There’s Something About This Year’s Hitters

Wednesday night, Justin Smoak hit another home run. It was a rather big home run, a dramatic ninth-inning home run, but as far as these purposes are concerned, all that matters is that a home run happened, nevermind the context. A couple months ago, it looked like Smoak could be breaking out, at last. He hasn’t slowed down even the tiniest bit. He’s been one of the more dangerous hitters in all of baseball, and for as much as the Blue Jays’ season has gone down the toilet, Smoak’s made for a great story. His career WAR is 3.4. His 2017 WAR is 3.1.

I’ve been thinking about Smoak a lot. But then, there are also other dots to connect. A story that’s similar to Smoak’s is that of Logan Morrison. In the AL West, Yonder Alonso has turned himself into an offensive weapon. There’s also been the unexpected breakout of Marwin Gonzalez, and while I don’t want to just go down a list name by name, there have been other big surges, and also a number of shocking collapses. Carlos Gonzalez has fallen apart. Jonathan Lucroy, too. Names and more names and more names.

It feels like hitters have been particularly unpredictable. But there could be a strong element of recency bias — I remember this year’s studies the best, and examinations tend to focus on the biggest surprises. So I tried to dig into the numbers. Turns out it’s more than just a hunch.

I don’t want to bore you with a whole bunch of methodological details. So I’ll try to be quick! I looked at how things have gone since 2002. I chose that season as the beginning because that’s as far back as I can access using our splits leaderboard. I focused my attention on year-to-year wRC+, to see how it’s held up. I set a minimum of 250 plate appearances, and for the second season in every sample, I included data only through July 25, to stay consistent with what we have for 2017.

That’s confusing. Let me try again. For every full season, between 2002 – 2016, I gathered data for players with at least 250 PA. Then I looked at how those same players hit in the following seasons, between 2003 – 2017, with the same PA minimum, and only using numbers through July 25. Do you get it? I hope so. Here is the overall plot, with a sample exceeding 2,800:

Here now is the plot for just 2016 – 2017:

I don’t know what, if anything, might jump out. In the second plot, you get a smaller slope. That’s indicative of something, but to go in a slightly different direction, perhaps you’re more accustomed to looking at r-squared values. Let’s do that. For the sample from 2016 – 2017, you get an r-squared of 0.08. Seems pretty small, right? But we don’t have any frame of reference. Let’s establish that part. Here are all of the r-squared values from the past decade and a half:

For the entire sample, the r-squared value is 0.21. The very highest mark is 0.32, from the first pair of seasons. The previous low was 0.14, for 2010 – 2011. This time around, again, we’re at 0.08. I’m not good enough at this to tell you exactly how significant that is, but the weakest relationship has been established. It’s not just anecdotal. Offensive numbers in 2017 bear the least resemblance to the same numbers from 2016, relative to the decade and a half under the microscope. It might not be anything earth-shattering, but it does look like something has been going on.

This is something that can be further explored when the season is over. Might as well let the last two months play out, in case the numbers feel like getting around to normalizing. And, ideally, at season’s end, this kind of examination can be performed comparing numbers against projections, instead of just numbers from the previous year. Projections and previous-year numbers will tend to look pretty similar, but the projections are always the recommended baseline. I didn’t use them here because I don’t have ready access to preseason projections from the past several years. I know they’re out there, though, and so this study could and should be repeated in November or December. It’ll be interesting to find out whether this season’s hitters have really been the least predictable. It’s something that could be a fluke, and it’s something that could be more than a fluke.

Take that drop in r-squared for 2010 -2011. The next year, the relationship bounced back. It wasn’t the start of a trend. But that year was that year, and this year is this year. Every year is different, and as you attempt to imagine explanations for why this might be a real thing, you’re probably drawn to the idea that there’s so much information available out there. Hitters are changing in response to Statcast data. Pitchers are adjusting, themselves, in response to similar information. It feels, at least in theory, like changes could be happening faster than ever, in either direction. Players can be more precise about identifying their strengths and weaknesses, so there’s less guesswork. Hitters have a better idea how they might improve. Pitchers have a better idea how hitters might be attacked. It’s informational warfare, to an extent never before experienced. It’s all kind of a grand theory, and nothing else, but it at least feels like it could be true. It’s a start.

Maybe it’s nothing. Maybe this won’t hold up at the end of the year; maybe this won’t hold up when the numbers are compared to the preseason projections. I don’t know, because the season is only two-thirds over, and I put this all together in just a few hours. This’ll eventually be examined in much greater depth, I assume. For now, just know, if you feel like it’s been a strange season for hitters, there’s a reason behind that feeling.