Enjoyable with WAR Math

How a lot WAR does FanGraphs mission Ronald Acuña Jr. for in 2026? It’s a extremely simple query. It must be particularly simple now that each one of our projections are out. However because it seems, it’s much less clear lower than it sounds at first, and clarifying it has two advantages. First, it’ll enable you to higher perceive our projections. Second, it’s enjoyable to play with math. So buckle up: We’re doing arithmetic.

First, let’s choose what the “FanGraphs projection” even is. Right here’s the related part of Acuña participant web page:

Eight projections, every with tons of numbers. That’s lots! However once I say the “FanGraphs projection,” I’m referring to the primary inexperienced row, the FanGraphs Depth Charts projection or FGDC. That’s the top-line projection we use wherever on the web site that pulls in projections to make predictions. Once you see “2026 (Proj),” it’s utilizing that quantity until in any other case acknowledged.

That’s settled then, proper? We’re projecting Acuña for five.4 WAR. Why did I’ve to waste your time with an article about it? It has to do with how we make that projection, a course of you’re about to find out about, in all probability in additional element than you needed.

Projection programs are, at their core, fairly easy issues. They may get there in sophisticated methods, however they’re all attempting to guess how good at baseball a given participant goes to be sooner or later, they usually all attempt to try this by predicting the outcomes that that participant goes to accrue on the sector. In case you have a projection system, you might have projections for homers, singles, strikeouts, walks, stolen bases, and so forth, in addition to a projection for plate appearances, so that you by definition even have a per-plate look projection for every of these statistics.

We don’t use a single mannequin in our official FGDC projections. We take a 50/50 mix of the per-plate look projections from ZiPS and Steamer. These projections work very well! I’m not going to enter the specifics of validation right here, however multi-model predictions constructed from good fashions are inclined to work effectively, and that’s what we use. The speed statistic projections we spit out are nearly as good as we are able to presumably get them, a minimum of in our estimation. We discard the enjoying time projections from ZiPS and Steamer, although, as a result of we have now a separate course of for projecting enjoying time.

Going from that blended fee statistic projection to the numbers you see on the web site requires enjoying time estimate. Fortunately, that’s what Jason Martinez and Jon Becker do at RosterResource. They mission enjoying time for everybody in baseball; we multiply that enjoying time by the per-plate look outcomes, and that offers us a projected batting line. It additionally provides us projected baserunning and protection, to not point out all of the related metrics for pitchers, however for at present’s instance, we’ll be specializing in Acuña’s batting statistics. You possibly can think about the identical course of enjoying out in each side of WAR.

How does the RosterResource crew assign enjoying time? They fill out every roster with their greatest guess of what that crew’s enjoying time will appear like. They don’t forecast accidents – how might they? Our depth chart projections appear like a depth chart, naturally sufficient. The starter on the depth chart is projected for a full-time function. That’s simply how it’s.

This makes good logical sense, although it would take a second to wrap your head round. Why not attempt to mission the imply consequence, with some probability of damage? Largely as a result of that’s not what persons are truly searching for in projections. “What number of residence runs do you mission Aaron Choose to hit this yr?” inherently assumes a wholesome season. “Solely 15, as a result of I mission him to get harm on Might 7” can be a extremely unusual reply to that query. Should you’re questioning what an on a regular basis participant’s projection appears to be like like, “in a full season” is an unstated norm. The one exception to this rule is a sound one: at present injured gamers see their enjoying time lowered by the estimated time it should take them to return from damage. Zack Wheeler, for instance, is down for 127 innings this yr, as a result of he gained’t be prepared for Opening Day. For at present wholesome gamers, although, we assume continued well being.

Take Acuña’s projection, for instance. He’s averaged slightly below 500 plate appearances per full season to this point, with a number of catastrophic accidents slowing him down. If he reaches his FGDC projected 651 plate appearances, it might be the third-most he’s taken in his profession. However we’re not attempting to guess how effectively Acuña’s knee would possibly maintain up in our preseason projection. We’re projecting what he’ll appear like if he performs a full season. That’s what folks need, in my estimation, the factor they’re truly asking for. Perhaps not each particular person. However within the mixture, I feel that’s clearly the commonest use case for a projection system, probably the most honest solution to reply the query of how good a man goes be, which is what these fashions had been constructed to deal with.

