How much risk do Kohli, Pant and KL Rahul take when they attack in the IPL?
Two of these players are more risk-averse now than you might think
Himanish Ganjoo24-Mar-2022In the 2018 IPL, KL Rahul averaged 56 at a strike rate of 139 in the first 16 overs of the innings. Since 2020, he averages 75, but his strike rate has dropped to 127. What is the source of this transition in Rahul’s scoring profile? An analysis of shot types can unearth the answers – in the process breaking up a batter’s record into a product of shot selection and execution.Run-scoring begins with intent – some batters show lower strike rates because they decide to be “anchors” or strike rotators in the middle overs, owing to team composition, while others decide to attack spin in the same phase. Virat Kohli and Nicholas Pooran are prime examples of the two extremes. Kohli’s hitting range against pace is exceptional, but like the more recent version of Rahul, he decides to play a sedate role for Royal Challengers Bangalore in the middle overs, milking the spinners, going at a strike rate of 114. Pooran, in contrast, goes at a strike rate of 180.3 in overs seven through 11 in the IPL. Data about shot types, recorded as part of ESPNcricinfo’s ball-by-ball commentary, can be used to measure these differences in attacking intent.The scatter plot below shows the runs-per-ball (RPB) and control percentage for all shot types, considering data since IPL 2018. The vertical line halfway down the X axis shows the average run rate in this time (1.29 runs per ball) and the horizontal line shows the average control percentage, 73%.Related
Do match-ups work in T20? The data says yes
Kohli vs Rohit: what the numbers say about their IPL impact
How many boundaries should a team attempt in a T20 innings?
What is the template for a successful chase in the IPL?
A first glance at the plot confirms a well-known principle of batting: risk is roughly proportional to reward. Control percentages decrease in the general direction of increasing RPB, with some exceptions. Secondly, the lines of average RPB and average control neatly classify the different shot types into three kinds. Non-attacking (NA) shots have high control and low RPB. These are defensive or strike-rotating attempts. Then we have safe-attacking (SA) shots, which are high-scoring, without much compromise on control, like the various drives. Finally, risky-attacking (RA) shots like slogs and upper cuts are high-risk and high-reward. For the purposes of this analysis, we include cut shots in the RA category.
The exact shot played off a particular delivery depends on the line and length of the ball and the field placement, but classification into these three broad categories helps distil the intent of a batter in a phase of the innings on average, isolating it from the results.As an example of the utility of this classification, let’s look at the middle-overs phase. In the younger days of T20, this was a period of consolidation and wicket preservation, but the more elite batters have now started attacking spinners in this phase to increase returns. The plot below breaks down the intent of different batters in overs seven to 16. The horizontal axis shows the percentage of attacking shots including safe and risky types.Attacking Percentage = percentage of SA + percentage of RA shotsA larger value means the batter attacks more. The vertical axis shows the difference in percentage between the two kinds of shots – a negative value means the batter prefers safer shots, even when deciding to attack.Flicks (18%) and cover drives (28%) are the most used shots in this middle phase, showing how batters look to minimise risk and work the spinners around. KL Rahul, Rahul Tripathi, Suryakumar Yadav and AB de Villiers stand out as batters who attack a lot but prefer these safer, classical shots. To their left, Virat Kohli also mostly plays safe shots, choosing to drive 32% of the time. In contrast, Shikhar Dhawan and Quinton de Kock prefer cut and pull shots, playing both these shots 26% and 24% of the time respectively. Ishan Kishan and Rishabh Pant also each play these two shots about 20% of the time.
In the powerplay, as shown below, most batters prefer safe-attacking shots with the aim to preserve wickets while using the gaps in the inner field to gather boundaries along the ground. de Kock, Jonny Bairstow and David Warner are the only currently active players who play more RA shots than SA shots in the powerplay while attacking more than 50% of balls faced. Ruturaj Gaikwad and Kohli both languish in the bottom left corner of the plot, lagging in attacking shot percentage and choosing mostly safe-attacking shots.
