Felix Hernandez and arbitrarily historic performancesby Jesse Sakstrup
August 17, 2012
Wednesday afternoon, as I made my way to the digital pages at ESPN.com to check some of the early box scores, my eyes were directed toward the main article on the site’s home page, notifying those who were unaware that Felix Hernandez was in the eighth inning of a “perfect game.”
I had been unaware that the Mariners’ righty was on the verge of making history. So what did I do immediately upon reading this news? I proceeded to check the box scores of the early games. Hernandez went on to complete his historical feat, and I didn’t watch a single pitch and it didn’t bother me.
As someone who delves into sabermetrics on a regular basis, it seems odd to me that we are still placing the perfect game on such a pedestal without question or further analysis. Granted, a pitcher who has thrown a perfect game has probably pitched well that night, but that doesn’t mean that he didn’t make any mistake pitches or that hitters didn’t get unlucky on balls in play.
Rewinding to Matt Cain’s perfect game earlier this season, Gregor Blanco made an outstanding catch, deep in the right-center gap, to preserve the unblemished outing. But, had Blanco been playing a step closer to the plate or a step closer to the line, or had he taken a slightly less optimal path to the ball, or had Aubrey Huff or Brandon Belt been playing right field that day, the ball would have dropped in for extra bases and Cain’s performance would have faded into history as just another forgotten one-hit, no-walk, 14-strikeout masterpiece.
Cain would have pitched the exact same game, and all of the balls in play would have been identical, but if Blanco didn’t make that catch, Cain’s performance would have been treated differently.
This, to me, is wrong. It is shallow analysis and it is emphasizing the result more heavily than the process. That isn’t how we measure two-week samples, or how we measure expected future value, so why do we measure performance in a given game this way?
The idea that the end outcome of a pitch doesn’t determine the quality of the pitch shouldn’t be an obscure idea to any analytical baseball fan. A pitch right down the middle is usually not ideal, even if the batter swings and misses or pops it up. The same can be said for a pitch down at the ankles that is put into play for a hit. Provided that that particular hitter doesn’t thrive on pitches at his ankles, then it was a good pitch, but a bad outcome.
Bill James’ game scores attempts to measure performance in a somewhat objective way, giving credit to strikeouts and innings pitched, while punishing for hits, earned runs and walks. This is a quick way to approximate the value of a start without actually watching it or looking at the advanced data, but we could probably do a lot better if we wanted to take the time or had access to the proper data.
So, if looking at process is a better way of evaluating a start than looking at results are, then what are the factors that should be considered? Well, the list could be just about as lengthy as you wanted, but we will stop short of forecasting the marginal value of what a player ate for breakfast. This is the list that I came up with:
The uncontrollable environmental factors: This encompasses weather and ballpark. Balls fly further and hitters typically hit better when the weather is warm, so a each out in 80-degree weather is probably more difficult to achieve than an out in 60-degree weather. Wind is also a major factor. If a ball is hit with precisely enough power to send it 400 feet in a context-neutral environment, but wind resistance causes it to travel 375 feet, which results in an out rather than a home run, the pitcher should not be credited for this positive outcome. The same goes for ballparks; how a ballpark affects home runs, hits and pitch movement cannot be ignored either.
Opponent: Who are you playing against? A perfect game at home against the Rockies is probably an easier achievement than a one-hitter with no walks on the road against the Yankees or in Texas. Not only must you factor in opponent, but the venue as well, as some hitters hit better at home than on the road, for whatever reason.
Additionally, handedness and platoon splits must be considered. A left-hander who saws through a lineup stocked full of lefties with a same-handed platoon split isn’t as big an accomplishment as the glowing box score might indicate.
Actual on-field performance: This is what the pitcher actually does. How hard is the pitcher throwing? How well are his pitches located? How well is he mixing his pitches? How much is each pitch moving? Is he locating pitches primarily in batters’ cold zones? Does the pitcher have a consistent release point and a consistent delivery?
This is really all a pitcher can control. Once the ball leaves the pitcher’s hand, the result is out of his control. A perfect swing is going to be able to hit the perfect pitch, because the swing is in response to the pitch. The hitter has a distinct advantage here in the ability to reduce luck or variance. A pitcher must hope his pitch is good enough that the hitter has little chance of hitting it. Pitchers can maximize their chances of a positive outcome with good velocity, movement, location, sequencing, exploitation of hitter weaknesses, and a repeated delivery, but these just reduce the chances of a negative outcome, they can never eliminate them.
