“It’s better not to know so much about what things mean.”
– David Lynch in Rolling Stone, September 1990

A few friends and I have a recurring movie night where we take turns choosing the featured film for the evening. Because one friend has decided to make his picks in the “campy horror” genre, last week we wound up watching Peter Jackson’s Dead Alive (yes, THAT Peter Jackson). Rotten Tomatoes describes it as a “delightfully gonzo tale of a lovestruck teen and his zombified mother,” while Wikipedia goes with “zombie comedy splatter film.” It deals in absurdity and surrealism and its favorite tool of the trade is fake blood. The production reportedly went through about 80 gallons of the stuff.

Absurdist storytelling launders its messaging through exaggerated extremes and by defying or subverting logic in ways frequently so morbid or dark that they surpass tragedy and come all the way back around to comedy. Extremes that defy logic exist in baseball too. A particularly rich source being players’ home/road splits. I went searching the 2024 season for the most extreme differences in player performance (minimum 200 plate appearances) between their home parks and road venues across a variety of offensive metrics. In my own act of defying reason, I don’t really have an explanation for choosing hitters over pitchers. Maybe I’ll do pitchers in the future. Maybe I won’t. Who needs symmetry or balance in the universe? Anyway, I found the largest disparities, and ignored the boring, expected ones like Rockies hitters clubbing a bunch more homers at Coors Field, and instead, locked in on the truly bizarre.

Certain occurrences earn their bizarre status not because of their unexpected nature, but rather because they take an expected outcome to such an extreme as to feel over the top, or a bit “on the nose,” as an editor might put it. Dead Alive depicts Lionel, a young adult man, still living at home with his mother, an overbearing type who domineers his life. Lionel and his mother portray the standard “momma’s boy” archetype, but exaggerated to nth degree — the film culminates with the supercharged zombie version of Lionel’s mother inserting her son back into her womb, where she can finally regain complete control over his life.

Like an overbearing mother, certain ballparks have a strong influence on the type of hitter who thrives under their care. Some encourage power, or prefer a certain handedness, while others look down on hitting and choose instead to emphasize pitching and defense. Petco Park in San Diego does not favor offense in general, but it is among the least friendly ballparks for lefties who hit a bunch of singles. Enter Luis Arraez, the singles hitter of all singles hitters.

The infielder/DH was traded to the Padres from the Marlins last May 4. Like Lionel, who in the early scenes of Dead Alive meets a nice young woman named Paquita and takes her on a date to the zoo, Arraez continued to do his thing for the month of May, hitting 38 singles, compared to the 30 he hit during the first month of the season. But then Lionel’s mother interrupts the date, gets bit by a Sumatran Rat Monkey, and chaos ensues, just as the influence of Petco Park eventually exerts its will on Arraez. He ended the season with a .268 average at home and a .359 average on the road, due in part to his hitting about 20% fewer singles (71 vs. 90) and almost 50% fewer doubles (11 vs. 21) at home compared to on the road. This placed him at the extreme end of Petco Park’s offense dampening effects, so extreme as to feel like the stadium stuffed Arraez inside her womb until he learned his lesson about hitting all those singles.

Batting average is one thing, but there are other stats that you wouldn’t necessarily expect to have extreme home/road splits; similarly, you wouldn’t necessarily expect a scene at the beginning of a movie that depicts the main character mowing the lawn at the behest of his mother to foreshadow a momentum shift in the big fight scene at the end. Nevertheless, Brice Turang’s stolen base success rate was 15 points higher at home than on the road, which was the largest differential among base stealers with at least 30 attempts (omitting Jazz Chisholm Jr. who was around 20 percentage points better on the road, but also switched home stadiums in late July). The Brewers second baseman stole 28 bases at American Family Field and was caught just one time there, while in away parks he stole 22 bases and was caught five times. The discrepancy becomes all the more notable when considering Turang reached base less frequently at home, posting a .290 OBP in Milwaukee compared to a .341 OBP everywhere else.

There aren’t too many data points to suggest why Turang was better at swiping bags at home, but as a player with just over 1,000 big league plate appearances, it makes sense that some of his visual timing and positioning cues might be more locked in at AmFam than they are elsewhere in the league. Things like the first base cutout in the infield grass and the sightlines behind the pitcher as he’s taking his lead from first are likely more dialed in at the place where Turang has taken the majority of his reps in the majors. Using one of his strongest tools (94th percentile sprint speed) and the comforts of a familiar environment, Turang almost completely compensated for his otherwise negative contributions on offense, just as Lionel, in defending his home from a horde of zombie partygoers, turned to a trusted tool — his lawnmower and its sharp, speedy blade — to mow through the walking dead.

The largest split I could find with respect to wRC+, which is already adjusted for park factors, belongs to Luis García Jr., who after several up and down seasons with the Nationals, spent 2024 as Washington’s primary second baseman. The lefty logged a 156 wRC+ at home and a 63 wRC+ on the road, a 93-point spread. This is where it’s helpful to know exactly how the park adjustment is applied to wRC+ and why that might make a fairly neutral hitting environment like Nationals Park seem like an oasis for one hitter in particular. Or, in other words, why a young lady like Paquita might continue to see someone even after his zombie mother ate her dog.

