T20 Player Value: Part IV
This is the fourth post in which I outline my approach to assessing player value. The first explains the overall objective: to measure the expected contribution of each player in runs. The second then details four main adjustments that I make to historic performances to remove any obvious biases in the data. The third walks-through an example and then adds a further three considerations on top of the ones explained previously: weighting, regression to the mean, and ageing
In this final post, I talk about replacement level, wicketkeeper value, and whether or not the progression towards ever higher T20 run rates affects my approach
Replacement Level
Replacement level is a concept widely used in baseball with both Fangraphs and Baseball Prospectus reporting their own Wins Above Replacement statistic. The idea is that a player’s value is most closely related to their production over and above what could be recruited from players not yet signed by other teams. Estimating replacement level is not a strict science but is usually done by looking at the lowest value players who are currently within the relevant league / competition
Considering the replacement-level player is not always totally necessary. Within a draft system or traditional signings, it shouldn’t make any difference to the ultimate decision: a 3-Run player will always be better than a 2-Run player. However, at an auction, where dollar amounts are at play, and getting value for money is a concern, it is a useful concept
Replacement level allows bowlers to be compared directly with batsman… it may be easier to find a replacement batsman who only costs the team 2 runs (vs. average) than it is to find a replacement bowler with the same level of production
It also allows us to neatly quantify the value of the all-rounder. A replacement batsman was worth roughly -1.7 runs across the last 5 iterations of the IPL, whilst a replacement bowler was worth -2.7. Chris Morris, on the other hand, is distinctly above average in both departments: +2.1 as a batsman and +1.9 as a bowler (this might slightly undersell his batting as he has improved considerably within the last couple of years). At the auction, a team bidding for his services is effectively bidding for two players worth a combined 7.6 runs above their respective replacements. Huge
Unfortunately, this is a slight over simplification. For example, it does not capture the value of a batsman who can make small contributions with the ball. Joe Root does not deliver more with the ball than a replacement bowler could. But he doesn’t need to. Few teams employ 5 true bowlers. Players like Root enable a stronger batting line-up without sacrificing too much from the other side of the game. There is no simple solution to this – cricket is a game with diverse playing roles and there is no avoiding that
Wicketkeeping
Source: Sports Analytics Advantage, Cricket Case Study - Selecting players for the 2020 English franchise draft
Currently, my methods do not include any attempt to value a wicketkeepers’ production behind the stumps. This article from Dan Weston suggests some sensible approaches and his results suggest large differences (up to 14 runs) between the best and worst English keepers
Without investigating myself, I will remain sceptical of such a cavernous gap between Bairstow and Foster. The key reservation is that it may be presumptive to assign full credit / blame to the wicketkeeper for byes, catches, and stumpings. Bowling style can also be a major cause of both byes and stumpings. I have previously discovered that bowlers who concede wide also tend to concede more byes (hint: they bowl fast) with Shaun Tait as the Extra-Chaos King
Year
T20 has evolved over time. Specifically, players are scoring faster and hitting more sixes. This has the potential to bias player evaluation in favour of recent batting performances and against recent bowling performances. I have ignored this. I rely on the assumption that, because current players are more or less affected equally, and because the effects are likely to be small, this will make almost no difference to the final results
There may be some difference in how the hitting evolution affects new players vs. more experienced players whose careers go back several years. And historical comparisons between players could also be affected. But overall, I am happy to ignore the passage of time as a significant effect on my analysis
That is it, everything that goes into my one-ultimate-catch-all-player-value-metric. There is a fair bit of code behind it. Once I get a chance to look through it all again, and confirm that my code is doing what I want it to, I will post a list of players ranked according to their contributions vs. the average player and vs. a replacement level player