Most "impactful" IPL batters
"Hit out or get out", they would scream in the late 80s, when Ravi Shastri was batting. Maybe it's time to bring back that slogan for his protege Virat Kohli
Long time! I’ve pretty much given up on this newsletter, primarily because I hardly watch cricket nowadays. And pontificating without watching is not the best way to go.
Anyway, there was a question that was asked on a WhatsApp group that I’m on, and to answer that I constructed a metric and downloaded some data, so I thought I should share it with everyone.
The trigger was the age old question - does Virat Kohli merit a place in the Indian T20 squad? On the one hand, he is scoring loads of runs, including wearing the orange (or is that purple?) cap this season. On the other hand, he is known to score slowly, leaving runs on the table, especially when the team is batting first.
So I came up with a fairly crude “metric” to calculate a batter’s impact on the game. For each innings, calculate the batter’s strike rate (runs per ball) and the strike rate for the rest of the game (including both teams’ batting). Now, calculate how many “excess runs” the batter has hit.
For example, if our batter scored 20 runs in 15 balls, and the overall game produced 360 runs in 240 balls, then our batter’s strike rate is 1.333 while the “remainder strike rate” is 340 / 225 = 1.51111. If our batter had scored at this “remainder strike rate”, he would’ve hit 22.667 runs in his 15 balls. And thus his “impact” is -2.667 (20 - 22.667).
Notice that this is a “volume metric” (or what I sometimes call an “area under the curve metric”), which rewards hitting out or getting out (a phrase Indians of my generation and older are familiar with thanks to Ravi Shastri). If your strike rate is lower than that of the rest of the game, it’s far better if you get out early - that way you leave more balls for the “rest” to hit.
Anyway, I computed this metric for each IPL game in the last three years (including last night’s LSG-MI game), and then for each batter simply summed up the impact over all the games that they have played. You can think of this impact as “runs added over average”. This makes it somewhat similar to baseball’s WAR (wins above replacement - I’d written about this last year).
I looked at the top 20 run scorers in the IPL over the last three seasons (2022 onwards) and looked at their overall total impact.
The way to read this is - across all the innings that he has played over the last three seasons Kohli has scored 232 fewer runs than he would have had he scored at the same strike rate as the rest of his teammates and opponents. Considering he has batted 38 times in this time period, Kohli’s slow batting means RCB have scored 6 runs fewer per innings in this period.
Kohli is not the only culprit here - you can see that Ishan Kishan, Shikhar Dhawan, Rohit Sharma and KL Rahul also have large negative contributions on their team’s scores. Another surprising “negative impact” player is Jos Buttler.
On the other side, we see that Shivam Dube and Nicholas Pooran are by far the most impactful batters, followed by Suryakumar Yadav.
Now let’s dig a little deeper into this - we know that (at least in ODIs) Kohli is a master chaser. Maybe his impact is different compared to whether his team batted first or is chasing?
Of the 232 runs that Kohli has cost RCB over the last three seasons, 202 have come when batting first, and the balance 30 while chasing. Even while chasing, Kohli underperforms the match average, but not by that much.
Elsewhere
This analysis took longer than I thought it will because all along the way I kept thinking “how will I teach my AI to reason and analyse like this?”. For the uninitiated, I quit my job towards the end of last year and have started building Babbage Insight. It’s fundamentally an AI model that can generate insights based on structured data, without being prompted.
At some weird approximation, we are building an AI model that can analyse data like I do.
So this means that nowadays whenever I analyse data, I do it deliberately, observing myself and trying to figure out “insights” that I can then train my model to work like. I’ve taken on a fairly large mandate, and it’s going to be a tough build, but it’s exciting for sure. Anyway, I’m documenting my journey of starting up here.