Fernando Tatis Jr. As Statistics Lesson

If you’re a time traveler who jumped forward from late May 2026 to today, you’d probably recognize most of what’s going on. Gas is expensive. People hate data centers. Super El Niño is coming, even if no one knows exactly what that means for the weather. But if you’re a time traveler from San Diego in particular, one thing will no doubt floor you: Fernando Tatis Jr. isn’t in the biggest slump of his career. He’s been the second-best position player in the majors since the calendar flipped to June, in fact. So let’s make a trade. You can give me the secrets of your time travel down in the comments. I, in turn, will try to use Tatis’ resurgence to explain variance. No pressure.

Tatis hit his first home run of the year May 30, in his 239th plate appearance of the season. Up to that point, he’d looked like a completely different hitter from his 2025 form. Here’s a big old smorgasbord of statistics that changed:

It’s almost too much to think about. He swung harder and hit the ball softer. He put it on the ground too much, but also didn’t pull it often enough. If you looked up what not to do in a baseball textbook, it’d probably say something like that. Now let’s add on the rest of the year:

Oh, huh. Now he’s better than he was in 2025? What in the world? Well, luckily, I can tell you one thing that changed: his batting stance. As Jared Greenspan detailed this June, Tatis went from an open stance in 2025 to a neutral stance in early 2026 and slumped hard. Then he changed it back in the middle of May and started hitting like his old self.

One way to think about this: Hey, mission accomplished. We solved the mystery, and it’s time to pack up shop. And honestly, Greenspan’s article did a great job of that already. But I’m interested in something that’s a little harder to pin down. Tatis was great, and then he was terrible, and then he was great again. In fact, plenty of players follow that pattern – or maybe they’re horrible and then great, or bad and then OK and then good, or OK then good then bad. And that fact is interesting to me, as someone who thinks a lot about how good baseball players might be in the future. Are these all cases of batting stance changes? Most? Some? Only this one? Is that even the right way to think about it? Tatis is the best example I’ve seen in a while, which is why he’s an excuse for me to write this article.

Why did Tatis change his stance from open to neutral in the first place? At the simplest level, probably because he wanted to hit better. And he didn’t just make that stance change in a vacuum. He started swinging harder, and he started swinging and chasing much more. Some of that might come down to his new swing mechanics changing what pitches he offered at. Some might come down to approach, or to how he’s pitched, or to an intent to swing more. And all of those factors are changing at the same time too. Tatis knows how pitchers pitch him – but pitchers know that he knows how they pitch him, and he knows that… well, you get the idea.

It just so happens that MLB’s fancy cameras measure the adjustment that Tatis made, but that’s not always going to be true. Hitters make all kinds of changes that we don’t measure. Heck, hitters make all kinds of stance changes that we don’t measure. Statcast gives us foot position, which is amazing, but it doesn’t record hand placement, or hip load, or initial bat positioning, or any of the myriad ways that a human being can stand with a bat in their hand while they wait for a pitch.

The full picture is even a little more complicated than that. In addition to all the things batters change that we don’t measure, there are a ton of things batters change that we do measure, and not all of them matter all the time. Even more confusingly, we don’t measure everything that matters. (If I knew what we were missing, I’d tell you, but it’s foolish to think that we’re capturing everything that makes a hitter tick.) I think it’s pretty clearly true that moving his left foot one step back isn’t the difference between a .266/.340/.318 batting line (through May) and a .300/.365/.550 one (since). But I’d be lying if I told you that a single other change made this happen, or that three factors were responsible. The system is too complex.

The news gets even worse. Even if a player’s talent level didn’t change, his results would fluctuate quite a bit from random chance alone. I took Tatis’ aggregate 2026 numbers and turned them into a weighted random outcome generator. No cold-and-then-hot shape, no change in true talent: I just told my computer that his 2026 home run rate, single rate, strikeout rate, walk rate, and so on were all his actual skill level. Then I asked it to simulate his season in two parts: all of his results up through May, and everything since. Roughly 5% of my May samples were worse than Tatis’ actual results. Roughly 10% of my post-May samples were better than Tatis’ actual results. In other words, neither is likely by chance, but neither of those is impossible either.

I came up with an even more interesting test, too. I split Tatis’ real season at the points that would create the greatest gap between before and after batting lines – May 18. The gap between his batting line up to that day and his batting line after it is a massive 134 points of wOBA. That’s the difference between Pete Crow-Armstrong’s 2026 and the worst qualified batting line in the majors this year. (Poor Ezequiel Tovar, you know things are bad when your non-park-adjusted batting line is this bad despite playing your home games at Coors Field.)

Then, I took all the actual results and shuffled their order randomly, producing a different season but with the same overall batting line. I did that a bunch of times, and each time, I measured the greatest split in those seasons – seasons exactly identical to Tatis’ in aggregate, but in nonsense order. You can think of this as a counterfactual: If I knew, beyond a doubt, that Tatis’ true talent didn’t change all year, how likely would it be that we’d find a set of hot and cold streaks with this much of a wOBA gap?

In 5% of the seasons I sampled, I found a gap that large or larger. And if you were interested in the through-May versus the post-May split, 24% of randomly ordered seasons had divergences of that size or greater. That’s not to say that Tatis didn’t change something. I feel pretty confident that his stance change helped him improve at hitting, because the timing change is so perfect, and that 5% number is encouraging. At the very least, I believe that he made an adjustment and started getting the better side of variance. But if we’re being honest about the data, we have to admit its limitations, and the truth is that batting results are so noisy that using them as proof of a change in talent mostly doesn’t work. Using them to measure the magnitude of a change in true talent? Crazy talk.

