by EJ Fagan
February 19, 2024
***
NOTE: The following comes from EJ Fagan's substack page and is shared with permission.
Please check out EJ's substack page for more great articles.
***
Baseball Prospectus released its PECOTA projections for 2024. They look fantastic for the Yankees:
These rankings include the Corbin Burnes trade. They are really down on the 99-win Rays and 101-win Orioles, which is basically in line with my less sophisticated analysis.
PECOTA projects the Yankees to win 94 wins, 4th in baseball behind the Astros (95), Braves (101) and Dodgers (102). There’s a pretty big gap between the Yankees and the 5th place Twins (89). The Yankees are basically tied with the Twins, Braves and Dodgers for fewest runs allowed in the league, and 4th in projected runs scored behind the Astros, Braves and Dodgers.
Their projection has a teeny tiny bit more variance than their AL East competition, but nothing to be too concerned about:
That’s all great news! PECOTA has long been the most accurate projection system in baseball. The AL East is projected to be very competitive again, but it doesn’t buy the Orioles right now.
What do projections actually tell us?
I hear what you’re thinking. The game isn’t played in PECOTA. It’s played on the field. I think it’s worth flexing my statistician muscles and talk a little bit about how projections work.
A typical back-of-the-envelope projection system, like ZiPS, averages a player’s recent seasons and applies a predicted age curve based on all players to predict their future performance. Older players lose a step. Younger players get better. That’s basically it.
PECOTA does a lot more. It takes the player’s past performance. Instead of using all players to predict how a player will age, it creates a custom prediction for each player based on how similar players performed at the same age. How does it come up with those comparable players? They don’t release the exact secret recipe, but it involves everything from their performance, strike zone control, playing time, physical attributes like height and weight, and position. I believe that human beings are involved in setting playing time expectations.
One consequence of PECOTA’s system is that players are much more heavily regressed to the mean. It tends to bet that most players will strongly move toward the average MLB players. That’s why some of the win totals projected by PECOTA seem a little too close to 82 wins. It expects players who had breakout 2023 seasons like Kyle Bradish or huge disappointments like Carlos Rodon to move strongly back toward the average player.
Another consequence is that PECOTA tends to miss on outliers. There are a lot of players in major league history who are similar to Alex Verdugo, Anthony Volpe or Carlos Rodon. There are fewer who are similar to Aaron Judge or Juan Soto. Take a look at Judge’s top comparables on Baseball Reference. Basically none of them scream Aaron Judge back at me.
Also, Soto’s comparables are fun, and might help PECOTA to predict him a bit, but he’s got very different physical skills from pretty much all of these guys:
Finally, we should emphasize that PECOTA doesn’t know some important things. It doesn’t know who is coming back from injury or what injury. It doesn’t know how well a team will manage it’s bullpen. It doesn’t know who is playing in a contract year. It doesn’t know anything about MLB rules changes. There’s tons of room for error.
Should We Doubt the Projections?
How do we feel about the Yankees and the rest of the AL East on these error metrics?
The Yankees have a few pretty unique players in Judge, Soto and arguably Gerrit Cole. PECOTA is probably low on all three. They also have pretty big question marks in Cortes and Rodon, but I’m not sure what their individual projections look like. I don’t see any big differenced there.
The Orioles have lots of young players who could be outliers to be better than their peers, but PECOTA has lots of young players with great rookie seasons to compare against. I’m not sure that I can see a genuinely weird player on the roster, or anyone coming off a huge injury that is now healthy other than John Means.
PECOTA doesn’t know that the Rays are geniuses at getting more out of their players than anyone else. Yandy Diaz might regress, but the Rays might just find some other crazy breakout player to play shortstop or something. I think that their magic has to wear off eventually, but it hasn’t yet.
The Jays are a pretty normal team in the grand scheme of things. PECOTA thinks that they are about as good as last year. The Red Sox are just bad again.
I think PECOTA did a pretty good job here. It can’t divine the future, but it’s pretty good at making an educated guess. There will be a breakout team or two that PECOTA is wrong on, and a few teams that does a 2023 Yankees and disappoints. But right now, I’d bet on PECOTA and picking the Yankees to win the AL East.
I disagree with PECOTA saying that Giancarlo Stanton is a "similar player" to Juan Soto. When healthy and when he is hot, Giancarlo is one of the best HOME RUN hitters in baseball. But even then, he is a high homer, high strikeout hitter, while Soto is a high contact hitter, with a much better eye for the strike zone. He will walk a lot more and strike out a lot less than Stanton ever will.
On the other hand, I would put Stanton as a "similar player' to Aaron Judge. In terms of size, size of their strike zone (because of their size), their tendency to strike out a lot, and their tendency (when healthy) to hit an enormou…
I don't believe the Yankees are 15 games better than the Orioles, period. Can New York win the Division? Sure. But everything will have to break right for them.
The Judge and Soto comps are striking. I remember Bill James saying that where there isn't a single comp above 900, you're dealing with a unique talent. Soto is a great player with 6 excellent players with 900 or above comps, including two HoF'ers and three other guys who I think should and will get it. Judge has 0 comps at 900 or better, and none of them were as good as he is. Judge is a unicorn.
First of all, wouldn't it be nice if all of US....in the "non-sports" world can have a PECOTA evaluation that will predict OUR success (or lack of success) in "real life" this year? That would be even more significant than using it to predict the success of sports professionals. I would personally want to know my PECOTA evaluation in how my own success (or lack of success) will be in my own life this year.
I LOVE that the Yankees are predicted to be in First Place this year based on the PECOTA calculations. But I envision the Baltimore Orioles as the favorites in the division, and the Yankees, with the additions they made, knocking the Orioles out of Firs…
Bill Pecota should have stayed being a below average ex utility player instead of graduating to a below average statistician. How exactly do the Yanks win 12 or 13 more games in 2024 and the Orioles lose 14 more games then they did in 2023, despite getting a legit ace? I Just don't see it.
still waiting for another starter.
don't think they get to 94 wins without another arm
do think that Rodon will return to the land of the living this season and do well