In the weirdest year of baseball history so far, 2020 featured a gigantic playoff field introduced right as the season began, turning a 10-team postseason into a 16-team format. Changing the basic structure of awarding the sport’s championship with no advance notice would have been an odd choice in a normal season. But given the 102-game reduction in the league’s schedule and its resulting small sample size season, it kind of made sense. When the decision was made, it wasn’t a surety that there would even be a season, to the point that people would have been happy if extra-inning games were decided by closers riding ostriches and jousting.

Before the World Series was even completed, commissioner Rob Manfred expressed the league’s desire to keep the new format in a normal season. The players need to agree to changes like this, of course, and that permission wasn’t granted after all MLB offered in exchange was a universal designated hitter. One of the concerns, not officially made public, is that a playoff system that is more of crapshoot will further reduce the already eroded incentives for teams to spend money to improve their rosters. That’s hardly a shock; at the 2020 trade deadline, 10 teams already had projected playoff probabilities above 97%. Combine that with the absence of the normal advantages afforded to higher seeds, and you had a trade deadline that saw only a single team, the San Diego Padres, aggressively improve, and even their moves were almost certainly made with an eye toward 2021 and beyond.

So what is the ideal playoff system? That’s a difficult question, one that’s impossible to answer to everyone’s satisfaction. I can only answer for myself, and for me, there are a few requirements that are particularly important. Basically, I want a system in which regular-season performance matters, thus maintaining one of the core aspects of the game. I also want a playoff system that more heavily awards quality over randomness without making the result a preordained one. The more a championship is decided by randomness, the less incentive there is for teams to innovate and invest. Read the rest of this entry »

This story starts, as all good stories do, with me recounting the time one of my coworkers and I discussed something. Okay, fine, very few good stories start that way — almost none, in fact — but bear with me. This (non-baseball) coworker, someone who I consider very bright and very interested in baseball, told me he didn’t really believe in wRC+, even after I’d shown him some articles describing it.

Why, I wondered, didn’t he believe in it? It’s so elegant! The math is right there! How can you not like something that wraps up performance at the plate in a single number? No need to compare apples to oranges — you can juice everything to a pulp and simply count calories. His answer was simple: it doesn’t consider batting order.

“You’re telling me,” he said, “that you’d rather have Mitch Moreland as a leadoff hitter than Xander Bogaerts?” It was 2017, and we were working in the Northeast, which explains why both players were Red Sox and why this question was even close. “His wRC+ is higher, but he’d be worse at leadoff. He doesn’t get on base enough.”

To be honest, it’s a compelling argument. I didn’t really have the intellectual tools or the time to counter it. I went with the old tried and true method: I vaguely mentioned something about context-neutrality in the long run, said I had some bonds to arbitrage or whatnot, and went back to work, ending the conversation without conceding defeat.

Fast forward to today, and I still don’t have a wonderful answer to my former co-worker’s point. I do have a computer program that simulates games, though, so I decided to come up with a quick and dirty check. What if we plugged real hitters with similar one-number batting statistics but who get there in wildly different ways into the lineup? Would we learn anything? Would I be able to write 1,500 words about it and entertain the masses? I guess we’ll find out! Read the rest of this entry »

You know it’s almost time for baseball season when all of the major projection systems forecast Kyle Hendricks‘ ERA one run per nine innings too high.

As much as this sounds like a knock on those who develop projections, it’s not. What Jared Cross (Steamer), Dan Szymborski (ZiPS), Derek Carty (THE BAT), and the folks at Baseball Prospectus (PECOTA) do is no small feat. If I weren’t too cowardly to even try to create my own projection system, I would be too stupid to design one that is half as effective as theirs. Glass houses and all that.

That said, I am just smart enough to know that projected ERAs ranging from 3.84 to 4.42 for Hendricks, who boasts a career ERA of 3.12 and has never finished a season with an ERA above 3.46 (except that dastardly 3.95 ERA in 2015), are too high. It’s easy to poke holes in the obvious outliers, but projections succeed by describing and then predicting the talents of most pitchers, not the ones whose talents deviate dramatically from expectation. Hendricks is every projection system’s known blind spot.