(Here is where I must note that there is a character in Dead Alive named Scroat. Unfortunately, I couldn’t make a baseball analogy to Scroat because I do not remember which character was Scroat, and an IMDb search through the movie’s cast list does not have a headshot next to the actor who played Scroat. Really, I just wanted a chance to write Scroat in a FanGraphs post, so here we are. Scroat!)

Anyway, the park factor applied to wRC+ is a single value that captures the run environment in the stadium overall, as opposed to the more granular component level park factors that consider the stadium’s influence on the individual components of offense, such as singles, doubles, triples, home runs, etc. Component park factors that take into consideration the batter’s handedness are also available. Digging into the components of García’s home/road splits reveals that when in D.C., he struck out less and hit more singles and homers. Component park factors explain part of why García might benefit more from hitting in Washington than an average hitter: Nats Park does suppress strikeouts relative to its peers, and left-handed hitters get a boost with respect to singles. Fewer strikeouts means more balls in play at a ballpark where a ball in play off the bat of a lefty is more likely to lead to a hit. However, Washington remains neutral on home runs for those hitting from the left side. Looking at García’s splits with respect to batted ball characteristics reveal his home run-to-fly ball rate drops from 19.2% at home to 6.7% on the road. But it’s not just that the ball is carrying better because, additionally, his hard hit rate increases from 24.1% on the road to 38.4% at home. That García’s strikeout rate drops 10 percentage points in his home ballpark relative to everywhere else, in conjunction with his improved contact quality on fly balls, seems to suggest he sees the ball better at Nats Park than anywhere else. And for what it’s worth, a special aptitude for vision is what kept Lionel’s girlfriend from abandoning him, as she believed the tarot reading done by her seer/grandmother that foretold a fated, long-term romantic entanglement with Lionel.

Many don’t believe in fate and instead subscribe to the nihilistic view that the universe is composed of randomness, which at times manifests as utter, uninterpretable chaos. T-Mobile Park in Seattle is one of the worst ballparks for hitters, both overall and across all individual components, unless, by chance, you happen to be Luke Raley. The Mariners outfielder/first baseman defied the natural order of the universe (to the extent that there is one) and posted a .393 wOBA, 166 wRC+, and hit 15 home runs across 229 plate appearances at home, with a .295 wOBA, 91 wRC+, and seven homers over 226 PA on the road. Looking at component factors does absolutely nothing to explain Raley’s performance at T-Mobile Park, since as a lefty, all of Seattle’s horrible hitting juju applies even more so than it does for righties. His BABIP hovered around .300 both at home on the road, suggesting that if there’s luck in his performance, it was distributed evenly at home and on the road. In terms of his batted ball profile, Raley did have a higher hard hit rate at home, and he also pulled the ball more and put it in the air more, which collectively signals an overall higher quality of contact. Perhaps like the tarot-reading grandmother, Raley possesses some special sight that allows him to see the ball in a way that no one else has mustered at T-Mobile Park, or perhaps, as is the messaging of much absurdist art, we must simply submit to the random, chaotic winds of the universe, blowing some fly balls over the fence and leaving others to die on the warning track.

Whatever force is tasked with inflicting chaos upon the masses, it seems to enjoy unleashing Yordan Alvarez as often as possible. It’s true that Houston’s lefty DH/left fielder was not involved with the Astros’ banging scheme scandal, but he nevertheless is a frequent recipient of boos at away ballparks due to his uncanny ability to launch game-winning, soul-crushing moonshots in front of opposing fans. Though the booing is more of a vibes-based response, the data show that Alvarez does tap into his power more frequently on the road, hitting both doubles and home runs at a much higher rate, leading to a road wRC+ that is 62 points higher than his mark at home (a road advantage topped only by J.P. Crawford of the Mariners).

Again, wRC+ already accounts for the overall run environment, but not the components by which a particular player might be more heavily impacted. Houston’s ballpark, which is now called Daikin Park, grades out as neutral to hitters overall and with respect to left-handed home runs, but for doubles, a lefty hitter should have an easier go of it (though it’s worth noting Alvarez pulls the ball at a below average rate for lefties). But despite the neutral or better home park environment, in 2024, Alvarez hit 12 doubles and 13 homers across 315 plate appearances at home, while hitting 22 doubles and 22 homers across 320 plate appearances on the road. Alvarez also walked slightly more on the road, while holding his strikeout rate constant, suggesting a more patient approach that led to higher quality contact; this is reinforced by his higher home run-to-fly ball rate (20.4% vs. 11.7%) and hard hit rate (46.1% vs. 36.1%) away from Houston.

