparent that openers have lower Power scores compared to middle-order players. Power and Certainty scores could also be developed by season. This two-dimensional measure will provide a systematic shape to a 20-over innings and perhaps allow observers to assess risk within a 20-over contest. Perhaps it will also allow administrators to tweak the rules, just like they have in ODI cricket in recent times. Eighty-six batsmen have scored at least 300 runs in their IPL careers. Divided into four categories, the players are grouped as follows:

JOURNEYMEN: Power and Certainty worse than the median:
Parthiv Patel, Naman Ojha, Paul Valthaty, DB Das, Adam Gilchrist, Sanath Jayasuriya, Mandeep Singh, Tirumalsetti Suman, Aaron Finch, Eoin Morgan, Mayank Agarwal, Ashok Menaria, Johan Botha, Sunny Sohal, Piyush Chawla.

BIG HITTERS: Power better than the median, Certainty worse than the median:
Virender Sehwag, Yusuf Pathan, Yuvraj Singh, Ravindra Jadeja, Keiron Pollard, Dwayne Bravo, Ross Taylor, Albie Morkel, Abhishek Nayar, Jesse Ryder, Wriddhiman Saha, James Hopes, Azhar Mahmood, DB Ravi Teja, Daniel Christian, Harbhajan Singh, Laxmi Ratan Shukla

ACCUMULATORS: Power worse than the median, Certainty better the median:
Jacques Kallis, Sourav Ganguly, Tillakaratne Dilshan, Manish Pandey, Herschelle Gibbs, Manvinder Bisla, Graeme Smith, Sachin Tendulkar, Rahul Dravid, Robin Uthappa, M Vijay, S Badrinath, Ajinkya Rahane, Brendon McCullum, Manoj Tiwary, Venugopal Rao, Mithun Manhas, Swapnil Asnodkar, Ravi Bopara, James Franklin, Luke Pomersbach.

STRONG BATSMEN: Power and Certainty better than the Median:
Suresh Raina, Rohit Sharma, Chris Gayle, Gautam Gambhir, Virat Kohli, MS Dhoni, Shaun Marsh, Shikhar Dhawan, Mahela Jayawardene, Shane Watson, Dinesh Karthik, Michael Hussey, Kumar Sangakkara, AB de Villiers, David Warner, Ambati Rayudu, Brad Hodge, David Hussey, Matthew Hayden, Irfan Pathan, JP Duminy, Andrew Symonds, Cameron White, Saurabh Tiwary, Dwayne Smith, Kevin Pietersen, Angelo Mathews, Steven Smith, David Miller, Owais Shah, Stuart Binny, Faf du Plessis, Mark Boucher

Eighty-two bowlers have delivered at least 300 deliveries in their IPL careers.

STRONG BOWLERS: Power and Certainty better than the median:
RP Singh, Lasith Malinga, Amit Mishra, Dale Steyn, Zaheer Khan, Munaf Patel, Ashish Nehra, Ashok Dinda, Anil Kumble, Morne Morkel, Ryan Harris, Sunil Narine, Dhawal Kulkarni, Parvinder Awana, Doug Bollinger, Shaun Tait, James Faulkner, Farveez Maharoof, Wayne Parnell, Mitchell Johnson, Shakib Al Hasan, Dimitri Mascarenhas, Mohit Sharma

ECONOMICAL BOWLERS: Power better than the median, Certainty worse than median:
Irfan Pathan, Praveen Kumar, Harbhajan Singh, R Ashwin, Muttiah Muralitharan, Ishant Sharma, Murali Kartik, Shane Watson, Yusuf Pathan, S Sreesanth, Rahul Sharma, Brett Lee, Daniel Vettori, Dirk Nannes, Shahbaz Nadeem, Bhuvneshwar Kumar, Johan Botha, Iqbal Abdulla, Ramesh Powar, Glenn McGrath, JP Duminy, Alfonso Thomas

WEAK BOWLERS: Power and Certainty worse than the median:
Jacques Kallis, Siddharth Trivedi, Rahul Bhatia, Umesh Yadav, Manpreet Gony, Ravindra Jadeja, Ajit Agarkar, Pradeep Sangwan, Suresh Raina, Angelo Mathews, Andrew Symonds, Chris Gayle, Daniel Christian, Thisara Perera, VRV Singh, James Hopes, Ryan McLaren, Rohit Sharma, Pankaj Singh.

WICKET TAKERS: Power worse than the median, Certainty better than the median:
Piyush Chawla, Pragyan Ojha, Vinay Kumar, Albie Morkel, Dwayne Bravo, L Balaji, Shane Warne, Keiron Pollard, Shadab Jakati, Yuvraj Singh, Harmeet Singh, Azhar Mahmood, Jaidev Unadkat, Anureet Singh, Kevon Cooper, Roelof van der Merwe, Yo Mahesh, Chris Morris

Note again that this is how these bowlers and batsmen place over their careers. For individual seasons, some of these players will probably be slotted into categories other than the one they currently occupy.

Kartikeya Date writes at A Cricketing View and tweets here

Comments have now been closed for this article

  • David on April 19, 2014, 10:33 GMT

    @Michael Jones, I was thinking the same thing - "Certainty" is just a new name for an old stat.

    As for "Power", it's not only based on a questionable premise, but the hypothetical table is far-fetched.

    To take the premise first (bringing it down to the manageable unit of one over), why is a batsman scoring 0 4 0 0 6 0 (10 runs, power=2.5) a worse T20 batsman than one scoring 2 2 2 2 2 0 (10 runs, power=10)?