It’s not just projections that struggle with Hendricks, either. We, the sabermetric community, frequently use ERA estimators as shorthand to characterize a pitcher’s talent level. If you frequent FanGraphs, you’re familiar with Fielding Independent Pitching (FIP), expected FIP (xFIP), and Skill-Interactive ERA (SIERA). By virtue of how they’re constructed, each metric makes assumptions about the skills a pitcher theoretically “owns”:

• FIP: strikeouts, walks, and home runs allowed
• xFIP: strikeouts, walks, and fly balls induced
• SIERA: a complicated combination of strikeouts, walks, net groundballs (groundballs minus fly balls), and their squared terms and interactions with one another

While each estimator features a batted ball component, they focus on trajectory (launch angle), not on authority (exit velocity). This is a fair assumption, frankly. I have illustrated how a pitcher can influence hitter launch angle, operating under the assumption they bear little to no influence over hitter exit velocity. It’s not quite that bleak; certified baseball genius Rob Arthur found that the average pitcher’s effect on a baseball’s exit velocity: roughly five parts hitter, one part pitcher. Read the rest of this entry »

If you follow the work of Alex Chamberlain at all, you’ve heard of the value of launch angle consistency. I’m not going to recapitulate his body of work on the subject, but briefly: hitters with tighter launch angle distributions routinely run higher BABIPs, and you can think of launch angle consistency as roughly a proxy for “hit tool.”

Most of this comes down to avoiding terrible batted ball outcomes. The two worst things you can do when you put the ball in play are to hit it straight down or straight up. Given that balls are, on average, hit mostly forward and with a tiny bit of loft — breaking news, I know — launch angle consistency is a great proxy for how often you avoid those, because the more -80s and +80s you put in your sample of mostly 10s and 20s, the higher the standard deviation gets.

One thing I’ve often wondered is whether this idea of consistency holds up for subsets of batted balls. Intuitively, it seems like it might. Take hard-hit balls, for example. If you’re hitting the ball 95 mph or harder, you really don’t want to squander it by hitting the ball on the ground or straight into the air. The distribution peaks at 30 degrees, but anything between 10 and 35 is a solid outcome.

With this in mind, I decided to look for batters who grouped their hard-hit balls most tightly. Having a narrow distribution seems like a great way to maximize good outcomes. Which player, you ask, has the tightest launch angle consistency (I’m just using standard deviation here) on hard-hit balls? I’m glad you asked — it’s Dee Strange-Gordon. Read the rest of this entry »

Hey there! I want to give you a heads up about this article, because it doesn’t fit into a normal genre I write. Today, I won’t be telling you some new insight about a player you like, or creating some new nonsense statistic that tries to pull meaning from noise. This is a story about how baseball analysis is changing right before our eyes. A group of scientists and baseball thinkers are redefining the way we think about pitch movement, and I think it’s worth highlighting even if I don’t have anything to add to the conversation yet, because this new avenue of research is going to be front and center in Statcast-based analysis over the next few years.

“Seam-shifted wake,” as Andrew Smith, a student of Dr. Barton Smith (no relation) coined it, is a source of pitch movement that the first attempts at understanding the physics of a pitched baseball overlooked. It has already changed the way that coaches and pitchers approach pitch design, and due to recent data advances, it’s about to be everywhere. So let’s go over how we got here, to this newly observable way that pitchers deceive hitters, by starting at the beginning and working forward.

At its core, baseball is a game about one person trying to throw a ball past another person. There are other trappings — bases and baserunners, umpires, a strike zone, the mythology of Babe Ruth, and a million other sundry things. At the end of the day, though, everything starts with the pitcher trying to throw a ball past the batter.

Accordingly, baseball analysis over the years has focused on describing the flight of that ball. For a time, that simply meant describing the shape of pitches — they don’t call them curveballs for nothing. The next step was velocity — radar guns let us appreciate fastballs numerically rather than merely aesthetically.

In the past 15 years, the amount and scope of pitch-level analytical data has exploded. First, PITCHf/x quantified pitch location and movement. When we report a pitcher’s chase rate or how often a batter swings at pitches in the strike zone, it’s because the location where each pitch crosses the plate is recorded and logged. When we say a pitcher has eight inches of horizontal break on their slider, it’s because new technology allows us to measure it.

