We Need to Talk About Replacement Level

When I started writing at this illustrious site earlier this year, I gave myself one strict directive: don’t write about WAR. That might seem kind of strange given that my writing has always been on the analytical side. It probably seems especially odd given that I just spent two weeks making a formal model of contract value that relies pretty heavily on WAR. Still, though, on day one, I was pretty committed to the idea of leaving WAR alone and focusing on stats that could show a player’s skills more directly.

That might seem like a pretty weird one rule to have- how about something like ‘write well’ or ‘don’t cover the same player too often?’ Eh, I mean, I had those rules too. The proscription against WAR came from a deep place of unease with how reductive and opaque WAR is. It’s pretty sweet to reduce every player to one number- makes analysis a lot easier. When I understand the building blocks, I’m a huge fan of all-in-one stats. My single favorite baseball statistic of all time, wOBA, is an all-in-one encapsulation of offense. What’s so different about WAR?

I could argue with you that it’s the folding of defense and offense into one statistic. Certainly positional adjustments have their own pitfalls, and you would do well to consider carefully how valuable a shortstop or centerfielder might be when switching to first base (looking at you, Ian Desmond). At some level, though, I understand the concept there, even if I might quibble with the exact numbers. No, my real issue with WAR has always been the hazily defined ‘replacement level,’ and how hard it is to explain. If you ask most proponents of WAR, there’s an easy definition ready. Replacement level is the amount of runs produced or saved by readily available players, the kinds of guys you have sitting around at AAA ready to come up to the majors. To quote the Fangraphs glossary: “If this player got injured and their team had to replace them with a freely available minor leaguer or a AAAA player from their bench, how much value would the team be losing?” At its core, that sounds like a pretty clean idea. It’s also a devilishly complex idea, though. How do we find those minor leaguers? How do we know how well they’d perform in everyday roles? How do we compare them to the performances we actually see? Like a 2005 Facebook relationship, it’s complicated.

I’ve laid out the way that people explain replacement level. Do you know how it’s actually defined, though? There are 1,000 WAR available per year. By definition. Why? Well, the easiest way to explain it is probably that a simulated team of minor leaguers wouldn’t score that many runs and would allow a ton, and that a team of minor leaguers playing a regular major league schedule would expect to win a little under 50 games a year. That’s something like 33 wins a team, or 1000 total wins above replacement available. That’s a neat little math trick, and it explains how WAR values are calculated today.

The biggest issue with these calculations is that they’re just way too opaque. Find me this team of minor leaguers. You can’t, buddy, because they don’t exist. The counterfactuals are easy to find, though. Find me some players who got plugged in from the minors and did well. Well, I’d like you to meet my friend Luke Voit. He’s standing right next to my friend Harrison Bader. Both of those guys were minor leaguers who got called up, and both of them did a hell of a lot better than replacement level. I’ve noticed that this line of reasoning comes up a lot in articles that lean on WAR, and it absolutely should. The concept of ‘replacement level’ is very nebulous, and writers like me haven’t done a great job of refuting this obvious line of argument. This is specifically why I stayed away from writing about WAR when I started writing. It wasn’t because I don’t believe in it. It wasn’t because I didn’t find it useful to analyze baseball. It was because the concept of replacement level is fraught and tends to derail discussion. I think a lot of the reason for that is that it’s just easy to wish replacement level players are good if you don’t look at their actual numbers. Well, today I’m going to do that.

Let’s think of some ways to define a replacement level player. The first one I came up with was players receiving extremely marginal amounts of playing time. If you have 50 plate appearances in the majors, it’s a fair bet that you weren’t your team’s first option at your spot. Is this a perfect filter? Nope. It’s honestly pretty solid, though. Teams very clearly view players like this as replacement level- they are literally using them as replacements, and giving all the plate appearances they can to better players. How did this group of marginal players fare in 2018? Well, players who came to the plate 100 times or less in 2018 (excluding pitchers) collectively produced a slash line of .203/.268/.320, good for a 60 wRC+. Let’s not mince words here- they were awful. Their total WAR? -10.9 wins across 7,124 plate appearances, a pace of -1 win per 600 PA. In a very real sense, the set of players who received AAA-scrub level playing time in 2018 were worse than replacement level. That probably sounds wrong to you. It’s pretty easy to remember straight up nobodies coming up from the minors and providing good value for the Cardinals. First of all, the Cardinals are potentially the best team in the majors at extracting value from the minors. Second, that cohort of players has been bad, even on the Cardinals. Here’s a chart of Cardinals replacements this year, along with their plate appearances and WAR:

These dudes are bad! No offense to Patty Smarts, but that’s not an imposing group of hitters. You’ll forgive me if I’m skeptical that they’re meaningfully different from replacement level, and that’s from a team that’s excellent at developing 1-2 win talent.

