The Official Hopefully-Not-Too-Erroneous 2023 ZiPS Projections

After all the rumors and money and projections and Carlos Correa signing announcements (and un-announcements), here we are, back at 0-0, with every team having at least some technical level of hope for the 2023 season. Until today’s games start the process of turning projections into history, the season is a blank piece of paper or canvas, a fresh layer of snow without a hint of gray, a home improvement project before the moment you remember you barely know how to assemble an IKEA end table. Now before I get too wistful and start sounding like Paul Harvey or weird AI Vin Scully, let’s get down to business.

Since I am the owner, caretaker, and occasional messer-upper of the ZiPS projection system, that’s the system I used to run the 2023 season a million times. The methodology I’m using here isn’t identical to the one we use in our Projected Standings, meaning there will naturally be some important differences in the results. So how does ZiPS calculate the season? Stored within ZiPS are the first through 99th-percentile projections for each player. I start by making a generalized depth chart, using our Depth Charts as a jumping off point. Since these are my curated projections, I make changes based on my personal feelings about who will receive playing time as filtered through arbitrary whimsy my logic and reasoning. ZiPS then generates a million versions of each team in Monte Carlo fashion (the computational algorithms, that is — no one is dressing up in a tuxedo and playing baccarat like James Bond).

After that is done, ZiPS applies another set of algorithms with a generalized distribution of injury risk that changes the baseline PAs/IPs for each player. Of note is that higher-percentile projections already have more playing time than lower-percentile projections before this step. ZiPS then automatically (and proportionally) “fills in” playing time from the next players on the list to get to a full slate of plate appearances and innings.

The result is a million different rosters for each team and an associated winning percentage for each of those million teams. After applying the new strength of schedule calculations based on the other 29 teams, I end up with the standings for each of the million seasons. This is actually much less complex than it sounds.

The goal of ZiPS is to be less mind-blowingly awful than any other way of predicting the future. The future is tantalizingly close but beyond our ken, and if anyone figures out how to deflect astrophysicist Arthur Eddington’s arrow of time, it’s probably not going to be in service of baseball projections. So we project probabilities, not certainties.

Over the last decade, ZiPS has averaged 19.9 correct teams when looking at Vegas preseason over/under lines. I’m always tinkering with methodology, but most of the low-hanging fruit in predicting how teams will perform has already been harvested. With one major exception, most of ZiPS’ problems now are about accuracy rather than bias. ZiPS’ year-to-year misses for teams are uncorrelated, with an r-squared of one year’s miss to the next of 0.000575. Now, correlations with fewer than 20 points aren’t ideal, but the individual franchise with the highest year-to-year r-squared is the Mariners at 0.03, which isn’t terribly meaningful. If you think that certain franchises have a history of predictive over- or underperformance, you thought wrong, and I’d bet it’s the same for the other notable projection systems.

Now, I did say there was one major exception, and that’s in-season roster changes. As you might suspect, ZiPS systematically underestimates teams that add value during the season and overrates teams that subtract value. With that in mind, it’s not surprising that when you total it up, two of the teams that have made the most in-season additions over the last 17 seasons, the Astros and Yankees, underperformed most often in ZiPS. I have not figured out a way to make an adjustment for future, possible moves that does not result in a horrible, unwieldy mess.

With most of the free agents capable of moving the needle signed before pitchers and catchers reported and a subsequent dearth of trades this spring, there hasn’t been drastic movement from the pre-spring training projections. The differences are largely due to injuries and the cumulative affect of changes in playing time assumptions. As such, I’m including both leagues in one post this year. Let’s start with the AL East.

The Yankees were the most volatile team in the standings when I ran these projections back in February and the uncertainty of the team’s rotation has just magnified that result. We shouldn’t necessarily panic about Luis Severino and Carlos Rodón at this stage, but their injuries are adding yet more mysterious fog of war to the mix. Combine a rotation with a lot of variance with the fact that there’s a lot of value tied up in a single player, Aaron Judge, and the Yankees are the division’s X factor. The Blue Jays are again the most stable team, but they win 100 games less often than either the Yankees or the Rays. I think there’s a risk in underrating the Red Sox; ZiPS at least thinks they’re the weakest team, but we’re not talking the A’s or Nationals here. They project as needing the most things to go right to win the division, but it wouldn’t be completely preposterous if it happened. In total, Boston has the best divisional odds of any team projected for last place.