As with some of the other extreme splits, the increased patience and improved contact might mean that Alvarez doesn’t see the ball as well at Daikin Park as he does elsewhere. Or this instance of absurd home/road splits might be trying to send a different message. Absurdist art and its close relative, surrealism, frequently serve to defy logic, or at least quantifiable logic. At the end of Dead Alive, Lionel cuts his way out of his mother’s womb using a talisman that Paquita’s grandmother gave him for good luck. She probably thought its magical properties would prevent anything bad from happening to him, rather than its physical properties allowing him to puncture zombie flesh. Even magic follows no logical order.

Meanwhile, when asked to describe the experience of playing in a big league stadium in front of a packed crowd during the highest leverage moments of the game, players frequently use the word surreal. And in the surreal world, there wouldn’t necessarily be a logical explanation for why Alvarez becomes more powerful on the road, why he happens to be holding a talisman that can puncture the hearts of opposing fans. Maybe he feels less pressure away from the home fans. Maybe he takes a twisted pleasure in making a stadium full of fans fall silent. Maybe, like the zombies in the movie, he takes a poison intended for animals that has the unintended effect of supercharging his abilities. I’m mixing my talisman and poison metaphors now, but as previously established, there are no rules and nothing matters, so just roll with it and instead linger on the thought that if Alvarez ever leaves the Astros, he may morph into a supercharged monster permanently.

While we’re defying logic, I did stumble upon one member of the Colorado Rockies with a home/road split worth mentioning. In 228 plate appearances at Coors Field, Michael Toglia hit eight home runs; in 230 plate appearances away from Coors Field, he hit 17 home runs. So, in nearly the same number of opportunities, Toglia smacked more than twice as many home runs on the road as he did at Coors Field, a park notorious for juicing fly balls. My best guess is that the stadium’s reputation is doing psychic damage to a 26-year-old first baseman with just one full season under his belt. His hard hit rate is still higher at home, suggesting maybe he thinks that all he needs to do is swing out of his shoes and the thin air will do the rest. Meanwhile, his Med% is higher on the road and he hits the ball to the opposite field more often, suggesting a more controlled, purposeful swing away from the influence of Colorado. Maybe he’s overthinking the atmospheric conditions, or maybe he made a deal with an evil imp that granted him 60-grade raw power everywhere except the Mile High City.

Sometimes chaotic occurrences exist purely for comedic relief, offering no larger societal lesson or commentary on humanity. At one point in Dead Alive, Lionel visits his mother’s grave because he knows she’s a zombie and, therefore, not actually dead, so his master plan is to administer sedatives to her indefinitely in order to keep her safely in the ground. When he gets jumped at the cemetery by a band of local hooligans, he’s saved by a priest (literally, not spiritually), who seems to have exactly one skill, which is, as the priest puts it, to “kick ass for the Lord.” He does single-handedly wipe out the hooligans with what appears to be self-taught kung fu, but then promptly gets conscripted to the zombie ranks. The kung fu priest of baseball is Mike Yastrzemski, right fielder for the Giants, whose extreme singular skill is striking out way less at home than on the road. All of his other splits are as expected, but when batting at Oracle Park, he strikes out 19.7% of the time, compared to 32.6% everywhere else. It’s the most extreme strikeout difference in the bigs by a couple of percentage points.

In the movie’s next scene, Lionel has rounded up the current group of zombies, including the priest and a nurse, who was originally dispatched to look into his mother’s ailments before her transition to undead was complete. The priest and the nurse take a liking to one another and wind up birthing a baby zombie. This leads to a scene that was not in the original script and serves no purpose to the larger narrative; really, it’s just there for the jokes. Jackson decided to add it after they’d finished filming everything else, because they were still under budget, and since then, he has called it his favorite scene in the movie. For no comprehensible reason, Lionel takes the baby to the park, pushing it along in a stroller and mimicking the actions of the mothers he observes interacting with their babies. Perhaps Lionel thought that a change of scenery and treating the baby like a regular human baby would coax it into acting like a regular human baby, but it did not. Instead the viewer is treated to a series of hijinks, where the baby drags Lionel all over the park, and Lionel has to act like tackling a baby is perfectly normal behavior.

New Orioles outfielder Tyler O’Neill is the zombie baby hoping that a change of scenery does prompt a transformation. O’Neill experienced an even stranger flavor of Yastrzemski’s strikeout split. It’s not particularly unusual for hitters to strike out less in San Francisco (though not to the extreme reached by Yastrzemski), and the same holds true for Boston, where O’Neill played his home games last year. But O’Neill flipped the script; instead of striking out less at Fenway Park, he struck out significantly more frequently, posting a rate of 39.7% as opposed to 27.9% on the road. O’Neill hasn’t always struck out more at home than on the road. For example, during his final two years with the Cardinals, he was better in St. Louis than he was away from it, which is interesting considering that Fenway is a much more hitter-friendly park than Busch Stadium. It’s pretty funny to think that Fenway of all places could act as one hitter’s kryptonite, but the Orioles are hoping that was the case here. Perhaps getting O’Neill into a different park will do for him what Lionel couldn’t do for the zombie baby. If O’Neill’s overall line winds up resembling something closer to last year’s road performance, he’s much more likely to be a productive contributor in Baltimore. The spike in strikeouts caused his on-base percentage to crater to .301 in Boston, compared to .369 everywhere else.

