Home Sport News ‘Pop’ science — Your guide to learning a new MLB stat

‘Pop’ science — Your guide to learning a new MLB stat

9
0
SHARE


In early March, MLB.com rolled out a new stat: catcher pop times. A combination of two measures (arm strength and the quickness of transfer from catch to throw), pop times use Statcast to time how long a catcher takes to ship a baseball to second base during stolen-base attempts.

A new stat! Throw it into the baseball-stat mix, along with launch angles, Deserved Run Average, expected weighted on-base average, tunnel distance, catch probability, spin axis and literally dozens of other measures that not only didn’t exist 50 years ago, but didn’t exist five or 10 years ago. That are, for that matter, based on understandings of the game that didn’t exist much more than five or 10 years ago.

A baseball fan in previous generations spent time learning players’ stats. Now, that fan has to learn the stats — literally, what they are, what they do, why they exist and how to pronounce them without sounding like a doof. Each year brings new data, better data, new interpretations of this new and better data, a lot of introspection and some skepticism of all this new data, and, ultimately, a lot of new stats. It can be exciting but also intimidating, and sometimes confusing.

It needn’t be. When a new stat comes out, we follow a process to determine not whether it’s a good stat, but how it’s a good stat — how to understand its purpose, how to allow for its limitations, how to appreciate it for exactly what it is and how to make it useful.

We ask five questions.

1. What does it look like?

Last year, the Padres’ young defensive star Austin Hedges had the fastest pop time in baseball, at 1.86 seconds on average, with 90th percentile arm strength and 75th percentile exchange times. Phillies rookie Jorge Alfaro threw the ball the fastest — 89.8 mph, and he’s even faster this year, at 90.1 mph. J.T. Realmuto throws 87.6 mph, the second-fastest, but even catchers with bad arms can be quick if they have fast transfers. Chris Stewart‘s throws don’t crack 80 mph, but he’s got the fastest exchange speed in the game, and his 1.94 pop time was one of the 10 best last year.

Yadier Molina and Salvador Perez were both faster than average last year, as you might have expected, but Gary Sanchez was even faster. Miguel Montero — who got in a feud in Chicago with his pitchers over who deserved blame for stolen bases — had one of the slowest pop times in the game. Tyler Flowers was the slowest. The difference between a 90th percentile pop time and a 10th percentile pop time is only .15 seconds, a blink that makes all the difference.

It looks like this:

2. What does it purport to tell me?

Here’s the official definition: “Pop Time measures the time from the moment the pitch hits the catcher’s mitt to the moment the ball reaches the fielder’s projected receiving point at the center of the base.” It’s how fast it takes the catcher to go from catching a ball to getting it all the way to second base.

MLB.com breaks this metric up from there: the amount of time it took for the catcher to make the exchange (i.e., time elapsed between catching the ball and releasing it) and the speed of the throw to second (which, of course, correlates with the time elapsed between his releasing the ball and the ball’s arrival at second). It also splits the pop times into two buckets: pop times on stolen bases (the catcher failed) and pop times on caught stealings (the catcher succeeded).

A stolen-base attempt is a race between the runner and the pitcher/catcher/middle infielder, so pop time tells us how fast the catcher ran his leg of the race. It measures, very precisely, a part of the game that hinges on hundredths of a second.

Now, the goal of the catcher isn’t to be real fast, but to throw the runner out, so one might argue that this metric is putting a number on something that is related to but not the actual objective. Which is fine: No stat answers every question, and every stat leaves something out.

3. What does it leave out?

At least three things, all of them crucial to passing final judgment on the catcher’s performance. One is how accurate the throw was. It’s impossible to throw a runner out, no matter how fast you are, if the throw is 85 feet above the shortstop. Not only does pop time not answer this, but a catcher’s speed could, in some cases, be an impediment to accuracy. Accuracy and speed might actually be in conflict with each other.

That’s due to the second significant thing this stat leaves out, which is how much time the catcher had. An average major league pop time is 2.0 seconds. If there were a hypothetical catcher whose average pop times were slower than that but who beat the runner by one-hundredth of a second every single time, we might deduce not that he’s a slower-than-average thrower, but that he has a great internal clock and knows exactly what he’s doing. He might be taking his time because he can afford to, and because it helps him throw more accurately.