    And as for the table, when you look at the scoring breakdown of the big innings, you see how it has been rigged to produce the desired outcome. Who makes 105 from 55 with 23 singles and only 3 twos? And who makes 60 from 42 with 5 sixes and only 1 four? - And how come the innings of 105 only contained 1 six more than the innings of 60?

    And in any case, another interpretation of the table is that player 1 single-handedly wins 40% of his team's games, while player 2 is a reliable, middling to good contributor. Team balance tells you which player you'd prefer to have.

  • c on April 18, 2014, 15:13 GMT

    On Kartikeya's note: My point is that all efforts at comparisons must lend itself to USEFULNESS of the information presented. In this particular post the inconclusiveness is palpable ALTHOUGH the data mining effort is commendable! That's a lot of effort - even if the info is readily available. I look forward to a future post on how your power/certainly aspect co-relates to results. I'd love to know who the top 10 most IMPACT-ful players in IPL are.

    An observation on power scores and strike-rates: A case, then, can be made Indias WT20 final score of 130 was a Power Score. They don't necessarily win matches. HOWEVER, Yuvraj had 11 singles, 10 dots in 21 balls. In the same period Kohli had 11 singles, 7 dots in 26 balls. Of the x11 Yuvraj ROTATED strike,Kohli had x3 dominating strike-rates for a total of 33 runs in 16 balls. India's scoring RATE was best while Kohli & Yuvraj were together. Meaning: the devil's being painted darker than he is by many regarding Yuvraj's inning.

  • Viyasan on April 18, 2014, 7:22 GMT

    Strong Bowler = Ashok Dinda...I don't think so haha

  • Dummy4 on April 17, 2014, 11:59 GMT

    'Certainty' is just a fancier-sounding name for average/strike rate (runs per dismissal/runs per ball = balls per dismissal), and 'power' is an arbitrarily chosen metric which seeks to reward a particular pattern of run scoring for no apparent reason. A lack of boundaries for several overs does far more damage to the scoring rate than the occasional dot ball. A batsman who hits two boundaries per over is doing a good job even if he does allow a few dot balls in between; a batsman who scores mostly in singles and twos is of little use in T20 even if he does score off almost every ball.

    If a batsman has his eye in against a particular bowler, it's better to keep him on strike against that bowler to hit a few boundaries than give the strike to his partner, who may not score as heavily against the same bowler. The supposition that rotating the strike is always a good thing is based on the idea that it somehow disrupts the bowler's rhythm; one batsman on song disrupts it far more.

  • Dummy4 on April 17, 2014, 10:21 GMT

    Whichever analysis doesn't include the team's progress in the match whenever the batsman is at the crease or the bowler is bowling is incomplete. It will be interesting to see which batsman while at crease the run rate of the team is more than 8. Players like Kohli, Sachin, Dhoni, Kallis even Gayle avoid risk taking when there is a big hitter going well on the other side. But take charge when the onus is on them. That is what makes them so much more valuable for the team than the other players. Same applies to the bowlers as well. When ravi jadeja or bollinger is attacking, I see ashwin slip into a controlling mode rather than attacking mode. Analysis of performance without context is incomplete.

  • Dummy4 on April 17, 2014, 9:35 GMT

    A great attempt to find a better mouse trap. Having said that no method can truly capture and account for evaluating a bowler or a batsman in this form of the game. This is high risk high reward game with too many variables which besides the skills of the players has also to do with situations. Not enough space to elaborate but lovers of the game can understand the numerous variables. Now the main purpose of measurement ultimately is to put the best resources available to play for your team/country. I think most countries are doing very well in putting their best eleven without I am sure much mathematical analysis. Knowledge of the game, leadership that motivates the players, prompt action based on gut instincts and good luck are prime factors. Captains take care of that. That is why Dhoni is a great captain. Lastly any analysis (if used)should be based on recent performances less than 2 years and not on career statistics which mainly serves to keep hypothetical Harsha Bogle busy.

  • c on April 17, 2014, 5:52 GMT

    I cannot find any use in the analysis even if I were an IPL franchise holder. A players' value is FOREMOSTLY of his ability to contribute to a win. All types of players listed above can lay claim to that PRIMARY value if they make a substantive contribution to a win in a given game. Is it possible to extend this analysis to include performances as a ratio to wins and how that may determined the most IMPACTFUL player - as opposed to types of players that the author himself says may have a different flavor based on the year. This is similar to the 'personality types' they hype up here in America. I am supposed to be a Driver. Some days I drive good, some days I don't! Verdict: there is more work to do. SO MUCH data collection to figure what comes out of a pre-fed computer program that speaks of nothing. It's all DATA, no SENSE.

    Some are heard saying "excellent analysis". Exactly HOW?

  • David on April 17, 2014, 3:15 GMT

    IPL season opening match today. Player of the Match? The Accumulator in Chief and Number One Accumulator on the List - Jacques Kallis. 72 runs off 46 balls @ SR 156.52. 11 dot balls. 52% of runs scored in boundaries. Power - 6.54. Certainty - 72.

    Not bad for an old man, huh?

  • Vishwanath on April 17, 2014, 2:26 GMT

    @landl47, and i'd imagine the other person telling you that that was data used in the 2008 and 2009 editions of the ipl in which sanath jayasuriya was aged 38 and 39. remind me of his impact in the two decades of international cricket before that please? :)