Should you’ll permit me a quick playoff odds tangent, I’ve truly spent a great deal of time excited about how this method interprets to odds projections. And over-allocating enjoying time to the highest gamers on this method nonetheless works effectively in our playoff odds mannequin. We simply add up the person contributions of each participant on a crew to determine how a crew tasks to play, then toss these outcomes into the BaseRuns formulation, which does a wonderful job of turning team-level outcomes into run scoring estimates. Certain, projecting Acuña for 651 plate appearances is just a little optimistic, however we’re symmetrically optimistic. Now we have Mike Trout down for his most video games since 2019. Now we have Alex Bregman down for 679 plate appearances, a quantity he’s solely hit as soon as within the 2020s. It is a function, not a bug: We mission groups at full power within the FGDC numbers. That playoff odds mannequin isn’t even WAR (effectively, the WAR-based odds mannequin is, however that’s a narrative for an additional day). It really works completely effectively merely by consequence statistics.

Our odds already deal with in-season accidents shortly; if a participant will get harm, their enjoying time declines to replicate that damage. We’ve additionally experimented with accounting for depth and probability of damage in our playoff odds. That technique feels promising, and I’ll have extra on it as we get nearer to the season. However that’s a small adjustment to an already-good mannequin. The prevailing playoff odds mannequin already handles the optimistic enjoying time allocation effectively, one thing we’ve examined on a number of events.

Alright, let’s get again to WAR. The FGDC projections for residence runs, singles, and the like are a fantastic estimate of what Acuña’s statistical outcomes will appear like if he performs a full season, of that I’m positive. However WAR isn’t an on-field statistic. It’s a derived statistic we add later that relates a participant’s contributions to what a replacement-level participant would do. In the course of the season, that is easy to calculate. We take precise outcomes and use them to outline substitute degree. Everybody’s efficiency will get measured relative to that league-wide substitute degree. The system works in the best way you’d count on. It reduces all of the totally different dimensions of efficiency right into a single foreign money of runs, compares these runs to a baseline set by league-wide manufacturing, and turns these inputs into WAR.

Projection programs don’t have that luxurious. They don’t know what the league common outcomes will probably be in 2026. They mission a WAR determine, in fact, however that’s relative to a baseline that they calculate of their fashions, based mostly on how that individual system handles enjoying time allocation. It’s solely pure: To calculate wins above substitute, you must know substitute.

On the particular person mannequin degree, that’s fantastic. However we’re not utilizing a person mannequin. We mixture fashions at a per-plate look degree after which assign our personal totally different enjoying time, which is typically meaningfully totally different than what both Steamer or ZiPS used. Once more, we try this on function. We achieve this as a result of it helps create team-level projections the best way we wish, and helps reply the “What would this participant appear like in a full season?” query that the majority everyone seems to be asking.

How totally different would the scoring atmosphere be if each star performed as often as our Depth Charts enjoying time suggests? Let’s cease for an instance. Take into consideration the most important league hitters who performed in 2024 and in addition in 2025. Should you take their 2024 outcomes and weight them by their 2025 enjoying time, you get a .311 wOBA. In different phrases, the expertise degree of the gamers who performed in 2024 and 2025, in 2024 phrases, is a .311 wOBA.

Strive that once more with our Depth Charts projections for 2026, and also you’ll get one thing totally different. The expertise degree of the gamers who we’re projecting to look in 2026 – their 2025 wOBA weighted by their 2026 enjoying time – comes out to a .316 mark. That’s a significant distinction, though it sounds small. Over 600 plate appearances, that’s an additional three or so runs. Over the 193,000 or so plate appearances that the gamers getting back from 2025 to 2026 are projected to accrue, it’s extra like 800 runs. In our hypothetical 2026 world, the one which we construct by taking fee statistics from the fashions and including our personal enjoying time, the offenses are all good.

