Flem274*
123/5
I'm playing around with numbers for an MVP rankings project. Despite using bowlers here I was initially inspired by opening batsmen, since finding good openers tends to be harder than finding good middle order players. I think the true value of a player is found both in their historical scarcity and how much better they are than their contemporary rivals. To use an exaggeration to demonstrate, if you have a spinner who averages 20 and the expected average of all spinners during his career is 33, that's an OP advantage that should be recognised in any MVP type system.
So I'm very interested in actual performance vs expected performance. For example, a right arm pace bowler who debuted on January 1 2011 and retired September 13 2021 is compared against the aggregate statistics of all right arm pace bowlers bowlers during that time. He may average 28 and be expected to average 30, so he gets a score of 2.
What is immediately noticeable is how hard this breaks at smaller sample sizes (using kiwis is great for breaking data because they have several bowlers with low averages and low games played), but I thought I'd post the numbers anyway. I think it works okay for larger sample sizes.
Kyle Jamieson 13.27
Jack Cowie 11.95
Shane Bond 11.86
Glenn McGrath 9.66
Malcolm Marshall 8.83
Pat Cummins 8.69
Richard Hadlee 7.36
James Anderson 5.37
Bruce Taylor 4.35
Neil Wagner 3.51
Jasprit Bumrah 3.39
Tim Southee 3.09
Matthew Hoggard 2.56
Trent Boult 2.29
Chris Cairns 1.37
Pat Cummins is a good example of game distribution messing with the raw data too. If you exclude his debut test from the career span his score drops to 5.64. Bumrah's score is so low compared to his raw bowling average because his expected average was 26.18. Bumrah is a lot closer to the average bowler of his short career than you initially think since his entire career is contained within the same period I discussed in my 35 averaging batsmen thread.
Excluding noise, I find a top 3 of McGrath, Marshall and Hadlee quite reasonable and down the lower end Anderson, Taylor, Wagner, Bumrah, Southee, Hoggard, Boult, Cairns also quite defendable for a raw dataset.
With some more depth than this blunt force raw data I think this could be quite fun, especially when factoring in scarcity. I think it will throw up a mixture of confirming what we already know and throwing in some curveballs.
I think the main argument against will be I am punishing players for being good in eras where there are lots of good players. Firstly I don't think that will be true. The 80s were a golden age for quicks and both Marshall and Hadlee have an excellent raw difference between actual v expected that I don't think will be beaten by many if any. Secondly, being very good doesn't always make you the most valuable player to have, and I'm thinking of including seperate national rankings anyway where Bumrah will absolutely dominate compared to all Indian right arm pace bowlers - national value doesn't always correlate to world value.
So I'm very interested in actual performance vs expected performance. For example, a right arm pace bowler who debuted on January 1 2011 and retired September 13 2021 is compared against the aggregate statistics of all right arm pace bowlers bowlers during that time. He may average 28 and be expected to average 30, so he gets a score of 2.
What is immediately noticeable is how hard this breaks at smaller sample sizes (using kiwis is great for breaking data because they have several bowlers with low averages and low games played), but I thought I'd post the numbers anyway. I think it works okay for larger sample sizes.
Kyle Jamieson 13.27
Jack Cowie 11.95
Shane Bond 11.86
Glenn McGrath 9.66
Malcolm Marshall 8.83
Pat Cummins 8.69
Richard Hadlee 7.36
James Anderson 5.37
Bruce Taylor 4.35
Neil Wagner 3.51
Jasprit Bumrah 3.39
Tim Southee 3.09
Matthew Hoggard 2.56
Trent Boult 2.29
Chris Cairns 1.37
Pat Cummins is a good example of game distribution messing with the raw data too. If you exclude his debut test from the career span his score drops to 5.64. Bumrah's score is so low compared to his raw bowling average because his expected average was 26.18. Bumrah is a lot closer to the average bowler of his short career than you initially think since his entire career is contained within the same period I discussed in my 35 averaging batsmen thread.
Excluding noise, I find a top 3 of McGrath, Marshall and Hadlee quite reasonable and down the lower end Anderson, Taylor, Wagner, Bumrah, Southee, Hoggard, Boult, Cairns also quite defendable for a raw dataset.
With some more depth than this blunt force raw data I think this could be quite fun, especially when factoring in scarcity. I think it will throw up a mixture of confirming what we already know and throwing in some curveballs.
I think the main argument against will be I am punishing players for being good in eras where there are lots of good players. Firstly I don't think that will be true. The 80s were a golden age for quicks and both Marshall and Hadlee have an excellent raw difference between actual v expected that I don't think will be beaten by many if any. Secondly, being very good doesn't always make you the most valuable player to have, and I'm thinking of including seperate national rankings anyway where Bumrah will absolutely dominate compared to all Indian right arm pace bowlers - national value doesn't always correlate to world value.
