April 3, 2013
Dissecting Darvish's Opening Day
Four Reactions to Tuesday's Almost-Perfect Outing
Yesterday on Effectively Wild, Sam Miller and I talked about Opening Day and the power of first impressions. By the beginning of April, we’ve been so long without baseball that the first performance by a player or team assumes a significance out of proportion to its actual import. If we’ve spent the offseason dreaming on a player doing well, or fearing that he’ll fail, we’re more liable to take it as confirmation of our hopes or fears if that player performs as we expect on Opening Day than we would be if he did the same thing on, say, August 13th.
Tuesday was Opening Day for Yu Darvish, and we had high hopes. The Rangers ace recorded the second-highest strikeout rate among AL starters last season, but he also had the third-highest walk rate and struggled with mechanical consistency throughout the season. As Doug Thorburn detailed last week, Darvish’s mechanics came together down the stretch, and he also stopped nibbling and started throwing more cutters. With an evolved approach, he ended the year on an eight-start run of 66 strikeouts and 10 walks over 57 1/3 innings. Darvish is hard enough to hit when he can’t throw strikes; with good command and control, he might be as close as pitchers come to unbeatable. Both PECOTA and the Baseball Prospectus staff picked him to be the AL’s second-best pitcher in 2013, and with a weak opponent in the Astros, we expected to see a dominant Darvish last night. He didn’t disappoint.
Darvish struck out 14 batters, a career high. He walked none, which he’d done only once before (last year’s Wild Card loss to Baltimore). He fanned Carlos Pena in the fourth on a 98.3-mph fastball, which was 0.3 mph faster than the recorded speed of any regular-season pitch he threw last year, and he threw a slider to Jason Castro in the second that had a fraction of an inch more movement than any of his 2012 offerings. Darvish had no trouble missing bats as a rookie—of the 48 pitchers who threw at least 3000 pitches in 2012, Darvish’s 29.3 percent whiff rate ranked first—but he blew by his previous single-game high of 19 swinging strikes with 27 against Houston. He looks like he’s leveled up since last year.
How much of that was his opponent? The Astros lineup Darvish faced, weighted by plate appearances, had a .252 projected TAv. If Darvish had gotten Marwin Gonzalez out to end the game, it would have been the fourth-worst of 20 lineups to be perfect gamed since 1950. (Surprisingly, lineups in perfect games have tended to be pretty good—only the 1998 Twins, 1981 Blue Jays, and last year’s Astros and Mariners were below league average, although overall productivity isn’t the best way to assess the likelihood of a lineup going a game without a baserunner). It’s a bad offensive team, and while the odds are against its being no-hit at some point this season, it’ll probably come close another time or two.
So we’re going to have to wait to see how Darvish does against better teams, and whether he can maintain his mechanics all summer. But considering how he finished last season and how well he’s thrown this spring—his velocity ticked up in spring training, too, even more impressive given that most pitchers thrower slower early on—expect more outings like last night’s ahead. —Ben Lindbergh
Given Darvish’s ultra-wide arsenal, it’s always interesting to study what he’s throwing at any given time. At the start of last season, he attacked hitters with a heavy dosage of four-seam fastballs while using the slider and cutter as his main secondary offerings. But as the campaign progressed, Darvish continued running into trouble with walks. He chose to simplify things as a result, eliminating the hitch from his delivery and narrowing his arsenal. Darvish began going with what he could command best––his four-seamer, cutter, and slow curveball. Unless he was ahead in the count, his 80-grade slider took a back seat simply because he had trouble throwing it for strikes.
Between spring training and Tuesday’s masterful start against Houston, Darvish once again appears to be taking a different approach with his repertoire. Of his 111 pitches on Tuesday, he threw just 19 four-seam fastballs, averaging 95.4 mph on the radar gun. As one scout in attendance commented, “The most impressive thing was that he had 94-97 whenever he wanted, but he didn’t even need to use it.”
Darvish threw 46 cutters (out of those 111 pitches) against the lefty-heavy Astros lineup, using it to get under the hands of left-handed hitters. He mixed in a whopping 32 sliders, as well. But the key to his success remains four-seam fastball and cutter command. While Darvish will show his slider at 0-0 or behind in the count on occasion, he’s unlikely to do it often. He was getting ahead with his fastballs constantly on Tuesday, and that enabled him to use his slider as a wipeout offering. A Darvish that’s constantly working with two-strike counts is a scary Darvish given the extreme nature of his repertoire––and the slider in particular. Houston’s hitters learned that the hard way last night. —Jason Cole
Darvish, who hadn’t thrown more than 78 pitches in any of his spring training outings and who would have been yanked even with a no-hitter still in place, even admitted himself that he was tiring in the ninth inning of his near-perfect game.