Hey, I can hear you say. Stop it. Your definition of replacement level is too low! All these guys are just bozos. I want the Yairo Munoz’s of the world, the serviceable major leaguers who consistently rotate between AAA and the show. I wanted to come up with a better way to think about which players are replacement-y, so I came up with a new test. Let’s imagine a normal major league roster. Eight spots are everyday starters. Those aren’t replacement level players- they’re bona fide major leaguers. Another two or so guys are often useful platoon bats or fourth outfielders, real players who have excess value. The last two guys on the bench in this era of 13-man bullpens are shuttle squad members- backup catchers and utility men and the like. We don’t care about those first ten guys on each team’s roster- they’re the players the team wants there. With that in mind, let’s take every non-pitcher who batted in a game this year and cut out the 300 who came to the plate most often(10 for each of 30 teams). The remaining players will form our replacement level population. These guys are better than the true bottom of the barrel, the sample above. They batted .222/.290/.352, a 74.4 wRC+. The best of the bunch was Mets revelation Jeff McNeil, who slashed a scorching .329/.381/.471 en route to 2.7 WAR. The worst? Old friend Magneuris Sierra, who ‘batted’ .190/.222/.211, good for a 19 wRC+ and -1.5 WAR. All told, the 329 players in this group accrued 33,184 plate appearances, about 20% of the total plate appearances by non-pitchers this year. In that playing time, they totaled 1.3 WAR. If you want to turn that into a per-600-PA number, it’s 0.02 WAR per 600 PA. In other words, they were replacement level. The top 300 plate-appearers? They totaled the other 80% of plate appearances and 557 WAR, 2.4 WAR per 600 PA.

This is a Cardinals website. Let’s see the Cardinals who fall into this group:

Still pretty uninspiring. I don’t think most people would consider Tyler O’Neill a replacement level player. He’s a top 100 prospect, someone with real value to the team. Even with him, however, this motley crew produced .7 wins in 617 PA, essentially replacement level. Even a team with as excellent a reputation for spinning straw into gold as the Cardinals isn’t producing much value on the major league fringes.

I repeated the exercise with pitchers just for the heck of it. I looked for pitchers with 30 or less innings pitched, and let me tell you, I had a whale of a time manually going through and excluding all the position players who pitched this year. It needs to stop, or at least be better categorized. In total, the remaining low-use pitchers threw 3600 innings this year. They were putrid- a 5.9 ERA, 5.28 FIP, 8 K/9 and 4.4 BB/9. They were worth -20 WAR, or -.8 WAR per 150 innings.

There’s a bottom line to these studies. Replacement level exists, and it’s easy to see if you take the time to look at the data. Sure, keep dreaming on finding the next Jeff McNeil. Individual players like that are bound to exist, because populations vary. The mean of these marginal players, though, is to have no value above replacement level. It’s easy to think that any old roster slot can produce a 1-to-2-win player. Heck, from time to time I’ve fallen into that pattern of thinking. In aggregate, though, it’s just not true. The 300 guys teams play the most produce pretty much all the value. It’s even more stratified than that, in fact. Batters 201 through 300 were worth 78.3 WAR, 1.4 WAR per 600 PA. All the real value is in the best 200 players in baseball. Maybe your team can identify those guys, and maybe they can’t, but don’t undervalue that first few WAR that a player produces.

Yairo Munoz, the Cardinals’ replacement player phenom? He was worth 0.0 WAR this year. He took up 329 PA to do that. If you could actually grow 2 win players on trees, the 950 PA the Cardinals gave to Yairo and the rest of the scrub squad would have been worth the 3 wins the Cards needed to make the playoffs. Wins are good. Replacement level players are bad. Don’t sleep on how important it is to have real major league players to take your team’s at-bats, or you might end up giving a thousand plate appearances to minor leaguers while being convinced they’re not replacement level.