For Saber-Terms, I’ve been working my way towards discussing Wins Above Replacement by setting the groundwork for a lot of that discussion. Today, I’d like to talk about one of the more controversial (even though it really shouldn’t be) concepts that those new to sabermetrics find hard to accept: replacement level.
What is replacement level?
A good companion to this Saber-Primer is Colin Wyers’ piece from 2009 on replacement level. I’ll be referring to it in just a bit.
In discussions I’ve had in the past, the mention of “replacement level” or “replacement player” at times will elicit rolled eyes and a scoff, likely because such a player is imaginary, just like all the other stats the saber-nerds use from their collective mothers’ basements. But really, the concept of replacement level is actually pretty elementary to how baseball teams would allocate resources.
Imagine you run an expansion franchise and have the task of filling out your 25-man roster from available free agents. You have meager pickings from your minor league affiliates, so you’re forced to sign much of your roster from available players. With your league-average budget, you might start with some well-known players, maybe a few league average guys for your staff or position players. You would probably pay those guys according to whatever market rate the going rate is at. However, as the roster spots dwindle, you’re left with weaker and weaker players to fill those slots. Eventually you reach a certain point where the talent level of the players you may be looking at does not exceed the talent level of some of the 30-year olds playing in your Triple-A farm system. How much would expect to pay those guys? Well, since they’re about as good as your Triple-A veterans, you would probably expect to pay them the minimum, around $400K.
In a nutshell, you just hit replacement level. That is, replacement level can be seen as the talent level at which you would pay the league minimum in the free agent market. Keep in mind that, because of the pay scale currently employed in MLB, not all players who play for the league minimum are replacement level; in fact, because salaries are deflated at the minimum for the first three years, many players who are much better than replacement level get paid the minimum every year. The true replacement level player is the player with the talent level for which you would pay the league minimum if he were a free agent.
Colin puts it in another way that may help understanding of the concept (bold is my own emphasis).
et’s try a little thought experiment. Let’s say that tomorrow morning Bud Selig declares that he’s exercising the “best intrests of baseball” clause to have Albert Pujols’ contract nullified, making him a free agent immediately. How many teams offer him a job? And how much money do they offer him?
Yeah, I know – that one’s too easy. How about Hanley Ramirez?
Again, too easy. Everybody is going to at least offer Hanley some kind of deal. And so what if you already have a shortstop? Hanley’s bat will play anywhere on the diamond, won’t it? You’ll make room for Hanley Ramirez.
What about, say, Randy Winn? Or Carlos Pena? Most teams are going to be interested of course, but some teams might reasonably decide that they can’t make room for these guys, and the intrest won’t be nearly as high.
What about guys like Jason Kendall, or Ross Gload? Sure, some teams will call, but possibly not even half of them. And the money won’t be near what those other guys would be offered.
Okay, now that we’re all on the same page, I’ll proffer yet another definition of replacement level:
Replacement level is the level of talent at which teams stop competing for your services, and you end up competing for the last handful of open roster spots.
This gets to the commonly-quoted definition of “freely available talent.” There are replacement-level players everywhere, in far more abundance than there are roster spots on MLB teams. Thus, players at this talent level are considered replaceable; there are many of them hanging around in the minor leagues, waiting for their cup of coffee. Colin’s chart in the linked article gives you an idea of why there are so many of these players available.
Can you give me an example?
Replacement level has been defined primarily through examples. Sean “Rally” Smith in late 2008 put up this article on THT that used positional adjustments and the best players in the minor leagues to provide an example. Dave Cameron over at FanGraphs has done a series pointing out the players closest to replacement level at each position.
The seminal example, as Tom Tango has always put it, is Willie Bloomquist. Take a look at Willie’s FanGraphs page. He’s currently projected at around a .300 wOBA. Over 600 PA, that’s around 15 runs below average. Take a look at the UZR section. If you mentally add the positional adjustments to his various UZR/150 rates at each position (I know, they’re all small samples, just play along), you’ll get a player who rates at better in the infield, worse in the outfield, overall not far from average. Recall that the adjustment for replacement level is 20 runs per 600 PA, the assumption being that your typical replacement level player is some 20 runs worse than the average player. Sure enough, Willie’s pretty close, at 15 runs below average expected and around average defense. And given his 1.0 WAR in seven seasons in the “FanGraphs era” of 2002 to 2009 (1857 total PA), I’d say he’s about as close as it gets.
Why evaluate versus replacement level?
This is an excellent question. Why even bother with replacement level? The short answer is that it’s a matter of taste; if you’d like, you could choose another evaluation baseline. But there are definite merits to choosing replacement level over other baselines that have been considered. For example:
Vs. Average: You could consider measuring players versus the average, as Pete Palmer did years ago. But the problem with evaluating versus the average is that we then cannot identify the value of an average player. It is clear when you see an example such as the one given by former Baseball Prospectus author Keith Woolner here. I could look like an average player even though I never played a single game, while even someone like Emilio Bonifacio provided more value than me in 2009.
Vs. Zero baseline: You could also measure players versus a zero baseline. But the problem with that is that no team has that alternative. The Marlins did not have a choice to man an empty spot at third base last year, so it would be silly to measure Bonifacio’s production against that baseline. The lowest possible acceptable baseline would be a scrub playing third base, and that’s your replacement level baseline.
The concept of replacement level isn’t some mythical thing that’s been made up by sabermetricians to qualify some obscure player’s production. It’s a real concept that teams undoubtedly have to deal with, even if they don’t call it “replacement level.” Of course, that’s what we call it here, and that’s why I took the time to help define it.