In reality, a ton of players make mechanical adjustments every year. Most of them don’t lead to huge slumps and then massive hot streaks. We’re talking about Tatis because he went from awful to amazing, but if someone made the same change and went from meh to meh, we’d never learn about it. So maybe he’s just the example we point to because luck went his way. It’s a spiritual sibling to the Bonferroni correction. Or what if the change in Tatis’ results is partially mechanical, partially game theoretical approach, and partially random variance? Good luck teasing that out of the data.

The truth is that numbers lie all the time, and the people explaining the numbers to you lie as well. They don’t always mean to. They get fooled by randomness, or they come in with a pre-conceived idea of what’s going on and find numbers that fit the narrative. I’m not just talking about other people; I do this too, though hopefully not as often as I used to. But, if you use baseball statistics to try to tell a story about baseball players, you’re exposing yourself to plenty of confusing and hard-to-measure randomness.

My mental model for player performance is that three different factors are always swirling around, making it difficult to pin down what’s real and what’s noise. There’s player true talent – we don’t observe this directly, but we can guess at it. Aaron Judge is a better hitter than Patrick Bailey. True talent isn’t stationary – players get better or worse all the time, and their health surely plays a part in this too – but it is fairly stable. Sure, some players get massively better from one year to the next, but most don’t. For the most part, how talented someone is now is a good predictor of how talented he’ll be in a year. Judge isn’t just getting unreasonably lucky every season, obviously.

There’s the metagame, the cat-and-mouse battle between pitcher and catcher. We don’t observe this one directly either, but it’s clearly there. A batter might sit fastball because he’s seeing a ton of them and prosper as a result, but pitchers probably won’t keep playing into his approach in the long run. Batter and pitcher strategies are always changing, and they’re game theoretical – the right answer for each player depends on the choices the other makes. That makes this factor incredibly noisy. Strategies are always changing on purpose, and even mixed strategies – ones where you use randomness to help make decisions rather than choosing one approach every time – often have unstable equilibria. A batter might have a great approach one month, then have to discard it a month later because pitchers are doing something different. In the long run, there are surely some players who are skilled at out-approaching their opponents – but even they have up and down periods, because how could they not when both sides are constantly changing how they attack the other?

Finally, there’s random variance, statistical noise. Baseball is full of this, to the point where it’s just assumed. Seeing-eye singles are named as such because they shouldn’t get through the defense but do anyway. Bloop hits are line drives in the box score tomorrow. Hitting evens out. Tatis has a .452 wOBA over the past two weeks, while Victor Mesa Jr. clocks in at .498, but I don’t think anyone expects Mesa to outhit Tatis going forward. Baseball is noisy. We all know it. Some of it is where the ball bounces, some of it is a round ball on a round bat, some of it is stubbing your toe on the way to the batter’s box and being distracted when you really need to be locked in.

The tricky part comes when you try to figure out how all three of these factors work together. Any change in performance could be a function of some or all of them. In fact, every change in performance is probably a mixture of all three. Maybe Tatis’ stance was interacting poorly with the way he was being pitched. Maybe it really did make him worse. Maybe he was adopting a poor strategy against the way he was being pitched, and changing his stance helped him change the strategy. Maybe the way the ball was bouncing early in the year made him doubt his strategy, and that was making him worse. We’ll never know, because all we see are the outcomes, and the outcomes aren’t clean functions of any one thing.

There’s a common criticism of baseball-statistics types like me: We don’t care about what happens on the field. I’d argue that’s getting it backwards, though. I care tremendously about what happens on the field. It’s the best data I get to improve my understanding of baseball, and it’s also the only way that I can test my predictions. The problem, as I see it at least, is that I’m painfully aware of how noisy the random variation piece of the puzzle is, and that makes me far less confident that what I’m seeing on the field today is a faithful preview of what will happen on the field tomorrow.

When I was assembling this year’s Trade Value list, I had a tough time placing Tatis. I did a lot of research and thinking before ranking him 50th. I read Greenspan’s article and found it compelling, and I looked at plenty of statistical data myself. But the biggest thing that made me think Fernando Tatis Jr. would hit well going forward wasn’t his reverted batting stance. It was the fact that he’s Fernando Tatis Jr., a man with a 135 wRC+ across nearly 3,600 plate appearances of major league baseball and a 133 wRC+ across the two seasons prior to this year. I might not know how he’ll figure it out, or what stance will best unlock his thunderous power, but I do know that he tends to hit pretty well over time. That instinct can lead me astray – sudden collapses and breakouts happen – but it’s served me pretty well in the past, because results are noisy and looking through them for signal is necessary if you want to be any good at the prediction game.

That’s my attempt at prioritizing true talent over the other variable parts of performance. Obviously, I’m not 100% confident that I got the call right. It’s impossible to be 100% confident of random processes like this. That’s why we statistical types are so maddeningly wishy-washy. And just so we’re clear, this isn’t me taking a victory lap on my placement of Tatis. If we’re being results-oriented, I was clearly too low! He has the highest wRC+ in baseball since the day the article came out.

I suspect that I’m preaching to the choir here. People who read FanGraphs are already self-selecting for this type of analysis. But it’s been on my mind lately, and when I started to analyze Tatis’ turnaround, I couldn’t stop thinking about this. If you asked me to project his performance for the rest of the year, I’d care some about his recent form. Like I said, he’s been the second-best position player in the big leagues since the start of June. But I’d care more about his long-term form, and I wouldn’t disregard the ice cold start to the season either. That’s part of the full picture of Tatis, as much as his recent scalding form is. I don’t think Tatis has the best bat in the game, regardless of what he’s just done. I didn’t think he was a below-average talent when he was struggling, either. The truth is somewhere in between, as it so often is.