No sooner had he buttoned up a Dodgers jersey than Yu Darvish changed his release point and realized better results on the mound. I originally didn’t know what to think of the change when I noticed it in November , even suggesting that it might be indicative of fatigue and, as a consequence, a potential injury. But we’ve learned that Darvish’s adjustments were actually part of a conscious effort by the Dodgers to optimize the righty’s pitches.
“The Dodgers have encouraged Darvish to change a number of things since he joined them from Texas – his arm slot, the rhythm of his delivery and his pitch mix,” explained Bill Plunkett of the Los Angeles Daily News.
Wow. That’s somewhat crazy to me. Here was the best team in baseball, who just traded high-value prospects to land Darvish for the stretch run, immediately telling the ace to change his mechanics. Risky? Probably. But, darn, did it work out for the best.
Darvish started blowing his fastball by batters with nearly 60 percent more frequency than he had earlier in the season with the Rangers. That’s, uh, that’s a lot.
He wasn’t throwing harder with LA and he didn’t really change his fastball frequency, either. How, then, was he able to increase the swing-and-miss factor to such a degree all of a sudden?
The answer is in the release point change. Darvish let go of the fastball closer to his body as a Dodger, as illustrated by the graph below. Of note here is how the error bars are wider in August, right after he was traded. This illustrates more inconsistency in the release point, indicating that Darvish was clearly messing around with his mechanics.
So does this different release point actually correlate with increased spin? Yes, it does, as we see in the following heat map. Blue boxes are negative correlations. Pay attention to release_pos_x (horizontal release point), where release points further away from the body are associated with less spin rate. And also look at adj_rel_y (vertical release point), where the vertical release point basically has little effect on spin rate. So this horizontal release point is a big factor for Yu Darvish’s fastball spin rate.
Okay, so does Darvish’s spin rate predict whiffs? Yes.
Darvish’s fastball revolves 2,499 times per minute. For every 30 RPM increase in spin, Darvish has a 30 percent greater likelihood of making the batter swing and miss. HUGE. The odds of this being a false equivalency and that we are actually wrong about this relationship between spin and whiffs are less than 2 percent. Velocity and all relevant variables were controlled for and the prediction was still significant. In short, we can say with a great deal of confidence that if Darvish increases his fastball spin rate, he’ll generate more whiffs.
For every 30 RPM increase in spin, Darvish has a 30 percent greater likelihood of making the batter swing and miss.
What’s weird to me is that spin rate only predicts Darvish’s fastball whiff rate; it has no predictive value with his slider, curve, cutter, or change. That’s probably because Darvish’s slider is so good even if the spin isn’t consistently tight.
With the fastball, though, it’s imperative that he’s throwing it with as much spin as possible. Less spin equals fewer whiffs, period. And if his fastball isn’t playing well, we can probably imagine that his secondary stuff might not be as effective.
What does this mean for the Cubs? To be honest, I’m not entirely sure. What I am sure of is that I’m very much looking forward to monitoring Darvish’s release point once the season starts. Beyond seeing whether and how he can generate more whiffs, I’d love to allay the nagging worry that these “new” mechanics have been associated with injury in the past.
If this release point truly won’t affect Darvish’s health, however, he has a good chance of being much better in a full year with the change.
I adjusted vertical release point by subtracting Darvish’s height minus raw vertical release point, as suggested by Eno Sarris, former FanGraphs writer. I ran a logistic regression on whiffs by including release point, spin, pitch location, and velocity as predictors. I calculated odds ratio for whiffs — one RPM increased chance for a whiff by 0.3 percent (i.e., 100 RPM by 30 percentl p=0.0138). I then ran a linear regression on all variables included in the above correlation matrix. Release point and extension were significant predictors (p<0.001), but horizontal release point explained the most variance (about seven percent). Data was scraped by the baseballR package. Code available upon request.