For as much as we’d like for everything in baseball and life to follow some logical, rational, and quantifiable natural order, it doesn’t always work that way. There are too many lurking variables, agents of chaos, and forces we don’t yet understand. Sometimes it’s incredibly funny when something happens that we can’t explain. Sometimes it teaches us something completely separate from what we set out to divine. Sometimes we just have to accept that we don’t know what a weird thing is really about.

Last week, Michael Rosen wrote about Jack Flaherty’s delayed free agency market. Michael advanced a number of theories about why Flaherty hadn’t yet signed a deal, and what that might mean about his fastball, teams’ perceptions of his fastball, and the trajectory of his career broadly speaking. I found that piece really interesting – and I also started thinking about what Flaherty not having signed yet means in a larger sense.

You don’t have to look any further than last year to get an idea of what could happen to Flaherty. Blake Snell and Jordan Montgomery both waited a long time before settling for short-term deals. The year before that, Carlos Correa’s multiple failed physicals kept him on the market until the very end. In 2022, Correa, Kenley Jansen, and Trevor Story all found themselves looking for employment well into March.

All of those players came into the offseason expecting a major contract, and all of them ended up getting less than anticipated, bringing to mind some classic FanGraphs articles from Travis Sawchik, back in the halcyon days of 2018. Those articles drew on a study by Max Rieper that separated free agents into pre- and post-New Year’s signings and found a large discount for the latter group. Read the rest of this entry »

Earlier this week, I threw some numbers together on the value of productive outs. I focused on Corbin Carroll, and rightly so: His electric skill set is a perfect entry point for explaining how hitters can add (or subtract) value relative to average even when making an out. Putting the ball in play? We love it. Avoiding double plays? We love that too. The Diamondbacks are a team full of speedsters, and Carroll’s productive outs gave their baserunners a chance to show off their wheels.

A quick refresher: I calculated the difference in run scoring expectation between the average out and a specific type of out (strikeout, air out, non-GIDP groundout, double play) for each base/out state. Then I had a computer program tag each out made in 2024 with that difference. For example, the average out made with a runner on second and no outs cost teams 0.35 runs of scoring expectation in 2024. Groundouts in that situation only cost 0.25 runs, a difference of 0.1 runs.

Thus, on every groundout that occurred with a runner on second and no out, I had the computer note ‘plus 0.1’ for the “productive out” value. A strikeout in that situation, on the other hand, lowered scoring expectancy by 0.43 runs, a difference from average of -.09 runs. So the computer noted ‘minus 0.09’ for every strikeout with a runner on second and no out. Do this for every combination of base/out state and out type, add it all up, and you can work out the total value of a player’s productive outs. Read the rest of this entry »

Not every out is created equal. Take this fly out from Corbin Carroll, for example:

A lot of things can happen when you make an out with the bases loaded. You could strike out, leaving every runner in place. You could hit into a double play, an inning-ending one in this case. You could ground out some other way, or hit an infield fly. But Carroll’s here was the most valuable imaginable; with one out, he advanced every single runner, including the runner who scored from third.

Mathematically speaking, you can think of it this way. The average out that took place with the bases loaded and one out lowered the team’s run expectancy by a massive 0.61 runs in 2024. That’s because tons of these outs were either strikeouts (bad, runner on third doesn’t score) or double plays (bad, inning ends). But Carroll’s fly out was far better than that. It actually increased the run expectancy by a hair; driving the lead runner home and moving the trail runners up a base is exquisitely valuable.

That’s not the only way this could have gone. Consider a similar situation, a groundout from Aaron Judge:

Like Carroll, Judge batted with a runner on third and fewer than two outs. In this situation, the average out is bad, lowering run expectancy by 0.514 runs. But Judge’s was obviously worse. It cost the Yankees all the expected runs they had left in the inning, naturally, which added up to just a bit more than 1.15. Read the rest of this entry »

In 2008, the first year of PitchF/X pitch tracking, 13.9% of all pitches across the major leagues were sliders. Ah, those were the days – flat, crushable fastballs as far as the eye could see. More or less every year since then, sliders have proliferated. Don’t believe me? Take a look at the graph:

Are you surprised? Of course not. You’ve seen Blake Snell pitch – and Lance McCullers Jr., Sean Manaea, five of your team’s best relievers, and pretty much anyone in the past half decade. Pitchers are flocking to sliders whenever they can get away with throwing one. It used to be a two-strike offering, then an ahead-in-the-count offering, and now many pitchers would rather throw sliders than fastballs when they desperately need to find the zone. Look at that inexorable march higher.