Various playing time assumptions have given the White Sox a couple of games back in the projections, but ZiPS still thinks that their 2023 season will be viewed as a disappointment. It basically comes down to how ZiPS handles depth, and this particular set of projections has the Sox with about as much depth as Steven Seagal’s annotations for Finnegans Wake. There’s still a great deal of upside in the scenarios where they’re blessed with excellent health, but counting on that isn’t the best idea. Cleveland and Minnesota are safer choices, but both organizations have weaknesses. The balanced schedule pushes everyone’s playoff probabilities down a bit and of the three Wild Cards, ZiPS projects that an average of 0.3 Wild Card teams will be from the AL Central.

The hardest team to gauge in the West remains the Los Angeles Angels. The 84-win projection here basically assumes the Angels can build a 69-win team around Mike Trout and Shohei Ohtani. That doesn’t sound like a Herculean task — the Pirates and Reds appear likely to accomplish that at less than half the price. But it’s a task the Angels seem to fail to complete annually. There are reasons to like the team — they’ve added some second and third-tier heft to the depth chart — but I’m at the point where I’ll believe it when it happens. The rest of the division hasn’t changed that drastically: the Astros are the favorite, the Mariners ought to be good, the Rangers are interesting but not quite there yet, and the A’s have managed to keep their phone service active.

ZiPS already thought Philadelphia’s depth was abysmal and losing Rhys Hoskins didn’t help things at all. I think I actually prefer the Pale Hose’s spare outfielders to the Phillies’. In the Atlanta vs. New York projections, the Edwin Díaz injury was enough to break the deadlock, but it would be a mistake to downgrade the Mets too much. ZiPS remains unimpressed with Miami’s offense and I doubt it’s a surprise to anyone that the Nats project to only really contend for the top pick in the draft.

Very little has changed in these projections from six weeks ago. ZiPS almost always likes the Cardinals, which is rarely a bad idea as they’re regularly one of the safest organizations in baseball. Milwaukee’s rotational health will be the biggest variable when it comes to whether the Brewers fight it out with the Cards or finish a respectable-but-distant second place. The Cubs are just on the very edge of respectability, but there are still significant weaknesses in the lineup and pitching staff that keep them as dark horse Wild Card contenders for the moment. I think there are a couple more teams in the division, but if their owners don’t give much thought to how often they win in 2023, I’m not sure why you should.

The Dodgers’ projection relative to the Padres and Giants is likely the most controversial thing here, but if you look at LA’s roster, there are simply a lot of scenarios where things can go really wrong. It’s not as if ZiPS has a habit of under-projecting the Dodgers. The Padres lineup looks strong, but again, it’s a team that has significant downside scenarios; the rotation doesn’t go as deep as it has the last couple years. The Giants have a very low ceiling, but it’s also a well-assembled roster, at least if you believe my projections. The Diamondbacks could be a very good team in a few years and there’s enough young pitching talent in the high minors that ZiPS actually projects them to win 95 games more often than the Giants. And that’s despite the Giants, at 88 wins, being considerable closer to 95. The Rockies remain the worst team in baseball that doesn’t realize it isn’t a contender.

An important thing to remember about projections is that they’re not projecting the total number of wins that will win a division or a Wild Card spot, but the over/under projection for each individual team. But we know that not every team is going to hit their exact 50th percentile projection. The average expectation is that three teams fail to hit their 10th-percentile projection, three finish between their 10th- and 20th-percentile projections, and so on.

For example, the best over/under in the AL East may be 89 wins, but that doesn’t mean that ZiPS believes that 89 wins will actually be enough to win the division, because you expect half the teams to beat their projections, and some of those by a large amount. So here’s a table with the probabilities that X number of wins will win each possible playoff spot:

To actually win the AL East, the target isn’t 90 wins, but 95 wins. And that’s not even the toughest division; ZiPS projects that 98 wins fails to win the NL East a smidgen over half the time.