So when he went out to start the ninth in Houston having thrown 107 pitches, there was an interesting element of game theory at play.
Darvish recorded 12 of his 26 outs on pitches out of the strike zone, so at that point, it was 12 out of 24. Here was his PITCHf/x chart from innings 1-8.
Darvish was tiring, though, and the Astros knew that he was tiring or at least that he probably wasn’t prepared to go really deep into a game yet. So there were two extreme approaches to consider.
There was the chance strictly to work his pitch count. Darvish is a pitcher whom they may have to see four or five more times this year as new division rivals, and one whose routine and well-being they wouldn’t have minded throwing off a bit.
Or there was what they did, which was figure that at 107 pitches, Darvish would be trying to get back to the dugout as quickly as possible, so start swinging away. When Darvish obliged with his own get-me-outta-here approach, the Astros made contact on all three swings and eventually got the hit.
Here’s Darvish’s PITCHf/x chart from the ninth inning, in which all four pitches were fastballs.
Jason Castro took strike one, and he was the only batter that inning to watch a pitch before grounding out on the second one he saw. Carlos Corporan grounded out on the first pitch, and Gonzalez got the 91-mph pitch denoted in light blue and smacked it between Darvish’s legs for the game’s first hit.
The path that Darvish took was the exact one that the Astros appeared to anticipate, and while the offense’s actions may have saved a rival from a tough decision or what they thought was an outing of dangerous length, at least they’ve still never been no-hit in Houston in their going-on-52-year history. —Zachary Levine
Darvish produces some fun PITCHf/x data to work with. And by “fun” I mean “hellacious.”
His strikeout of Chris Carter in the eighth seemed to feature a variety of looks on the slider—intentionally or otherwise. Let's compare the action in PITCHf/x. The numbers indicating the vertical and horizontal movement are larger than those MLBAM typically shows as “pfx_z” and “pfx_x,” since our method of measuring movement examines a longer slice of the ball's flight and also includes the effect of gravity. While this approach may not be the norm, it was developed by Mike Fast and Matt Lentzner to show a more real-world version of movement. In this case, we can compare the numbers to some GIF animations and find out whether PITCHf/x sees what we see. The velocity numbers were recorded at 55 feet from home plate, which is closer to the actual release point and a little faster than what you see reported in Gameday.
We're just looking at pitches in the at-bat with glove-side break, so we'll label the horizontal movement “sweep.” Vertical movement is “drop,” since everything drops. Our “sweep” numbers may be about three inches off due to camera calibrations in Minute Maid Park, but it's safe to compare these pitches to each other, as they're from the same park, game, and at-bat.
The Gameday real-time tag for the first pitch was a slider, but it's actually a cutter. Take a look.
Speed: 88 MPH; Sweep: 3 inches; Drop: 18 inches
The 0-1 pitch was a slider. You can see it hook, and it doesn't have the carry the cutter has.
Speed: 82 MPH; Sweep: 19 inches; Drop: 36 inches
On 2-1, Darvish threw a slower version of his slider. It moved a little bit more—let's see if we can tell.
Speed: 79 MPH; Sweep: 21 inches; Drop: 38 inches
It may look like it moved more, but it was down in the zone. That gives it more depth to the eye.
Now we jump ahead to the full-count sequence of four straight sliders. Let's first look at the numbers on those together:
The first one looks flat to me, despite the drop being tied for best of the four. But I feel like I can see the lack of sweep.
The second 3-2 slider was a touch faster with more sweeping action, but the same drop.
Since he started it inside, you can see the sweep a little better than the others.
Next up was a slider with about the same speed and sweep, but less drop.
It looks like the same pitch, same set-up by A.J. Pierzynski behind the plate, but he left it up and Carter yanked high and foul. Hey, that fits the data! Same everything, except the drop wasn't there.
Last one, basically the same as the previous one. But this time Darvish starts it over the plate and it moves away.
Yes, to answer my own question, PITCHf/x does match up with what we see. But you have to keep camera angle and pitch location in mind, too. Classifying Darvish's pitches isn't much easier than hitting them. —Harry Pavlidis
Jason Cole is an author of Baseball Prospectus.