Only, maybe it’s not so inexorable anymore. Between 2015 and 2023, the average increase in slider rate was 0.9 percentage points year-over-year. The lowest increase was half a percentage point; each of the last three years saw increases of a percentage point or more. But from 2023 to 2024, slider rate stagnated. In 2023, 22.2% of all pitches were sliders. In 2024, that number only climbed to 22.3%, the lowest increase since the upward trend started a decade ago.

That’s hardly evidence of the demise of the slider. For one thing, the number is still going up. For another thing, it’s one year. Finally, 2024 marked the highest rate of sliders thrown in major league history. If I showed you the above graph and told you “look, sliders aren’t cool anymore,” you’d be understandably unmoved.

Not to worry, though. It might be January 9, but I won’t try to pass that off as genuine baseball analysis even in the depths of winter. I’ve got a tiny bit more than that. Raw slider rate is a misleading way of considering how pitcher behavior is changing. There are two ways to increase the league-wide slider rate. First, pitchers could adjust their arsenals to use more sliders and fewer other pitches. Second, the population could change – new, slider-dominant pitchers could replace other hurlers who throw the pitch less frequently.

For example, Adam Wainwright retired after the 2023 season. He threw 1,785 pitches that year, and only five were sliders. Plenty of the innings Wainwright filled for the Cardinals went to Andre Pallante, who graduated from the bullpen to the rotation and made 20 starts in 2024. Pallante actually threw fewer sliders proportionally in 2024 than he did in 2023 – but his pitch count ballooned from 1,139 to 1,978. Similarly, Michael McGreevy made his big league debut in 2024 and threw 311 pitches, 19% of which were sliders.

The numbers can lie to you. Pallante, the only one of our three pitchers to appear in both years, lowered his slider rate. But in 2023, Pallante and Wainwright combined for a 7% slider rate. In 2024, Pallante and McGreevy combined for a 17.1% slider rate. That sounds like a huge change in behavior – but it’s actually just a change in population composition.

The story we all think about isn’t Wainwright retiring and handing his innings to McGreevy and Pallante. It’s Brayan Bello going from 17.5% sliders to 28% sliders while pitching a similar innings load – something that also happened in 2024, just so we’re clear.

To measure how existing pitchers are changing their slider usage, we shouldn’t look at the overall rate. We should instead look at the change in each pitcher’s rate. That’s a truer reflection of the question I’m asking, or at least I think it is. And that answer differs from the chart I showed you up at the top of this article.

There were 315 pitchers who threw at least 50 innings in 2023 and 2024, and threw at least one slider in each of those two years. Of those 315 pitchers, 142 increased their slider usage, 24 kept their usage the same, and 149 decreased the rate at which they threw sliders. The story was similar from 2022 to 2023. There were 216 pitchers who fit the criteria in those years; 90 increased their slider usage, 19 kept theirs the same, and 107 decreased the rate at which they used the pitch. From 2021 to 2022, the effect went the other way; 122 pitchers threw sliders more frequently in 2022 than they did in 2021, 22 kept their usage the same, and 74 decreased their usage.

Put that way, the change is quite striking. The slider craze kicked off in earnest in 2017. From 2016-2017, 114 pitchers increased their slider usage and 89 decreased theirs. That rough split persisted in 2017-2018 and 2018-2019. Everything around the 2020 season is a little weird thanks to the abbreviated schedule, but the basic gist – more pitchers increasing slider usage than decreasing slider usage – was true in every pair of years from 2014-2015 through 2021-2022.

That sounds more like a trend than the overall rate of sliders thrown. Graphically, it looks like this:

Let’s put that in plain English. From 2015, the start of the spike in slider usage, through 2022, there were far more pitchers increasing their slider frequency than decreasing it. On average across those years, 1.3 pitchers threw more sliders for every one pitcher who threw fewer. In the past two years, that trend has reversed; more pitchers are reducing their reliance on sliders than increasing it. The population is going to continue to change – they don’t make a lot of Adam Wainwrights these days – but on a per-pitcher basis, the relentless increase in slider usage has halted.

I tried a few other ways of looking at this phenomenon. I held pitcher workloads constant from year one and applied year two slider rates to each pitcher (pitchers who only threw in year one obviously keep their rate unchanged). The same trend held – the last two years have seen a sharp divergence from the boom times of 2015-2022. I looked at the percentage of starters who started using a slider more than some other pitch in their arsenal and compared it to the ones who de-emphasized it; same deal. I also should note that I’ve grouped sweepers and slurves among the sliders for this article, so this reversal is not about pitchers ditching traditional sliders to get in on the sweeper craze.

No matter how you slice it, we’ve seemingly entered a new phase of pitch design. For a while, most pitchers took a hard look at what they were throwing and decided they needed more sliders. Now, though, it appears that we’ve reached an equilibrium point. Some pitchers still want more. Some think they’re throwing enough, or even a hair too many. Now splitters are on the rise, and hybrid cutters are starting to eat into sliders’ market share.