When Statcast debuted in 2015, it added another wrinkle: radar tracked the spin rate of each pitch in flight, putting a numerical value on something that had previously been only qualitative; a pitcher’s ability to generate movement through spin. Doctor Alan Nathan has written several authoritative studies discussing the value of this spin data. Read the rest of this entry »

Planning a starting rotation for 2021 carries innumerable pitfalls. Nearly every pitcher in the league saw a reduced workload last year, and they did it in strange circumstances to boot. It’s not merely that the short season set everyone’s innings back — though that’s a huge component. A large number of cancelations and postponements also meant more doubleheaders and more cobbled-together games, another way to throw pitchers off their rhythm.

Put it all together, and protecting arms sounds like an appealing plan for 2021. The Mariners announced that they’ll use a six-man rotation next year, a continuation of the plan they leaned on for all of 2020. The Red Sox are talking workload management. Since initially publishing this piece, Jeff Zimmerman pointed out that the Tigers will use a six-man rotation as well. Is an embiggened rotation the solution to this universal problem? Let’s do the math.

It depends, first of all, on what you give up. The innings tradeoff of a six-man rotation is straightforward. Giving your pitchers an extra day off limits their workloads, naturally enough. The math on that is straightforward if you assume it doesn’t affect their in-game workload. Take a pitcher who averages six innings per start. In a five-man rotation, that’s 192 innings of work. Adding a sixth pitcher to the rotation cuts that down to 162 innings.

How much do those 30 innings of work matter when it comes to health? I’ll level with you — I’m not sure. We simply don’t have the data to say with any amount of certainty, because the number of comparable situations is so small. Pitchers have light workloads all the time, but in most cases it’s due to age or injury. Looking at what a 21-year-old pitcher did in first building up stamina probably can’t tell us much about how many innings Jake Odorizzi, to pick a random example, should throw in 2021. Likewise, a pitcher’s workload in his first year back from Tommy John surgery can’t tell us how many innings Trevor Bauer can be effective for. Read the rest of this entry »

Consider the plight of the base stealer. In the 1980s, their role was sacrosanct. Get on base first, then cause havoc. For fans of speed and baserunning, it was a veritable golden age. Rickey Henderson and Vince Coleman each stole 100 bases in three separate seasons. Since 1980, the 13 top seasons in terms of stolen bases per plate appearance were 1980 through 1992.

Alas, the scurrilous forces of math and efficiency conspired to dethrone the stolen base. As it turns out, advancing one base is less good than creating an out is bad. It’s bad enough, in fact, that you need about three successful stolen bases to make up for the downside of getting caught once. The very best thieves managed that level of efficiency, but in aggregate, the league only crested a 70% success rate once from 1980 to 1992. Steals simply weren’t advancing teams’ goal of scoring as many runs as possible.

For a time, there was a reasonable counter-argument: what if attempting a stolen base has positive value that isn’t solely contained in reaching second base? Perhaps the pitcher has steals on the brain, or the defense loses its cohesion while attempting to cover the base for a throw. It doesn’t need to add much edge to make the math add up.

In 2007, the authors of The Book took up this question. They found a large advantage to batters when a runner was on first — exactly what proponents of steals suggested. There was a big problem, however. That advantage was for all runners on first base. The faster the runner, the smaller the advantage. In addition, actually attempting a steal carried a huge hit to the batter, more than enough to offset the advantage of having a runner on base. Read the rest of this entry »

A couple of weeks ago, I was chatting with PitcherList’s Alex Fast about four-seam fastballs swinging strike rates (SwStr%) and their relationship to pitch height — or, perhaps more specifically, their lack of relationship. At the pitcher-season level (e.g., “2020 Clayton Kershaw“), the correlation between SwStr% and pitch height appeared weak at best. When you consider that no fastball is created equal and then introduce small-sample variance to the equation, the relationship could, understandably, become blurred at the pitcher level.

