The Problem With Blaming Analytics
Analytics got both teams to the World Series, and Kevin Cash’s truly bad decision went against a recent analytical trend.
First are foremost, credit is due to the Los Angeles Dodgers. They are arguably the most talented team in baseball, and they earned their 2020 World Series rings.
The turning point in Game 6 that led to the Dodgers securing their championship came in the sixth inning. Through 5 1/3 innings, Rays ace Blake Snell held the Dodgers to just one hit, punching out nine. However, Snell allowed a base hit to Austin Barnes, bringing up the top of the Dodgers order for the third time. With Mookie Betts about to step in as the go-ahead run in a 1–0 game, manager Kevin Cash made the move to pull his big lefty. Snell was visibly shocked and upset as his skipper made his way to the mound.
Nick Anderson relieved Snell, and he immediately allowed a double to Betts, a wild pitch to allow Barnes to score from third, and a ground ball to first base that gave the speedy Betts enough time to score. All of a sudden, it was a 2–1 Dodgers lead.
John Smoltz and the other old school analysts at Fox immediately took the opportunity to pounce on how the move backfired for the analytically-inclined Cash, blaming his reliance on the numbers for the sudden turn of events.
Let’s get this out of the way right from the get-go: both teams made it to this stage because they rely heavily on analytics.
On offense, the Dodgers built a powerful lineup by loading up on hitters who crush the baseball (they led baseball in barrels, hard-hit rate, and average exit velocity).
On pitching, the Rays run-prevention machine consisted of flame-throwing relievers — all of whom threw from different arm slots — and throwing conventional pitching roles out the window. They ranked fifth in hard-hit rate, fourth in barrel rate, third in average exit velocity, and seventh in strikeout percentage. The Dodgers staff is made up of guys who throw gas and induce weak contact (first in average fastball velocity, barrel rate, and hard-hit rate). For goodness sake, they bullpened Game 6, holding the Rays to just one run.
In the field, the Dodgers shifted more than any other team in baseball, and no matter what John Smoltz tries to tell you, shifts work. L.A. held lefty hitters to a .255 wOBA when they shifted and righties to a .302 wOBA. Both the Dodgers and Rays were second in their respective leagues in Defensive Runs Saved.
Both teams have also used analytics to discover undervalued diamonds in the rough and develop players. The Dodgers turned Justin Turner and Max Muncy into All-Star corner infielders by encouraging them to lift the ball in the air. The Rays identified Randy Arozarena and traded for him because they knew that he could hit the baseball with authority. They unlocked a better version of Tyler Glasnow by instructing him to elevate his fastball and tunnel it with curveballs in the dirt.
Another problem with these ex-players ranting about analytics is that they assume the numbers favored pulling Blake Snell for Nick Anderson. Let’s look at what the analytics actually had to say.
Cash’s justification for pulling Snell was that the top of the order was coming up for the third time. The belief that most pitchers become less effective as hitters see them more often in a game is one of the staples of new-school baseball. For his career, Snell’s OPS against increases from .711 the second time through the order to .742 the third time through. In 2020, opponents started crushing him to the tune of a .977 OPS by the second time through.
The argument against removing Snell is that he was locked in and mowing down the Los Angeles lineup, as if the metrics couldn’t detect that. They did. Observe the analytics on Snell’s stuff from Game 6 in comparison to the regular season.
The metrics confirmed that Snell’s stuff was better than usual on Tuesday night. His velocity was up on all of his pitches, his fastball was generating a ton of empty swings, and opponents couldn’t make solid contact against him. Additionally, his 40% CSW rate (called strikes plus whiffs) was higher than any of his previous starts all season.
However, that doesn’t mean that Snell was guaranteed to continue dealing. It simply meant that up to the moment he was pulled, he had been pitching better than he had all season. He had made Mookie Betts look foolish in his first two at-bats, but Betts is really damn good and certainly capable of making adjustments or punishing a mistake.
The final fastball that Snell threw was 94.3 mph. That was his slowest heater of the game, the only one he threw under 94.7, and about two mph slower than his average velocity that night. Perhaps Cash took that as evidence that the stuff was beginning to degrade. Snell at 94–95 is not as overpowering as Snell at 96–97. One bad pitch could have changed the criticism to Cash leaving his starter in for too long. Meanwhile, in 2020, the Tampa Bay bullpen had held hitters to a .643 OPS when facing them the first time in a game.
Personally, my leash for Snell would have been short, but I would have given him one more baserunner. The worst-case scenario is that he allows a home run to Betts, at which point you’re only down by one run and have an even more legitimate reason to remove your starter. The Rays were going to have to score more than one run to defeat a potent Dodger lineup anyway.
The reality is that pulling Snell was not Cash’s true mistake. It was who he replaced him with. The skipper called on Nick Anderson, who was without a doubt his best reliever in 2020 — during the regular season. In the postseason, it was a different story.
Entering Tuesday night, Anderson’s sample sizes from the regular season and the postseason were identical — 58 batters faced — and the results were vastly different. His fastball location was off all October. Instead of keeping it elevated, he was leaving it right over the middle of the plate.
As a result, Anderson was not finding his typical level of success. In the postseason, he had poor location, a lower strikeout rate, a lower whiff rate, a higher xwOBA, and had allowed at least one run in six straight appearances. Guess what? Those are analytics, and they proved that Anderson was not just getting unlucky in his recent outings, but was legitimately throwing the ball poorly. Instead, Cash acted as if he was still in regular-season form, which is ignorant by both the eye test and the metrics.
The merits of pulling his starter so early can be debated, but Cash’s true mistake was going to the wrong reliever. Tampa Bay’s bullpen is deep and effective, but he used an arm that was struggling. Diego Castillo, the team’s next best bullpen arm after Anderson, would have been a much better choice to finish off the inning and work through the heart of the Dodgers order. In this scenario, it would have been much more difficult to argue against removing Snell.
Instead, the Rays manager will have to live with the consequences of his decision, mainly the understandable frustration directed toward him from his best pitcher. Had a different reliever entered and escaped the inning unscathed, Snell would have likely forgiven his manager. Instead, he watched a move that he was highly skeptical of turn out poorly for his team.
Kevin Cash definitely made a poor choice in Game 6, but it was not removing Blake Snell from the game. It was replacing him with a struggling reliever. It’s not fair to use this game as a cautionary tale of “over-relying on analytics” or “letting a computer manage” — these two teams got to the World Series thanks in large part to advanced metrics, and there was data that indicated that Anderson wasn’t right.