It’s far too early to say that sliders are on the decline. Factually speaking, they’re not. But to me, at least, it’s clear that the last two years are different than the years before them when it comes to the most ubiquitous out pitch in baseball. Sure, everyone has a slider now – but in the same way that four-seam fastballs were inevitable right until sinkers made a comeback, the slider is no longer expanding its dominance among secondary pitches. An exciting conclusion? I’m not sure. But it’s certainly backed by the evidence.

As of the time I’m writing this article, roughly half of our Top 50 free agents have signed new contracts this offseason. That sounds like a great time to take a look at how the market has developed, both for individual players and overall positional archetypes. For example, starting pitchers have been all the rage so far, or so it seems. But does that match up with the data?

I sliced the data up into three groups to get a handle on this: starters, relievers, and position players. I then calculated how far off both I and the crowdsourced predictions were when it came to average annual value and total dollars handed out. You can see here that I came out very slightly ahead of the pack of readers by these metrics, at least so far:

To be fair, none of us have done particularly well. The last two years I’ve run this experiment, I missed by around $1 million in average annual value, and the crowd missed by between $1 and $2 million. Likewise, I’ve missed by roughly $10 million in average annual value per contract, with the crowd around $18 million. This year, the contracts have been longer than I expected, and richer than you readers expected, though you did a much better job on a relative basis when it came to predicting total dollar outlay. We were all low on every category, though, across the board.
Read the rest of this entry »

Sports Info Solutions has been tracking every pitch thrown in Major League Baseball since 2002, and since the beginning, those pitches have been hitting the strike zone less and less frequently. You can check the tumbling year-over-year numbers over on our pitch-level data leaderboard, but if you want to spare yourself a click, I pulled them into the graph below. It paints a damning picture of the command of today’s stuff-over-stamina, throw-it-hard-before-your-elbow-explodes pitchers. Don’t go near this graph if you’re on roller skates:

If you ever feel the need to shake your fist at young pitchers and mutter about loud music and fastball command, this is the graph for you. SIS has documented the percentage of pitches that hit the strike zone dropping from the low 50s to the low 40s over the last 20 years. Combine that with the game’s ever-increasing focus on velocity and stuff, and you’ve got a nice, tidy narrative: today’s pitchers are too focused on throwing hard to know where the hell they’re throwing the ball. However, the truth is a bit more complicated. It’s important to keep in mind that the SIS numbers come from real life human beings who analyze video to track pitches, while the friendly robot that powers Statcast has its definition of the strike zone set in digital stone. Read the rest of this entry »

Right after FanGraphs published my piece on the Kirby Index, the metric’s namesake lost his touch. George Kirby’s trademark command — so reliable that I felt comfortable naming a statistic after him — fell off a cliff. While the walk rate remained under control, the home run rate spiked; he allowed seven home runs in May, all on pitches where he missed his target by a significant margin.

Watching the namesake of my new metric turn mediocre immediately following publication was among the many humbling experiences of publishing this story. Nevertheless, I wanted to revisit the piece. For one, it’s December. And writing the story led me down a fascinating rabbit hole: While I learned that the Kirby Index has its flaws, I also learned a ton about contemporary efforts to quantify pitcher command.

But first, what is the Kirby Index? I found that release angles, in concert with release height and width, almost perfectly predicted the location of a pitch. If these two variables told you almost everything about the location of a pitch, then a measurement of their variation for individual pitchers could theoretically provide novel information about pitcher command.

This got a few people mad on Twitter, including baseball’s eminent physicist Alan Nathan and Greg Rybarczyk, the creator of the “Hit Tracker” and a former member of the Red Sox front office. These two — particularly Rybarczyk — took issue with my use of machine learning to make these predictions, arguing that my use of machine learning suggested I didn’t understand the actual mechanics of why a pitch goes where it goes.

“You’re spot on, Alan,” wrote Rybarczyk. “The amazement that trajectory and launch parameters are strongly associated with where the ball ends up can only come from people who see tracking data as columns of digits rather than measurements of reality that reflect the underlying physics.”

While the tone was a bit much, Rybarczyk had a point. My “amazement” would have been tempered with a more thorough understanding of how Statcast calculates the location where a pitch crosses home plate. After publication, I learned that the nine-parameter fit explains why pitch location could be so powerfully predicted by release angles.

The location of a pitch is derived from the initial velocity, initial release point, and initial acceleration of the pitch in three dimensions. (These are the nine parameters.) Release angles are calculated using initial velocity and initial release point. Because the location of the pitch and the release angle are both derived from the 9P fit, it makes sense that they’d be almost perfectly correlated.

This led to a reasonable critique: If release angles are location information in a different form, why not just apply the same technique of measuring variation on the pitch locations themselves? This is a fair question. But using locations would have undermined the conclusion of that Kirby Index piece — that biomechanical data like release angles could improve the precision of command measurements.

Teams, with their access to KinaTrax data, could create their own version of the Kirby Index, not with implied release angles derived from the nine-parameter fit, but with the position of wrists and arms captured at the moment of release. The Kirby Index piece wasn’t just about creating a new way to measure command; I wanted it to point toward one specific way that the new data revolution in baseball would unfold.

