Occam’s Razor and Jackson Holliday’s Demotion

Baseball is back! Well, I guess it never really left, what with winter leagues and spring training and all. But major league games that count are back, and I’ve been parked on my couch watching as much as I can all week. Naturally, then, it’s time to talk about a guy who hasn’t debuted in the majors yet.

The Orioles are one of the best teams in the American League, and before the season, it seemed like they were going to debut one of the top prospects in the game for a third straight year. Jackson Holliday tore up the minors in 2023, and though he only got a shot of espresso at Triple-A, he was the team’s presumptive starter at second base.

At first glance, nothing Holliday did in Sarasota this spring changed that likely path. He hit .311/.364/.600 in 48 plate appearances while largely playing second base. Then the Orioles did something no one saw coming: they sent him down to the minors.

For years, sending a good prospect down for a few weeks at the start of the season – to “work on their defense” or “learn to hit lefties better” or some other fig leaf – was an economic decision. Players who earned less than a full year’s service time – 172 days out of the roughly 187-day season – reached free agency a year later, with that near-year giving them Super 2 status: an early dip into arbitration, more or less.

The most recent Collective Bargaining Agreement specifically addressed this practice by creating something known as the Prospect Promotion Incentive plan, or PPI. It’s a carrot-and-stick approach. Should a prospect win Rookie of the Year after a) accruing a full year of service time and b) appearing on preseason Top 100 prospect lists (two out of three of MLB Pipeline, ESPN, and Baseball America), that player’s team gets a draft pick just after the first round in the next draft. There are a few other ways to get that pick – winning or placing highly in MVP or Cy Young voting prior to the player qualifying for arbitration – but basically, touted rookies who spend the full year with the major league club and win awards share the wealth with their team.

The stick side of things is that players who finish in first or second place in Rookie of the Year voting accrue a full year of service time, regardless of how many days they spent on a team’s 26-man roster. Adley Rutschman, for example, didn’t get called up until May 21 of 2022, mostly due to injury. That worked out to 134 days of service time. But he finished second in Rookie of the Year voting, and you can see that he’s listed as having two full years of major league experience on his player page.

Teams have responded to this new model by starting their best prospects in the majors. If things click right away, hey, draft pick city! If they don’t, the odds are high that the team will send them down at some point. After all, they aren’t clicking. Jordan Walker debuted on Opening Day last year, but he wasn’t quite ready for prime time, and ended up accruing only 149 days of service time.

This feels like a comfortable decision for most clubs. The Orioles did it with Henderson last year, for example, and got a draft pick out of it. His case was complicated by the fact that he’d already debuted in 2022, so he was likely picking up a year of time either way, but the point is that the PPI changed team incentives, bringing them into closer alignment with prospects’ desire to make the majors earlier (and thus hit free agency earlier down the road).

That’s the background. The Orioles don’t seem to be doing that with Holliday. So what are they doing? Normally, I’d apply Occam’s Razor: the simplest solution is the correct one. But the thing is, neither of the options here are simple, or even logical:

• Baltimore’s front office is manipulating Holliday’s service time and also making a bad financial decision
• Baltimore’s front office doesn’t think Holliday is ready

You probably think there are other options. Maybe this is just some new, heretofore-unseen service time manipulation plan. Maybe there’s actually a secret way that the O’s can have their cake and eat it too. Maybe. I don’t think so, though. I can’t figure out what they’re doing, in fact. Let’s walk through why not.

First, could the Orioles be planning on threading the needle, holding Holliday back long enough that he doesn’t finish in the top two for Rookie of the Year voting while still contributing to the team? Perhaps, but that seems essentially impossible to control. The three betting favorites for AL Rookie of the Year are Wyatt Langford, Evan Carter, and Holliday. That matches our projections. The rest of the group of rookies – Colt Keith, Colton Cowser, Ceddanne Rafaela, Junior Caminero, and a host of others – are a step behind, both in the markets and according to the Depth Charts.

How many games would Holliday need to miss to drop his odds of finishing in the top two substantially? I wrote a simple computer program to assess the problem. I gave each of the top three players a 2.5 WAR/600 projection, as well as two other contenders a 1.8 WAR/600 projection. I asked the computer to simulate 1,000,000 seasons with randomness injected into each of those projections, so that we didn’t get an exact tie every time. Specifically, I bumped each projection using a normal curve with a standard deviation of 1.5 WAR per 600 PA.

If each of the top three players have 600 plate appearances given these inputs, Holliday would win the award 25% of the time and finish in second an additional 18% of the time. The same is true for Langford and Carter. The last two contenders – I called them Keith and Raffaela, but you can pick whoever you’d like, it’s just a simulation – each win 13% of the time and finish second 23% of the time.

That’s with everyone playing a full-ish schedule. Now, let’s start docking Holliday some playing time. At 550 plate appearances, he’s down to a 20% chance of a win and a 20% chance of a second-place finish. At 500, he’s down to 15% odds of winning and 22% odds of getting second. At 450 plate appearances, he’s only 11% likely to win the award, but now 25% likely to get second place. Here are those odds in a grid:

The issue here is that if Holliday is as good as our projections think, holding him down longer doesn’t accomplish what you might expect. It makes him less likely to win the award, sure, but he’s still fairly likely to finish in the top two if he has a good season. If he has a bad season (as determined by my variance term), he’s unlikely to finish in the top two regardless. Playing time does more to affect his chances of winning Rookie of the Year than his chances of finishing in the top two.

