Making sense of Alexander Colomé, part 1

Rays closer Alex Colomé hasn’t been very good this year. He’s only pitched four innings so far, and all sorts of crazy things happen in four innings, so the numbers themselves aren’t really a reason to be concerned. Still, Rays fans are concerned, because while Colomé was absolutely one of the best relief pitchers in baseball back in 2016, in 2017 he was just good enough. The results so far in 2018 fit neatly into a worrying trend.

This is the start of a three part series to investigate what happened to Alex Colomé. Why was his 2017 worse than his 2016? Which version is the real one? Can the downward trend be reversed?

Part one of this three part series will focus on describing the easily observable differences between Colome’s two years as a backend reliever.

Setting the Question

Throw out the small sample size struggles of 2018 and concentrate on the deterioration of of Colome’s overall results in the years previous.

Across the board of ERA and its estimators, Colomé got worse. The clearest evidence was the strikeout rate which dropped over 10%, while Colomé walked a few more batters as well. His ERA was saved from an unacceptable 4.00+ rate by the fact that while most of his peripherals went in the wrong direction, fewer than expected of the fly balls he allowed left the yard in 2017.

Pitch Usage Percentage: Fastball vs. Cutter

One difference is obvious, and possibly important. Since his move to the bullpen, Colomé has relied on two pitches, a mid-90s fastball that can reach the upper 90s at times, and a high-80s cutter. The usage rates have changed.

In 2016, Colomé threw his two pitches in nearly equal amounts, with the fastball offered 52% of the time and the cutter 48% percent of the time. That’s already a high usage rate for a breaking pitch (no, a 90 mph cut fastball is not generally called a “breaking pitch” but I think it’s the appropriate term based both on the movement of the pitch—big for a cutter—and how and where Colomé deploys it), but in 2017 Colomé took things a step further, settling on the cutter as his primary offering and throwing it 67% of the time. So far, in 2018, the usage rates have grown even more extreme.

It’s tempting to stop here, saying, “Alex Colomé did better when he threw fastballs and cutters in equal amounts, and did worse when he threw more cutters than fastballs. Therefore, he should return to 2016 usage levels. Q.E.D.”

Maybe that’s right. Maybe, while the cutter is by most measures Colome’s best pitch (more on that later), it’s best used playing off of a well-established fastball.

That feels overly simplistic though, and it doesn’t address the question of why Alex Colomé started throwing more cutters. The Rays are smart, and they pay attention to pitch mix numbers like this. Why would Colome change a process that was working? Why would he not return to the old approach once his results got noticeably worse?

I don’t really have an answer to that, but it’s a set of questions to keep in mind before we get lost in the weeds of Colomé’s pitching. Also remember that baseball doesn’t stand still. When a player is as good as Alex Colomé was in 2016, everyone else notices, and they all try to figure out how to adjust. The thing that worked once upon a time is not guaranteed to work again.

Now further into the weeds.

The Stuff Hasn’t Changed

The other obvious place to look is at the movemement and velocity of Colome’s pitches. There’s just not a whole lot there though. Velocity has stayed mostly the same.

As has vertical movement.

The horizontal movement looks a little more interesting, partly due to a weird March for the radar at Tropicana field (the horizontal movement on everybody’s pitches was reported as abnormally armside for a righty, before the system was recalibrated), and it’s certainly possible that Colomé wore down in the last month of the 2017 season (based on some changes to his cutter profile in September), but I don’t think the overall answer to the Colomé question lies in pitch velocity or movement.

Let’s look elsewhere.

Fewer Whiffs on the Cutter, and a Major Change in Location

One big change for Colome was that batters stopped missing his cutter at the outrageous rate they had before. Data from Brooks Baseball:

The smaller change was against right-handed hitters, and I don’t really have a good explanation for why, other than that he threw it more and caught a little bit more of the plate (which seems like a plausible explanation, if not an interesting one). You can stare at the graphs yourself if you’d like. Here is Colomé’s cutter location against righties in 2016, and here it is against righties in 2017.

But take a look at what Colome was trying to do against lefties, where the bigger change happened. Here’s 2016:

And here’s 2017:

Observation: that’s different.

Colomé went from emphasizing his breaking ball down and in on the back foot to opposite-handed batters to instead living on the down and away, threatening the back door, and he ended up with exactly the results one would expect. Generally speaking, a backfoot breaking ball gets whiffs. It’s somewhere pitchers go when they’re aggressively looking for a strikeout. The backdoor, and especially the low back door, is a location pitchers use to minimize hard contact against them, while also picking up some called strikes.

And that’s basically what happened. Here are a bunch of numbers thrown at you all at once. For scale, we’re talking about 160 cutters with 26 put in play in 2016, and 328 cutters with 56 put in play in 2017.

xwOBAcon (expected wOBA on contact, based on velocity and angle off the bat), wOBAcon (actual wOBA on contact), and Called Strikes% are calculated from Baseball Savant data. The rest is from Brooks Baseball.

By changing location from the back foot to the back door, Colomé seems to have traded whiffs for fouls and called strikes. When the ball was put in play against him, it was most often on the ground, and soft.

Assuming this was a conscious trade-off being made by Colomé and the Rays pitching department, did it work? Narrowly speaking, we can look at overall results on the cutter vs. lefties (meaning strikeouts, walks, and balls in play) and say that it made things marginally worse.

But that’s probably too narrow of a focus, so let’s keep moving.

Correction: An earlier version of this article displayed a warning that the wOBA numbers in the above article were incorrect, and stated that they would soon be updated. The editor who published that “correction” was confused. The above numbers are unchanged, and remain correct, at least as far as the calculation goes (arguments about their conception would be considered and welcomed). That editor was also the author. Basically he forgot what he had done the day before. These things happen. Both parties are sorry.