But enough about that. It’s time for the leaderboards. I removed all pitchers with fewer than 500 fastballs. Here are the top 20 in the Kirby Index for the 2024 season:

And here are the bottom 20:

A few takeaways for me: First, I am so grateful Kirby got it together and finished in the top three. Death, taxes, and George Kirby throwing fastballs where he wants. Second, the top and bottom of the leaderboards are satisfying. Cody Bradford throws 89 and lives off his elite command, and Joe Boyle — well, there’s a reason the A’s threw him in as a piece in the Jeffrey Springs trade despite his otherworldly stuff. Third, there are guys on the laggard list — Seth Lugo and Miles Mikolas, in particular — who look out of place.

Mikolas lingered around the bottom of the leaderboards all year, which I found curious. Mikolas, after all, averages just 93 mph on his four-seam fastball; one would imagine such a guy would need to have elite command to remain a viable major league starter, and that league-worst command effectively would be a death sentence. Confusing this further, Mikolas avoided walks better than almost anyone.

Why Mikolas ranked so poorly in the Kirby Index while walking so few hitters could probably be the subject of its own article, but for the purposes of this story, it’s probably enough to say that the Kirby Index misses some things.

An example: Mikolas ranked second among all pitchers in arm angle variation on four-seam fastballs, suggesting that Mikolas is intentionally altering his arm angle from pitch to pitch, likely depending on whether the hitter is left-handed or right-handed. This is just one reason why someone might rank low in the Kirby Index. Another, as I mentioned in the original article, is that a pitcher like Lugo might be aiming at so many different targets that it fools a metric like the Kirby Index.

So: The Kirby Index was a fun exercise, but there are some flaws. What are the alternatives to measuring pitcher command?

Location+

Location+ is the industry standard. The FanGraphs Sabermetric library (an incredible resource, it must be said) does a great job of describing that metric, so I’d encourage you to click this hyperlink for the full description. The short version: Run values are assigned to each location and each pitch type based on the count. Each pitch is graded on the stuff-neutral locations.

Implied location value

Nobody seems particularly satisfied with Location+, including the creators of Location+ themselves. Because each count state and each pitch type uses its own run value map to distribute run value grades, it takes a super long time for the statistic to stabilize, upward of hundreds of pitches. It also isn’t particularly sticky from year to year.

The newest version of Location+, which will debut sometime in the near future, will use a similar logic to PitchProfiler’s command model. Essentially, PitchProfiler calculates a Stuff+ and a Pitching+ for each pitcher, which are set on a run value scale. By subtracting the Stuff+ run value from the Pitching+ run value, the model backs into the value a pitcher gets from their command alone.

Blobs

Whether it’s measuring the standard deviation of release angle proxies or the actual locations of the pitches themselves, this method can be defined as the “blob” method, assessing the cluster tightness of the chosen variable.

Max Bay, now a senior quantitative analyst with the Dodgers, advanced the Kirby Index method by measuring release angle “confidence ellipses,” allowing for a more elegant unification of the vertical and horizontal release angle components.

Miss distance

The central concern with the Kirby Index and all the blob methods, as I stated at the time, is the single target assumption. Ideally, instead of looking at how closely all pitchers are clustered around a single point, each pitch would be evaluated based on how close it finished to the actual target.

But targets are hard to come by. SportsVision started tracking these targets in the mid-2010s, as Eno Sarris outlined in his piece on the state of command research in 2018. These days, Driveline Baseball measures this working alongside Inside Edge. Inside Edge deploys human beings to manually tag the target location for every single pitch. With these data in hand, Driveline can do a couple of things. First, they created a Command+ model, modifying the mean miss distances by accounting for the difficulty of the target and the shape of a pitch.

Using intended zone data, Driveline also shows pitchers where exactly they should aim to account for their miss tendencies. I’m told they will be producing this methodology in a public post soon.

Catcher Targets (Computer Vision)

In a perfect world, computers would replace human beings — wait, let me try that sentence again. It is expensive and time-intensive to manually track targets through video, and so for good reason, miss target data belong to those who are willing to pay the price. Computer vision techniques present the potential to produce the data cheaply and (therefore) democratically.

Carlos Marcano and Dylan Drummey introduced their BaseballCV project in September. (Drummey was hired by the Cubs shortly thereafter.) Joseph Dattoli, the director of player development at the University of Missouri, offered a contribution to the project by demonstrating how computer vision could be used to tag catcher targets. The only limitation, Joseph pointed out, is the computing power required to comb through video of every single pitch.

There are some potential problems with any command measurement dependent on target tracking. Targets aren’t always real targets, more like cues for the pitcher to throw toward that general direction. But Joseph gets around this concern by tracking the catcher’s glove as well as his center of mass, which is less susceptible to these sorts of dekes. Still, there’s a way to go before this method scales into a form where daily leaderboards are accessible.