We can convert these percentages into dollars if we make a few assumptions. Off the cuff, I approximate that the Orioles would gain something like $20 million in value through lower salaries and increased team control if Holliday hits Super 2 this year instead of getting a full year’s service time, using these rough values for arbitration pay. A draft pick just after the first round is worth roughly $10 million. So let’s re-cast that chart in terms of dollars:

It might look like holding Holliday down is the clear best answer, but there’s an offsetting cost: less playing time for Holliday in the 2024 season. That matters a lot! There’s also a missing benefit in that first line. If Holliday struggled out of the gate, the O’s would probably send him down anyway; they’re in a tight divisional race, after all, so they can’t afford to keep giving a struggling prospect important at-bats. They might end up giving him less than a full season’s playing time not in some plan to manipulate his time, but merely because he isn’t ready. Having him up only to send him down if he looks overmatched would mean burning an option, but if the Orioles are sweating how many options Holliday has a few years from now, other things will have gone very wrong in his development; the option will be the least of their worries.

I threw some very rough values on these outcomes based on my rough understanding of the marginal value of wins, plus a simulation of how frequently Holliday will realize an outcome poor enough to get sent down. To be clear, these are guesses. I don’t have an infallible calculator or anything. I think they’re pretty good guesses, based on win leverage and value over replacement and the aforementioned simulation, but I want to be clear that they’re only approximate. The O’s might have completely different values for some of these things I’m estimating. That’s just how it goes sometimes. Anyway, my guesses:

Reducing the extra expected wins the O’s would get from a few weeks of Holliday to a monetary amount feels ridiculous, I know. Honestly, it kind of is ridiculous. We don’t know how close the Orioles will be to the playoff cut line at the end of the year. They could make it by 10 games, or miss the playoffs even if Holliday were up the whole time. Turning that binary into a single number is strange. You could ignore that column if you wanted, but it’s my attempt to capture the decline in playoff probability from not putting your best team on the field. In fact, maybe you could ramp that number up; missing the playoffs by a few games would be pretty devastating if there were an easy way to add a win here or there.

You’ll notice that the expected values of these options are extremely close. They’re so close, in fact, that tiny changes in my estimates of the cost or benefit of each effect could change the optimal outcome. They’re pretty close to equivalent options, in other words. And if they’re equivalent options, then the Orioles should always pick the one that doesn’t mess with their top prospect’s service time in a transparent way.

Why? There are a ton of reasons, but the most obvious one is that dealing with prospects isn’t a single-shot game. Holliday is going to be on the Orioles next year, and for quite a few years after that. Messing with his compensation in a ham-handed way isn’t free; he’ll remember it in his future dealings with the team. What’s more, this doesn’t occur in a vacuum. Everyone else on the Orioles will see what the team does with their top prospect. If you’re a fellow high performer, seeing the O’s play service games with your compatriot might make you think twice about signing a contract extension.

That’s a much bigger deal than the tiny savings you might or might not achieve by gaming Holliday’s service time to get him into Super 2 status rather than accruing a whole year’s worth of time. Contract extensions are often mutually beneficial for team and player. They give the team meaningful economic savings while providing the player financial certainty. Both sides win on that deal – but you can’t make the deal without trust, and that can get harder to come by in a hurry if you have a high-profile case of manipulating the rules of service time to limit player compensation.

Maybe this sounds like a minor effect to you, but given that the benefits of keeping Holliday down an extra few weeks are tiny to begin with, even the chance of affecting a single extension negotiation makes keeping Holliday down a bad expected value play in my book. The O’s have a lot of guys they’d love to extend in the coming years; signing any one of those players to a team-friendly deal would give them vastly more financial savings than anything they do with Holliday this year.

To put it a little more forcefully: I believe that keeping Holliday down through the Super 2 deadline would be not just a bad faith decision but a bad economic one. The savings just aren’t high enough to offset the potential future costs. The Orioles obviously will never tell us publicly if they agree with my assessment, but I feel pretty good about my math here. The new CBA didn’t completely change the calculus of service time manipulation, but it certainly tilted the scales towards calling players up as soon as they’re ready, and the O’s are in the perfect place for that tilt to matter. They’re locked in a tight race for the playoffs, their top prospect is better than the other options they have at his position, and they also have a ton of players they’re hoping to sign to extensions sooner rather than later.

The other option in this case isn’t particularly likely either, but it’s at least possible that the O’s simply don’t think Holliday is ready to play in the major leagues every day. The case here is plate discipline. Sure, he hit well in spring training, but he struck out 30% of the time with swing-and-miss numbers that suggest it wasn’t a complete fluke. He’s 20 years old and has batted only 255 times above A-ball. The presence of Henderson on the big league roster has also necessitated a move off shortstop, where Holliday had seen the bulk of his minor league defensive innings, to second base (975.1 innings at short vs. 215.1 at second entering this year). I think that Holliday is probably ready, but I’m not certain about that opinion. I’m more certain that demoting him for service time reasons is a bad call; the numbers simply don’t lie.

I’m not here to tell you which of the two scenarios is true. I don’t think either option makes that much sense. The math is no good on the service time manipulation front, and Holliday also looks ready to me. It’s clearly a very small sample, but he’s tearing up Triple-A to start the year and doesn’t seem to have a strikeout problem. Meanwhile, Tony Kemp is on the major league roster but the O’s don’t seem interested in using him. Ramón Urías looks like a weak link in the starting lineup, to boot.

Whatever Baltimore is doing, it’s strange in my eyes. If they’re trying to squeeze out a few extra dollars, it’s too much risk for too little return even if you think there are no ethical implications. If they think their super-prospect isn’t ready, I think they’re being too sure in their own assessment. Whether he needs more time or not is, in the end, unknowable. We and the team can only make estimates. Even if they think there’s only a 20% chance that he’s ready, the relative costs and benefits of finding out at the major league level suggest to me that he should be up already. We’ll never know exactly what the Orioles front office is thinking, but I’m perplexed by their decision here. The simplest explanation is often correct – but in this case, there’s no simple explanation at all.