The pitching staffs are all good too, in fact. However our Depth Charts rendition of expertise doesn’t measure this interplay. The hitters are projected to be actually good, as a result of their projections are towards a decrease baseline. The pitchers are additionally projected to be actually good, as a result of we do the identical sort of over-allocation of enjoying time to stars on that facet of the ball. However since we simply take Steamer and ZiPS fee statistics reasonably than performing some sort of adjustment for opposition on our finish, we don’t dock any batters for his or her anticipated more durable competitors. The result’s a sort of Lake Weobegone projection: We predict that each batters and pitchers will probably be above common in 2026.

That sounds fairly bizarre, nevertheless it’s not an issue on the subject of projecting the season. Once we run our playoff odds, we take crew runs scored and runs allowed estimates produced by our depth charts, flip that right into a successful share, and thus get a measure of crew power that doesn’t account for opposition. To simulate a sport, we simply examine the crew strengths of the 2 groups enjoying. Since everyone seems to be over-estimated and we’re solely evaluating groups comparatively, the projections nonetheless work correctly. We could be projecting a way more proficient league in 2026 than in 2025, however there aren’t any further wins to go round on the finish of the day, and since we’re measuring everybody relative to their opposition, it simply works.

There’s some bizarre math stuff right here that you must maintain in your head. There’s no approach that each one of our WAR projections may be proper. It’s actually unimaginable. There’s a set quantity of WAR to go round yearly. Now we have our hitters down for greater than that. If we’re truly proper about enjoying time, the baseline degree will merely be larger, and the identical quantity of offensive contribution will simply be value fewer wins above substitute.

Why then don’t we simply modify the substitute degree baseline up in our FGDC projections? It seems like that will be the best answer. However should you cease and give it some thought, that’s positively not the fitting factor to do. If we modify the baseline larger, we’re implicitly saying one thing concerning the worth of manufacturing in 2026. The speed stats? These projections are good, and we’re not touching them. If we alter the mapping between that offensive manufacturing and WAR, although, we’re beginning to do one thing that is not sensible.

What’s it going to appear like if Ronald Acuña Jr. performs a full season in 2026? We predict that it’ll appear like a .286/.393/.512 slash line with 31 homers. If he accrues these numbers, what do we expect his WAR will probably be? Our 5.4 estimate might be fairly good, though the best way we get to it appears bizarre. That’s as a result of the precise noticed expertise degree within the majors in 2026 isn’t going to be wherever close to what our purposefully-optimistic enjoying time projections would counsel.

Right here’s the actual query persons are asking once they take a look at a WAR projection: How helpful will this participant be, relative to an inexpensive league common baseline, if he performs a complete season? If we re-centered our preseason WAR estimates across the larger common degree of competitors that we mission, we’d be answering a unique query: How helpful will this participant be if he and in addition each different star in baseball performs a complete season? That second query isn’t the one which’s high of thoughts. Clearly, then, that technique wouldn’t work.

The query we reply, the one which I feel matches what folks need, has issues of its personal. Our outcomes are unimaginable by nature. They gained’t be true within the mixture. Happily, although, none of these issues truly matter all that a lot in the actual world. Our projections nonetheless work as a result of they take a look at relative crew power. The Depth Charts would possibly allocate an excessive amount of WAR to each crew, nevertheless it doesn’t allocate an excessive amount of WAR to any star: Every projection is that star’s baseline consequence in a wholesome season, precisely what we wish it to be.

I can inform, studying again over this text, that it’s going to be a complicated one. It’s laborious sufficient to speak about how WAR works after we’re already-accrued outcomes; layering in projections and re-centering and allocation clouds the matter even additional. The excellent news is that the takeaways are easy. Our projections do job of determining relative crew power. Additionally they do job of projecting participant expertise degree. They take a couple of mathematical shortcuts to get there, however these shortcuts don’t intrude with both of the 2 principal issues the projections are for: answering the questions “How good will this man be?” and “Will my crew make the playoffs?”