The Ball Threatens to Overshadow Baseball

With the postponement of Wednesday’s scheduled ALCS Game 4, the de-juiced baseball remained a hot topic of discussion, particularly in New York, where the Yankees appeared to catch a game-changing bad break on Didi Gregorius‘ fifth-inning fly ball that instead of becoming a three-run homer that would have swung the lead in their favor, was caught at the warning track. Players and managers are talking about it, fans are talking about it, analysts are talking about it — here in New York City, even on the 10 pm local news. Like it or not, it’s an issue that won’t go away, in part because of MLB’s stubborn insistence that nothing has changed, even in the face of evidence to the contrary, as well as the reality that the league actually owns about 25% of Rawlings, the manufacturer of the baseball (the other 75% is owned by Seidler Equity Partners, founded by Peter Seidler, the leading investor of the Padres). While none of this invalidates what the players on the field are accomplishing, everybody is suddenly playing or watching a very different game than we’d grown accustomed to during the regular season.

On Saturday, our own Craig Edwards reported that Cardinals manager Mike Shildt said that his team’s analytical department had noted that fly balls are traveling four-and-a-half feet less far than normal. ESPN’s Jeff Passan reported on Wednesday that officials from two unnamed teams (quite possibly those of the Astros and Yankees, since Passan was reporting from New York) “concurred that whatever batch of balls has been used during October is not performing the way the ones in the regular-season did.”

The change concerns the drag on the baseball, which according to the work of Baseball Prospectus’ Rob Arthur has increased sharply relative to the regular season, strongly suggesting that different balls are being used. Arthur estimated that 50% more homers would have been hit if the regular season ball were being used. On Thursday, Arthur published a new piece showing that the postseason baseballs are also affecting pitchers, producing slightly less vertical break (by about 0.4 inches) on curveballs, and that “Sliders are cutting across the zone a little less; sinkers staying a little more buoyant… [The changes] all seem to hover on the border between just large enough that baseball’s tracking system can detect them and just small enough that a major-league hitter probably wouldn’t care.”

I don’t doubt Arthur’s math or modeling, but I do wonder if certain pitchers might feel differently, particularly when he writes this:

These results may also suggest that previous episodes of juicing (2017) and de-juicing (2018) the baseball didn’t do much to pitchers either. Perhaps whatever way the baseball’s drag is varying simply doesn’t affect a pitch grip very much. Maybe it’s only a few thousandths of an inch of seam height, or a very slightly tweaked surface texture–these kinds of alterations, microscopic though they may be, might have been enough to propel massive spikes and falls in the league’s home run rate.

As Dr. Meredith Wills noted in connection with her findings regarding the 2019 regular season baseball’s construction, we’ve heard numerous complaints from pitchers such as Sean Doolittle, Jon Lester, and Noah Syndergaard regarding the surfaces of the balls, and stories such as the one about Masahiro Tanaka struggling with his split-finger fastball due at least in part to the lower seams. That said, those pitchers’ complaints probably have more to do with their overall command of their respective pitches than the modest effects of reduced or increased drag.

Arthur’s latest piece also updated the expected home run tally; “An improved model of batted balls that takes into account temperature as well as exit velocity, launch angle, and park effects shows the same takeaway as before. There have been 64 home runs in the playoffs versus a total of 100 the algorithm predicted.”

My own less rigorous research showed that even in the face of Arthur’s findings, average fly ball distances and exit velocities had increased relative to the regular season. After a closer look at that, keyed by a dialogue between the two of us as well as a larger sample size of fly balls, it appears that my distance issue has come out in the wash.

Through games of October 9, fly balls had averaged 328 feet, up from the major league average of 324 feet during the regular season. However, the latter average probably isn’t the most apt figure to use for comparison, particularly considering that only one team from the bottom half of the rankings (the 18th-ranked A’s, who lost the AL Wild Card game) made the postseason. The fly balls of the 10 teams that reached October averaged 327 feet during the regular season. Weighting their regular season fly ball contributions by the actual number of fly balls they’ve hit in October, we would expect an average distance of 326 feet thus far, but instead, with another week’s worth of games underfoot, they’ve averaged 324 feet, about 0.7% less than expected.

We can even take this a little bit further by considering the fact that the playoffs feature the teams that are also better at suppressing fly ball distances. In this case, nine of the 10 teams ranked among the majors’ top 21 (the NL Wild Card-losing Brewers were 27th), and those 10 teams allowed fly balls averaging just 322 feet. Weighting their regular season fly balls by the actual number of fly balls they’ve allowed in October yields more or less the same figure. If we can infer based upon Arthur’s finding that hitters have about five times the influence over exit velocities as pitchers that the same is true for distance (Arthur hasn’t specifically studied that but said he expected it to be true), we would expect an average distance of 325 feet; the drop is about 0.5% from what’s expected.

While those figures don’t exactly jibe what Shildt said, it’s a safe bet that teams have more data to bring to bear with regards to direction, temperature, and their own models. As for exit velocities, the playoff teams are similarly concentrated away from the bad ends of both the hitting and pitching leaderboards. The 10 playoff teams, which all rank among the majors’ top 19 in batter exit velocity on fly balls, averaged 92.6 mph off the bat for such drives, compared to the overall major league average of 92.0. Weighted by teams’ number of postseason fly balls hit, we’d expect an average of 92.5 mph, while we’ve gotten 93.0 mph. Bringing the pitchers into this, the 10 playoff teams ranked among the majors’ top 20 in pitcher exit velo on flies, averaging 91.7 mph. Weighted by teams’ number of postseason fly balls allowed, that drops to 91.6 mph, and if we again use the 5:1 ratio, we get an expected exit velo of 92.3 mph, 0.7 mph less than what we’ve actually seen. So that sign is still pointing in the wrong direction relative to Rob’s model, but the gap is now less than half of the 1.5 mph difference from my previous piece.