Indeed, the full pop-time data has a way of confounding analysis. Last season, about two-thirds of catchers were slower on caught stealings than on successful stolen bases. Slower throws were more likely to lead to outs than faster throws. That suggests there’s a lot of this going on: When the catcher knows the runner has a great jump and likely has the base stolen, he puts a premium on quickness — perhaps at the expense of accuracy, in a desperate calculation. When he knows the base stealer has a poor jump or is a slow runner, he makes sure he delivers the ball accurately, at the expense of his pop time. As Orioles catcher Caleb Joseph put it to me this month, “I could go out and throw a 1.79, because I know how to cheat the system” — but, he said, those wouldn’t necessarily be his best, most accurate throws, and hopping out of his crouch early could cost his pitcher a called strike.

Consider, for example, two Russell Martin throws this season. In one, he is recorded as having a 1.93-second pop time, a mark only two catchers were faster than (on average) last season.

In the other, he has a 2.36-second pop time, a time you’d expect to see at a high school game:

But, of course, the 1.93 wasn’t fast enough — Martin hopped out of his crouch early, perhaps a desperate measure to try to catch up to Rajai Davis‘ excellent jump — and the 2.36 was in plenty of time. On a spreadsheet, a 2.36 makes Martin look terrible. But, in context, the 2.36 was Martin performing to his objective rather than to our stat.

The third significant thing left out is where the pitch was. Here’s a truly elite pop time: Realmuto’s 1.80, one of the fastest of the year:

An 88 mph throw! A 0.54-second transfer. Fantastic stuff, something to appreciate. Is it as impressive as this, though?

That’s a 2.33-second pop time, a 1.01-second transfer and an 80 mph throw — but on a pitch in the dirt, on an all-arm throw. It’s incredible. As Joseph put it, miming the challenge of turning a dirty slider into a transfer and throw: “This is the game.”

It’s important to be precise with how we think of these things. “What it leaves out” is not “why it’s bad,” or even “why it’s sometimes wrong.” Rather, it’s “how do we use this as data without being led into a trap?” This question is important for every stat there has ever been. Even WAR, which attempts to include everything that can be measured, leaves out a ton — most notably, everything that can’t be measured.

4. What type of stat is it?

We use one word, but “stat” covers a huge range of measures, with different purposes and different levels of human intervention. Until the 1980s or so, most of us knew three types of stats:

1. Counting stats: hits, strikeouts, outfield assists
2. Average stats: batting average, ERA, fielding percentage.
3. Credit stats: saves, wins, game-winning RBIs

We were aware of but didn’t generally have consistent access to:

4. Objective “scouting” stats: pitcher velocity, runner’s time from home to first
5. Split stats — counting or average stats, chopped up by independent variables: batter vs. pitcher matchup stats, vs. lefties, at home, etc.

And along the way came Bill James, Pete Palmer, Project Scoresheet, online hobbyists, Baseball Prospectus, The Book and Baseball-Reference.com sabermetrics. A lot of what they developed were a more sophisticated version of previous stat types — wOBA, for instance, is a more sophisticated average stat based on a more sophisticated counting stat, and UZR is a more sophisticated fielding percentage with a more sophisticated denominator. Win Probability Added is a more sophisticated credit stat. But they also introduced new types:

6. Context-adjusted stats, such as OPS+ or wRC+, which don’t merely count events, but adjust them based on how much easier or more difficult the context (ballpark, era, level of competition) makes each accomplishment.
7. Predictive stats, such as xBABIP, Fielding Independent Pitching, catch percentage or hard-hit ball rates, which move away from results to estimate how well a player’s performance “should” have turned into value.
8. Player value stats, such as WAR, which take every piece of the chicken, mash it up and reconstitute it into a delicious chicken tender.

And, finally, Statcast, PITCHf/x and commercial scouting services have given us the ability to describe every aspect of player and in-process movement:

9. Neutral descriptive information, like changeup usage, fielder positioning information and release point, all of which might have been observed in previous generations but not measured.
10. Semi-neutral descriptive information, such as horizontal pitch movement, non-fastball pitch velocities and launch-angle descriptions that might have qualitative value in combination with other metrics, or in changes over periods of time.
11. Probably non-neutral descriptive information, such as release-point consistency, spin rate, exit velocity, route efficiency and tunneling. Though we might call these stats more sophisticated, more accurate versions of Category 4 stats: objective scouting information.