Fewer Whiffs on the Fastball Too

It wasn’t just Colomé’s cutter that produced whiffs less frequently in 2017 than in 2016. His fastball followed a similar trajectory.

The zone profile pictures for Colomé’s fastball are less clear than those for the cutter, especially against lefties, but there are still two interesting changes in his fastball approach and results vs. righties (which is, perhaps coincidentally, where the bigger drop in whiffs took place).

The first is location. In 2016, he lived down and away to righties.

Not so in 2017. While there was still some concentration of pitches on the bottom and outer thirds of the plate, the overall emphasis moved from down-and-away to up-and-in.

Now look at the whiffs. These are presented with total pitches as the denominator, rather than with swings as the denominator, so as to give a better understanding of the small sample sizes we’re dealing in.

In 2016, while that previously-discussed down-and-away location wasn’t a source of whiffs, Colomé was able to get swings and misses with his fastball basically everywhere else, throughout the entire zone.

But in 2017 batters rarely fanned at it overall, and when they did, it was only in the traditional strikeout spots of at/near the top of the zone, and in on the hands.

A pitcher who can throw his fastball by hitters no matter where he puts it is in a good place. A pitcher who must hit certain whiff-inducing spots to get a punchout has a much harder path to trod.

If this was the first bit of data I’d looked at, I’d have guessed that Colomé had lost either velocity or movement on his fastball between the years, and that it simply became a lesser pitch. When a pitch is thrown less often, as was Colomé’s fastball in 2017, the per-pitch results are supposed to get better, not worse. But we’ve already checked and that doesn’t seem to have happened.

Is it possible that those fastballs down and away in 2016 were getting hitters leaning over the plate and setting them up to miss when they later got a fastball within the zone? Maybe. That sounds hard to prove. Harder to disprove.

To try to avoid missing the forest for the trees (a few swinging strikes), here are some overall numbers for Colomé’s fastball vs. right-handed batters. These numbers are compiled from 240 pitches in 2016, and 105 in 2017.

Well that certainly complicates things.

Are we really so sure that Alex Colomé pitched worse in 2017 than he did in 2016, particularly when we’re talking about about his fastball to righties? Yes there were fewer swings and misses, and fewer ground balls, but there were more called strikes, and overall less damage done (also note that if one believes in xWOBA as a predictive stat, one might conclude that Colomé had bad ball-in-play luck in 2017).

It’s not surprising that the move away from the down-and-away location lead to fewer groundballs and more fly balls. The surprising thing, though, is that in terms of ball-in-play results, that was an improvement for Colomé. While hard hit fly balls are what all the cool hitters seem to be looking for these days, the ones righties got off Colomé were not hit hard. Although pop-ups account for a negligible number of the pitches under consideration here, I included them in the table to make the point that not all balls hit in the air are dangerous. Going by xwOBA, Colomé’s were not.

Making the point further, his HR/FB against righties in 2017 was a miniscule 2.5%.

This is an unexpected result, so hold on to it for later, but remember that “fastballs against right-handed batters” is a limited view. We might still be in the trees.

The Sequence Has Changed

Brooks Baseball has a really great tool for visualizing pitcher approach. Here is Alex Colomé in 2016, and here is Alex Colomé in 2017. Open them up side by side.

Colomé started throwing his fastball less and his cutter more, overall, but the shift was was especially pronounced against righties, where the 2017 season saw not just a minor adjustment of the ratios but rather a largescale change in approach.

In 2016 Colomé was a traditional “set-them-up-with-the-fastball, sit-them-down-with-the-breaking-ball” pitcher (STUWTF,STDWTBB for short) against both hands. In 2017 he threw more cutters to everyone, but he pitched backwards against righties.

Maybe pitching backwards isn’t the right term, as that might denote going back to the hard stuff once he was ahead in the count. To be more precise, Colomé stopped offering early fastballs, using his cutter as the primary pitch in all counts to righties, and only throwing enough fastballs to keep batters honest.

Whether or not he succeeded in keeping them honest is an open question.

Conclusion and Next Steps

Pitching is hard, and so is analyzing pitching. Results are noisy, approaches (both of the pitcher and the batter) change, and true talent is a moving target. In some ways, as a two-pitch reliever with a clearly delineated “good year” and “not as good year,” Alex Colomé is easier to approach than many pitchers, but that doesn’t mean there are clear answers shouting from the spreadsheet.

Here is a summary of what I’ve observed:

• Colomé was more dominant in 2016 than in 2017.
• Colomé threw roughly the same number of fastballs and cutters in 2016, but in 2017 he threw more cutters than he did fastballs.
• That drop in fastball usage took place across the board, but it was by far most dramatic early in the count to right-handed batters.
• There was a significant change in the location of Colomé’s cutters to left-handed batters, with those cutters clustered on the “back foot” (down and in) in 2016, and on the “back door” (down and away) in 2017.
• Lefties hit Colomé’s cutter a bit less hard in 2017, but they also struck out less against it and walked more. The overall change in results on the pitch was slightly negative for Colome.
• Along with the overall lower usage, there was a subtle change in the location of Colome’s fastballs to righties, as he stopped emphasizing the bottom-outside corner of the strike zone.
• In a similar pattern to that of his cutter against lefties, Colomé missed fewer bats with his fastball against righties, but overall was able to avoid having the pitch hit hard.

To this point I’ve almost exclusively examined Colomé’s pitches on their own. That’s necessary for describing some of the changes in process, such as pitch location, but it ignores the obvious reality of pitch relationships and interactions. The resulting numbers from such a context-free analysis are fractured and suspect.

In parts two and three I’ll attempt to put these observations int a larger context.