The variation of intent of a batter in different phases can be modelled by the Intent Index, which measures the relative frequency of playing a class of shots in a phase of the innings.Intent Index = 100 x (frequency of kind of shot for batter divided by frequency of kind of shot for all batters) – 100Here, the baseline considers all players. The bar chart below shows the Intent Index for Kohli. Looking at the first bar, Kohli has an Intent Index of -43.5 for risky-attacking shots in the powerplay, which means he is 43.5% less likely than the average batter to play these shots. The Intent Index thus helps us normalise and quantify the attacking intent of a batter.
Moulded in an ODI and Test batting framework, Kohli mostly eschews risky shots, preferring safer options even while deciding to attack in the middle and slog overs in T20. He plays non-attacking shots 19.4% more frequently than the average batter. On the other hand, Pooran shows heightened hitting intent in the middle overs, playing high-risk attacking shots much more frequently than the average batter.
This intent profile can also be useful in analysing the varying roles of a batter in a line-up, which is where we come back to KL Rahul and his transition to being a safer batter in recent seasons of the IPL. The arrow plot below shows the changes in Rahul’s Intent Profile between the two eras.Firstly, he has become much more conservative in the powerplay. In 2020 and 2021, his RA Intent Index was -16, which meant he played 16% fewer risky-attacking shots than average, whereas in 2018 and 2019, it was -9.6. His NA Intent Index too has gone up in the first six overs, indicating that he is defending and working the ball much more.In the first five-over block of middle overs, the earlier version of Rahul used to play risky-attacking shots 36.5% more often than the average batter, while in the past two seasons he plays them 6.5% often. In place, he plays more safe-attacking shots now. However, in the second half of the middle overs, his RA and SA Intent Indices have both shot up, although he still plays risky-attacking shots with average frequency overall in the powerplay.
Another interesting study in the disparity in intent between the first and last two seasons of this period in the IPL is Pant. In 2018, he had a monstrous season, striking at 173.6 while averaging 52. In 2019, his strike rate was still a humongous 162. This figure dropped to 114 in 2020 and went back up a bit to 128 in 2021. The execution of his attacking shots deserves a separate discussion, but as the plot below shows, Pant’s frequency of RA shots dropped drastically across innings phases. While in 2018-19 he was starkly clear of the average batter in terms of RA shots played, in 2020-21 he fell to below average in the powerplay and close to average in the final nine overs. In place of RA shots, his NA shot Indices have increased massively. The reason for Pant’s strike rates falling off is mostly a change in intent.
The Intent Index can be employed to visualise the attacking intent of a particular batter against different bowling styles, helping identify match-ups for bowling sides and areas of concern for the batting group. Here is the list of Intent Indices for Kohli and Pant, broken down by bowling type and phase, for the previous four IPL seasons.
For example, facing left-arm pace, Kohli is 42% less likely than average to hit an RA shot, and 19% less likely to hit even an SA shot in the first six overs. He hits far fewer RA shots than average in all phases against all bowling types. Pant, on the other hand, hits more RA shots than average at nearly all times in the innings facing all bowling types.
Run output begins with intent, which we have spoken about, but it ends in execution. We have seen above how certain batters have a preference for certain shots, which is dictated firstly by the areas in which bowlers bowl to them, but also by their shot-making strengths. Different players employ different shots as attacking or strike-rotating options, and the balance of run output and control on each shot type varies between batters.Breaking a batter’s record down by shot type enables an analysis of these minutiae – which shots are productive, which ones risky, and which ones give middling outputs. Below is the break-up of Virat Kohli’s record on attacking shots facing pace. The Strike Index shows how far away from the average batter Kohli is playing that shot against pace. The Control Index shows how his control fares in relation to the average player playing that shot.Strike Index = 100 x (Batter’s SR for shot vs pace or spin divided by the average batter’s SR for shot vs pace or spin) – 100Control Index = 100 x (Batter’s control percentage for shot vs pace or spin divided by the average batter’s control percentage for shot vs pace or spin) – 100 Kohli’s classical approach to hitting ensures that when he decides to slog, he maintains 37.2% better control than the average batter versus pace, returning a strike rate of 200. On the cover drive, he is 10.2% less likely to be in control compared to the average batter, but his strike rate is 25.5% better. His whippy, bottom-handed on-drive is also a productive shot, giving him a 25% better strike rate than average, while his control remains close to average.
