By Sam Waters
Measurements of player value like WAR lean heavily on the concept of the replacement level player. Subtracting replacement level production from a given player’s production allows us to estimate that player’s marginal contribution to his team on top of what a minimum-cost alternative would provide.
When outside observers evaluate players using WAR, they are mostly concerned with describing the true value of each player in a vacuum, so they focus on neutralizing the effect of each player’s context. To this end, WAR states a player’s hypothetical value under average conditions, which include an average run-scoring environment and average replacement players. For the purposes of an objective, descriptive analysis, this is appropriate.
But when a front office assesses the value of a player for the purposes of an actual decision about roster construction, they must consider the context in which they find themselves. If certain teams systematically provide minimum-cost players with higher levels of true skill as replacements for their big league rosters than others, those teams would have a higher replacement level. This, in turn, would alter all of their player value calculations and subsequent decision-making. For this reason, it is important to think about whether teams might have different replacement levels.
I can think of three avenues for systematically producing better replacement players than other teams, despite expending the same resources in terms of salary or draft pick on each replacement player:
1. Acquisition through superior talent evaluation
2. Development through superior training once acquired
3. Increased selectivity for replacements when fewer replacements are needed
The first two are pretty self-explanatory, so let’s focus on the third option. The wage floor in the MLB might disguise differences between replacement level players in terms of salary, but in reality there are varying degrees of skill within this pool. Teams with more above-replacement-level players have less roster slots to fill with replacements, which allows them to be more selective and use only their better replacement options.
Now, let’s say we have 30 teams with equal abilities in the acquisition and development departments and identical pools of replacement options of varying skill. One of these teams plays in a big market with a high payroll, which they have used to acquire many above-replacement players, leaving them with just two open slots on the big league roster for a replacement level player. Meanwhile, a small market team with a low payroll has not acquired many above-replacement players, and has fifteen open roster slots to dedicate to replacement level players. The big market team gets to pick its two best minimum-cost options as the replacement level players in those last two slots, while the small market team has to pick its best fifteen options. The average replacement of the big market team is likely to be much better than the average replacement of the small market team in this scenario. In general, if a team’s roster construction requires more replacements and we measure replacement level by average expected replacement production, we would expect its replacement level to be lower when we hold all else equal.
That is one reason why a star player might be worth more to a team like the Yankees or the Red Sox. Since they can be choosier about their replacement players, their overall replacement level should be higher. This would leave mediocre players with a smaller share and elite players with a larger share of the league’s available marginal production from the high-payroll team’s vantage point than from the perspective of a low-payroll team.
In general, differences in the expected production of replacement players among teams should create discrepancies between each team’s valuations across the entire player pool. Such differences need to be taken into account when teams make personnel decisions or when outside observers analyze these decisions, rather than simply basing analysis on each player’s projected value in a vacuum.
Great point! Don’t know if the valuation changes would be very large, but it’s the right way to think about the problem.