The Powers method

Absent a raft of public information about actual pitcher targets, there instead can be an effort to simulate them. In their 2023 presentation, “Pitch trajectory density estimation for predicting future outcomes,” Rice professor Scott Powers and his co-author Vicente Iglesias proposed a method to account for the random variation in pitch trajectories, in the process offering a framework for simulating something like a target. (I will likely butcher his methods if I try to summarize them, so I’d encourage you to watch the full presentation if you’re interested.)

The Powers method was modified by Stephen Sutton-Brown at Baseball Prospectus, who used Blake Snell as an example of the way these targeting models can be applied at scale to assess individual pitchers. First, Sutton-Brown fit a model that created a global target for each pitch type, adjusting for the count and handedness of each batter. Then, for each pitcher, this global target was tweaked to account for that pitcher’s tendencies. Using these simulated targets, he calculated their average miss distance, allowing for a separation of the run value of a pitcher’s targets from the run value of their command ability.

“Nothing”

On Twitter, I asked Lance Brozdowski what he saw as the gold standard command metric. He answered “Nothing,” which sums up the problem well. This is a challenging question, and all the existing methods have their flaws.

There are ways that the Kirby Index could be improved, but as far as I can tell, the best way forward for public command metrics is some sort of combination of the final two methods, with active monitoring of the computer vision advancements to see if consistent targets can be established.

But one would imagine the story is completely different on the team side. By marrying the KinaTrax data with miss distance information, these methods could potentially be combined to make some sort of super metric, one that I imagine gets pretty close to measuring the true command ability of major league pitchers. (In a video from Wednesday, Brozdowski reported on some of the potential of these data for measuring and improving command, as well as their limitations.) The public might not be quite there, but as far as I can tell, we’re not that far off.

Editor’s Note: This story has been updated to include Vicente Iglesias as a co-author on the 2023 presentation, “Pitch trajectory density estimation for predicting future outcomes.”

The Winter Meetings always feature trades, but two stood above the fray last week. First, the Guardians traded Andrés Giménez to the Blue Jays in a two-part transaction that briefly left Cleveland with three lefty-hitting first basemen. Then the White Sox traded Garrett Crochet to the Red Sox for four prospects. The best of that group, Kyle Teel, happens to play catcher, the same position as Chicago’s top prospect Edgar Quero. They even have the same future value grade of 50, which is the cutoff for top 100 prospects.

The Guardians made an extra trade to avoid doubling up on similar archetypes, sending Spencer Horwitz to the Pirates for three young pitchers, but the White Sox just kept both catchers. I heard a lot of murmured questioning of that decision as I walked around the Dallas hotel that briefly hosted the center of the baseball universe. But I think both teams were acting rationally, and that worrying about Teel and Quero overlapping is silly. I can’t prove it for you – but I did come up with some data that will hopefully sway your opinion.

Cleveland’s case was straightforward. Steamer projects Horwitz as a 2.5 WAR/600 PA player. It projects Kyle Manzardo as a 1.8 WAR/600 PA player. Josh Naylor? Steamer has him down for 2.4 WAR/600 PA. Three players for two positions — first and DH. (Yes, Horwitz has played second base, too, but he really shouldn’t be a second baseman, and I don’t think the Guardians would’ve used him there.) One of them would ride the bench despite being an above-average contributor, a poor decision for a team that’s trying to maximize its resources. Something had to give.

On the other hand, there are the White Sox. They, too, traded a young star, and the best player they got back plays a position where they already had a similar option. Teel was our 42th-ranked prospect on our updated Top 100 list in 2024, a polished all-around catcher who we expect to reach the majors at some point in the next two years. Quero was our 40th-ranked prospect, and you’re never going to believe this, but he’s a polished all-around catcher who we expect to reach the majors at some point in the next two years.
Read the rest of this entry »

I don’t know the name for this phenomenon, but I’m guessing everyone has experienced it at some point. You hear something enough times, and you start to repeat it without really thinking critically about it. My example: the breakeven stolen base rate. I’ve heard that term so many times over the years, often in connection with whether teams were stealing too much or not enough, that I incorporated it into my thought processes like it was my own.

But then someone asked me why the optimal stolen base success rate was around 70%, and I realized that I’d been wrong. It was a bolt-of-inspiration kind of moment – you only need to hear the counter-argument once to re-assess your old, uncritically assumed thought. Why should teams keep stealing so long as they’re successful more than 70% (ish) of the time? I couldn’t explain it to myself using math.

The other side of the coin, the notion that teams should be successful at far better than the breakeven rate in the aggregate, is incredibly easy to understand. There’s a difference between marginal return and total return. Consider a business where you’re making investments. Your first investment makes $10. Your next one makes $8, and then $6, and so on. You could keep investing until your business breaks even – until you make a negative $10 investment to offset that first one, more or less ($10+$8+$6+$4+$2+$0-$2-$4-$6-$8-$10). But that’s a clearly bad decision. You should stop when your marginal return stops being positive – when an investment returns you $0, you can just stop going and pocket the $30 ($10+$8+$6+$4+$2+$0). Read the rest of this entry »