If you look at the break-up for the same shots versus spin, Kohli uses the various drives to gather runs along the ground in the middle overs. Also, he has a low strike rate and low control when he slogs spinners. His Control Indices hover around the average, except on the cut where is somewhat better, and the slog, where he is worse, and his Strike Indices are mostly negative. Kohli’s struggles against spin are not just a matter of intent; his execution is sub-optimal as well.
Take a look at Rohit Sharma’s shot-making tendencies against pace bowling. He pulls quick bowlers often off the front foot; the shot makes up 10.3% of his strokes, fetching him a strike rate that is 29% better than average.Against spinners as well, Rohit’s pull shot has a Strike Index of 56 and a Control index of 8. His other productive shot is the cover drive, which he uses to pierce gaps in the off side during the powerplay. Unlike Kohli, he fails at executing the slog well against the faster bowlers, with about 50% less control than average.
A batter will choose to execute the shots he feels most comfortable with to maximize his returns. Bowlers, armed with a breakdown of a batter’s shot types, should ideally seek to make him hit shots he is not adept at. In summary, a batter’s execution record will show the overall result of these individual strategic battles.To summarise how well a batter has been able to execute his chosen shots, we can calculate his Average Strike Index and Average Control Index. These numbers are weighted averages of the SI and CI for each attacking shot played by the batter. They tell us how often a batter plays his best shots (a batter playing a shot he executes well a high number of times will automatically improve these numbers), and how well he plays them compared to the average batter, in terms of both run output and control. For the ASI and ACI, we will consider the batting hand and bowling style for each shot. Note that since we only use attacking shots, the additional context of the innings phase is not needed. The innings situation only decides a batter attacks. We are now concerned with what happens they do.
The plot shows the Average Strike and Control Indices. A positive value means that the batter does better than average while attacking, compared to the average batter. The value indicates the percentage by which he is better or worse. For example, Pooran strikes 16% faster than the average batter when he attacks, while exerting 4% higher control.The first remarkable thing about this plot is the presence of Chris Gayle, Kieron Pollard and Andre Russell in a cluster far along the positive ASI values. These three mostly gather their boundaries from conventionally classical shots: drives. While the average batter uses drives to take doubles and boundaries along the ground, these three employ them in a more attacking manner – finding average control but high returns. The fourth West Indian revolutionary bat, Sunil Narine, is an all-out attacker, by far the worst in terms of control but 12% better at run-scoring than the average batter.At the top, Sanju Samson, de Villiers and Pooran have high control and high run-scoring execution. Slightly lower than them in the same quadrant lie KL Rahul, Hardik Pandya and Pant. Rahul and Pant might have chosen to become judicious with the number of attacking shots they play, but their execution remains top-notch. Kane Williamson and Warner attack more sedately, choosing their attacking shots and sacrificing high run rates for high control. In Warner’s case, although the Attacking Distribution graphs show that he chooses more RA shots than SA shots, his execution data shows that he plays his attacking shots more safely, exercising high control but compromising on run rates.
****
Information about the choice and execution of shots is a gold mine for strategists as they jostle to find perfect match-ups and ploys to control each phase of a T20 innings. This article shows a few such applications of dissecting batter records by execution and intent, to answer questions both academic and tactical. The obvious next application of this data is to simulate counterfactuals – for instance, what happens if Kohli decides to attack the spinners a little more? My next article will seek to answer such “what if” questions via simulations using the granular, personalised data of shot preference and execution we have seen here.