As a retort, I sent him the following graph, which shows SwStr% by pitch height for the three broad pitch classes as defined by Statcast, the source of the data. For reference, I’ve added black lines to indicate the average bottom, heart, and top of the strike zone:

If we zoom out and consider the question at the macro level, independent of context (what’s the average swinging strike rate for all fastballs by pitch height?), we can see that fastballs generate more swinging strikes up in the zone, a phenomenon our own Jeff Zimmerman touched upon here. This finding is mildly interesting in and of itself. But as I considered the matter further, the importance of swing frequency (Swing%) to SwStr% became clear (both use all pitches as a denominator). Regardless of efficacy, more swings will afford more chances for swinging strikes. As such, I anticipated that fastballs probably induce more swinging strikes up high than down low simply because hitters swing more frequently at high fastballs. Similarly (but inversely), non-fastballs would generate more swinging strikes down low instead of up high. The next graph all but affirmed my intuition:

Although the peaks of the bell curves cluster near the heart of the zone, we can see distinct differences in swing rate by pitch class at the thresholds of the strike zone. At its bottom edge, hitters are half as likely to swing at fastballs as they are at non-fastballs; at its top edge, twice as likely. Read the rest of this entry »

I didn’t want to write this article. One of my favorite things to do, back when I was a full-time Cardinals fan and part-time writer, was wait for the first few weeks of the season and then start ogling velocity changes. There’s almost nothing that made me feel so unabashedly happy as seeing an extra tick or two out of some arm I’d written off the previous year. Why spoil that magic by looking into whether it actually matters?

Nothing fun can come of using data to look at incuriously held beliefs, but that’s never stopped me before, so I decided to examine pitchers who experienced velocity gains from one year to the next. Do their fastballs grade out better? Do they strike out more batters? Walk more? Do they hold the gains from one year to the next? I had no clue, but I decided to find out.

First things first: 2020 goes right out the window. The season started in late July, and no one had anything approaching their normal offseason routine. Temperatures were weird, workloads were changed on the fly, and some teams were affected by COVID-related cancelations; trying to tease something out from that noise is pointless and unnecessary. I’ll just use 2015 through 2019 instead.

Why 2015? That’s when Statcast first arrived, and with it a new tracking system. I could, I suppose, use data since 2008, but I wanted to minimize the chances of false readings stemming from the change in systems. 2020 also featured a change — to camera-based readings instead of radar — but we’re already throwing it out anyway, so no big deal there.

In each year, I looked at the population of starters who threw at least 500 four-seam fastballs. I then found the year-to-year changes for each pitcher-season combination. Justin Verlander, as an example, averaged 93.4 mph in 2015, 94.1 mph in 2016, 95.3 mph in 2017, 95 mph in 2018, and 94.6 mph in 2019. That means his ‘15-’16 change was 0.7 mph, his ‘16-’17 change was 1.2 mph, his ‘17-’18 change was -0.3 mph, and his ‘18-’19 change was -0.4 mph. This gave us a database of 236 pitcher-seasons from 2016 to 2018 — I’m leaving out changes between 2018 and 2019 because I want to know what happens the year after a pitcher gains velocity. Read the rest of this entry »

Before we get going, a heads up: this article is about a mathematical model that explains tanking. I’ll give you a preview. The way baseball is currently set up, it’s no surprise teams tank. If you want to fix that, the game’s competitive structure must change. I suggest a few ways to accomplish that change at the end of the article, but be warned: the bulk of this is a no-nonsense dissection of why teams keep tanking even as the returns go down. Personally, I think the game should make the changes I suggest, because the boom and bust cycle of team contention makes for a fraught fan experience. The way the game is set up now, however, it’s no surprise that teams do it.

The logic behind tanking is straightforward and solid. Being in the middle is the worst; there are no prizes for winning 85 games, flags fly forever, and so on and so forth. You’ve surely heard it enough times that you don’t need a repeat, but just for completeness’s sake, we’ll do it one more time.

By trading present concerns for future value, you make your team better in the future. As a byproduct of trading present concerns away, your major league roster gets bad — bad enough, hopefully, that you’ll move to the top of the draft. Additionally, with no pesky need to be competitive, you can use your major league roster to give borderline players extended tryouts. Hit on a few of them, and that’s even more tailwind for the future.

In essence, tanking is making a bet that taking a step back now will let you take two steps forward sometime in the future. Even if that isn’t the case, being quite bad for a while and then quite good for a while sure sounds better than being mediocre the whole time. Tanking works on both axes, which explains its continued appeal. Do you think your team will win something like 77 wins? Blow it up! 75 wins? Blow it up! 70 wins? You guessed it. Read the rest of this entry »