I want to circle back to the Gregorius fly ball, as well as another from the same game that had players talking, a fourth-inning drive by the Astros’ Martín Maldonado that — based upon the conditioning we’ve all received by intently watching a season in which a record number of home runs were hit — looked like it was going out, but wound up a double off the left-center wall.

Recall the situation: with the Yankees trailing 2-0 in the fifth but with two on and two outs — the fourth time in five innings that they had at least one runner in scoring position against Gerrit Cole — Gregorius hit a 98.6 mph fastball from Cole that at first looked to be yet another short porch special given the cozy dimensions of Yankee Stadium’s right field, which is 314 feet down the line. The drive momentarily elevated the hopes of the Yankees and their fans but it wound up settling into right fielder Josh Reddick’s glove just in front of the wall for the third out of the inning. The Yankees would never get a better opportunity, and lost 4-1. Here, again, is the drive, which according to Statcast (Baseball Savant’s search feature, rather than the in-game feed, which produced slightly different estimates) left the bat at 101.4 mph, with a launch angle of 41 degrees and traveled an estimated 342 feet.

As Alex Bregman told reporters after the game, Maldonado had two drives to left field that similarly raised the hopes of the Astros’ dugout, the aforementioned double and a sixth-inning fly out. The former was aided by what might charitably be termed an adventurous route to the ball — whether due to the wind, changed expectations caused by the differing postseason ball, or simple human error — by left fielder Brett Gardner:

Maldonado’s drive, which ultimately went for naught as far as the scoring was concerned, was measured at 101.5 mph off the bat, with a 36.4 degree angle, and an estimated distance of 378 feet.

Rather than cite the percentage of drives with similar specs that wind up home runs — a rate that obviously can be affected by the direction of the ball, since the distances down the foul lines are much shorter than those to center field — I drilled down into Baseball Savant to get a better idea of the average distance of such drives, bracketing them somewhat so as to increase the sample sizes and produce what is hopefully a truer read of how far we could have expected the balls to travel.

Let’s start with Gregorius’ drive, considering only fly balls with similar Statcast specs:

The Savant search interface can be daunting, and it has some quirks. For one, when you’re specifying the metric range, you can’t just set it to equal a single number; you need to use greater than or equal to one number (or less than or equal) or a range of two numbers. For a metric such as exit velo, which is generally cited to the first decimal, one can use the two nearest whole numbers. Launch angle is generally reported as a whole number, though in Savant queries, decimals exist. To enlarge the samples, I went one degree in either direction, and then combined them. I did this for both 2019 data and for 2015-19 data, and as you can see, the end results weren’t terribly different; the two average distances are 11 and 12 feet longer than Gregorius’ drive.

We can improve upon this, obviously at the expense of sample size, by limiting the search to fly balls that were pulled, either by righties to left field or, as was the case with Gregorius, lefties to right field:

Ignore that last line for a moment. What we see here is that pulled fly balls in this range do travel farther, by an average of eight feet using 2019 data (albeit a pretty small selection of drives) and by an average of three feet using 2015-19 data — all of which suggests the likelihood that Gregorius got hosed.

For that last line, I’ve used a metric hidden within the dropdown menus, launch direction, which measures the degrees to the left (negative) or right (positive) of the axis running from home plate to second base. Through an iterative process — which basically means I kept banging the keyboard until I found something — I bracketed Gregorius’ launch direction at somewhere between 23 and 24 degrees. As it turns out, over the past five seasons just three balls have been hit to similar specs in the regular season, which clearly won’t do for a sample. Widening to 22-25 degrees gives us six drives, going from 21-26 degrees gives us 10, and from 20-27 gives us 11, ranging from 338 feet to 374, with an average of 353 feet. Even with that wider range, Gregorius’ ball is the only similar one from the postseason.

While each of those searches leads to a similar conclusion — that Gregorius’ drive traveled significantly less far than expected — the same isn’t true for Maldonado’s drive, which was also pulled (with a direction between -17 and -18 degrees, which is to say to the left of second base).

As the Savant results put the launch angle at 36.4 degrees (compared to Gregorius’ 40.9 degrees), it seems acceptable to go with a narrower range in the search. All of the methods I used produced larger sample sizes than for Gregorius, and only a couple of them produced average distances longer than his drive. In other words, it’s less clear that Maldonado lost out on a home run relative to what we could have expected during the regular season, but that’s not to say that the Astros haven’t occasionally gotten the short end on fly balls in other games.

Given the home run-dependent nature of scoring in the postseason — 41.4% of this October’s runs have scored via the longball, slightly ahead of the 1995-2018 average of 39.6% — and our increased understanding of the changed baseball’s influence on home run rates, it’s inevitable that all of this remains a topic of discussion while the spotlight is shining most brightly. That shouldn’t detract from our enjoyment of what the likes of Gerrit Cole, Max Scherzer, and others are doing, but so long as MLB fumfers its way around the issue, the consequences of their lack of oversight and/or vision won’t go away.