Catcher pop times are a Category 4 stat. It’s a classic objective scouting metric, captured by every scout with a stopwatch for decades. Statcast finally gives us access to the scout’s stopwatch and it gives us more precise times, and it does so for almost all throws, not just the ones you or a scout happen to be sitting on. They’re also a Category 10 or 11 stat.

5. Does it tell me who is good? Does it do so better than my own eyes and/or what’s already out there? How should I use it?

Pop times describe how well a catcher physically performed a baseball action — threw to second base — without necessarily showing how effectively he did it. In a lot of ways, it’s comparable to sprint speed, which can’t tell you how good a player is at stealing bases or chasing down fly balls but can definitely tell you the speed he sprints.

Unlike stolen bases, though, there isn’t already a traditional stat that tells you with much certainty how well a catcher throws out baserunners. Pitchers have a bigger influence on the running game than catchers, so a catcher’s caught-stealing percentage is dramatically affected by his pitchers — and, therefore, can be misleading. For that matter, his caught-stealing percentage is dramatically affected by how many baserunners attempt to steal against him. A catcher’s reputation can shut down the running game before the runner ever takes his lead.

A new, more advanced stat at Baseball Prospectus — Throwing Runs — winnows all those variables down until the catcher’s role is isolated, a more carefully thought-out metric for measuring catchers than CS%. It’s the most complete measure, it’s based on actual results and it’s probably the best stat if your question is, “Which catcher added the most value by stopping the running game?”

But pop times can answer questions Throwing Runs can’t, about individual stolen-base attempts, about a catcher’s specific strengths and weaknesses and about his physical tools. If a veteran catcher’s throwing success goes down one season to the next, it might be a blip, a fluke. But if his arm strength goes from 85 mph to 80, it’s probably a physical change. And it appears to be correlated to throwing success. As MLB.com’s Mike Petriello wrote of his site’s new stat,

“Pop time does matter in preventing stolen bases, though it’s also unsurprisingly a pretty noisy relationship. Based on the data, 0.1 seconds of pop time changes the caught stealing rate by 10 percentage points. Only one of the slowest 10 catchers had a caught-stealing rate of 40 percent. Only one of the top nine — Hedges — didn’t. It’s also a skill that correlates pretty well year to year. [Martin] Maldonado, for example, was 1.92 in 2015, then 1.91 in ’16 and 1.93 in ’17. At the other end, someone like Stephen Vogt was at 2.08 in ’15, 2.10 in ’16 and 2.11 in ’17. It’s a combination of two different skills, as you’ll see, but we tend to see the same names at the top each season.”

There’s a pretty convincing match between catchers’ average pop time and how well they rank on Baseball Prospectus’ advanced throwing stats. BP’s top 10 throwing catchers last year all had faster-than-average pop times.

Don’t use it for what it’s not. It’s not the final say on how good a catcher is at stopping the running game. It probably can’t tell you, without any other context, how good or bad any individual throw was. It won’t prove one catcher is better than another. It doesn’t aspire to be any of those things, at least not now.

What it does do is introduce you to a catcher’s physical skills. In combination with your own eyes, or over the course of dozens of throws, it can illuminate the reasons a catcher was successful at stopping the running game and help separate his skills from those of his pitchers. It can undoubtedly tell you who has the strongest arm, if not the most effective one. It’s primarily descriptive but can be paired with other information to help pass judgment. And as we get more data from Statcast, it could end up even answering the bigger questions about catcher performance, once we can pair the pop times with pitcher delivery times, catcher accuracy, infielders’ tag times and baserunners’ jumps and sprint speeds.

For now, it’s a source of data that gives great color (Alfaro’s average throw to second is 90 mph!) and offers some new insight into how the game is played. For instance: The fact that pop times are faster on unsuccessful throws than on successful throws is new to me! It might seem intuitive now that I know, but it’s definitely not something I already was intuiting. Pop times taught me something about catcher behavior. I imagine it won’t be the last time.



Source link

LEAVE A REPLY

Please enter your comment!
Please enter your name here