One thought: My guess is that in the "runs conceded per wicket vs expectation" metric, you're effectively just subtracting the bowler's average from the total average of all the other bowlers who played in the match. Empirically, I've seen taking a ratio proves to be a much better indicator of how a bowler has performed. Ratios prove to be better predictors of match outcomes than differentials.
For e.g. — a bowler who averages 15 when the others average 30 has generally done just as well as a bowler who averages 20 when the others average 40. But in your calculations — the latter bowler will be privileged by 5 runs lower than the former. It would be more accurate to say that both bowlers have performed twice as well as the rest.
I definitely take your point on ratios, and I know that’s e.g. how Jarrod Kimber presents it in his work. Working from a purely descriptive perspective, I find framing it in terms of runs to be a presentation of the data that makes it a bit easier to get your head around — but it’s the first time I’ve worked with bowling data this way, and will have a think about doing it differently going forward for the reasons you outline.
One thought: My guess is that in the "runs conceded per wicket vs expectation" metric, you're effectively just subtracting the bowler's average from the total average of all the other bowlers who played in the match. Empirically, I've seen taking a ratio proves to be a much better indicator of how a bowler has performed. Ratios prove to be better predictors of match outcomes than differentials.
For e.g. — a bowler who averages 15 when the others average 30 has generally done just as well as a bowler who averages 20 when the others average 40. But in your calculations — the latter bowler will be privileged by 5 runs lower than the former. It would be more accurate to say that both bowlers have performed twice as well as the rest.
Thanks for taking the time to read! And yes that’s what I’ve done in effect (my code is online here if you’re into that sort of thing: https://github.com/plottheball/ptb-newsletter/blob/main/newsletter/2025/ptb-72/ptb-72-bumrah.R).
I definitely take your point on ratios, and I know that’s e.g. how Jarrod Kimber presents it in his work. Working from a purely descriptive perspective, I find framing it in terms of runs to be a presentation of the data that makes it a bit easier to get your head around — but it’s the first time I’ve worked with bowling data this way, and will have a think about doing it differently going forward for the reasons you outline.
Thanks again and all the best!