Type of contact: While batters can, and often do, make good contact with good pitches, it happens a lot less against quality pitches than it does with poor pitches. Tracking the velocity, angle and direction off of the bat would be a much better quantification of a pitch result than simply applying the hit or out tag to a ball in play.
Why is it that we can see a line drive that gets caught and say that the pitcher got lucky, or we can see a dribbling ground ball stop half way between home plate and third base, allowing the batter to leg out an infield single, and say that the pitcher got unlucky, and yet, at the same time, we (some of us) look at box scores and draw conclusions about the performance of a pitcher?
If all you knew was that Felix Hernandez threw a perfect game, could you make a reasonable estimation about the quality of his performance? You could conclude that he probably pitched well, but to say it was one of the top 23 performances of all time would be overestimating the accuracy of the box score line. This, however, is how history evaluates performance. Hernandez’s perfect game will go down as one of the better perfect games of all time and may be considered a top-10 pitching performance.
Balls not put in play: Does the pitcher have swing and miss stuff? Swinging strikes are a pretty good indicator of how a pitcher’s stuff is working. The more a batter swings and misses, the better the likelihood that the pitcher is throwing difficult pitches to hit. Similarly, the more called strikes a pitcher gets is usually a positive reflection on his pitch location and his pitch sequencing. A pitcher never tries to split the heart of the plate, but if he does and the hitter is frozen, then he probably did a good job of sequencing his pitches and fooling the hitter. This isn’t great execution of a pitch, but it is good execution of being unpredictable.
Strikeouts are the best outcome a pitcher can get, and, while a walk isn’t the worst outcome for a pitcher, throwing the fourth ball outside the strike zone guarantees that a batter reaches base, and that is bad. If the batter puts the ball in play, the expected value of the average outcome is not as bad as a walk, so a walk is the worst result that a pitcher has control over.
Perhaps an algorithm could be created that could encompass all of these data—and anything I am leaving out—into an expected outcome value based on what usually happens when a pitcher does certain things. But for now we are going to have to draw our own arbitrary conclusions about the data we have.
Lets put the Felix Hernandez perfect game under the microscope to the extent we can.
The game-time temperature was 79 degrees and the wind was 4 mph. The temperature certainly wasn’t optimal for pitching, but not terrible, either, and the wind was probably a non-factor. He pitched at home, in Safeco, a park that increases strikeouts and decreases all four types of hits, so this was definitely beneficial.
Taking the field in Hernandez’ opposition were the Tampa Bay Rays. Here is the batting lineup they rolled out and their wRC+ against righties and away from home in 2012.
1. Sam Fuld (93 against righties for career in 367 at bats; and 87 on the road)
2. B.J. Upton (96 and 87)
3. Matt Joyce (154 and 131)
4. Evan Longoria (105 and 94)
5. Ben Zobrist (123 and 140)
6. Carlos Pena (101 and 78)
7. Jose Lobaton (61 for career and 95)
8. Elliot Johnson (96 and 129)
9. Sean Rodriguez (55 and 58)
The starting lineup that Tampa put out there had a collective wRC+ of 98 against righties and a wRC+ of 100 on the road. There was basically no advantage gained by Felix in facing this version of the Rays.
Hernandez’ actual outing was a smooth, 113-pitch performance. He struck out 12 hitters (44.4 percent), he induced 26 swing and misses (23 percent swinging strike rate) and his fastball registered an average of 93.2 on the radar gun. Felix is striking out 24.4 percent of all batters this season, is garnering swinging strikes 10.4 percent of the time and is averaging about one mph less on his heater for the season than he did this outing. On the surface, his stuff appears to have been tremendous on Wednesday.
He mixed his pitches well, too, throwing 28 fastballs, 24 curves, 23 change-ups, 18 sliders, 13 cutters and seven sinkers. Looking at the PITCHf/x data, there may have been a couple of incorrectly classified pitches, but the general idea is the same: Hernandez made himself unpredictable by mixing in a great variety of pitches.
Here is the pitch plot of his pitches from his perfect game.
He stayed near the edges pretty well, but there were plenty of pitches in the middle and near the middle that were likely mistakes and he was probably lucky to get away with every single one of them.
Here is the pitch plot of his previous start on Aug, 10, in which he allowed five hits, four earned runs and one walk.
His location here isn’t quite as good—there are more pitches in the center area of the plate—but it doesn’t seem like too substantial a difference.
Since we don’t have access to HITf/x data and have no way of determining ball angle off of the bat and velocity off the bat with any exactitude, grading the type of contact that the Rays made on Felix’s pitches will lack precision. We can look to the batted ball classifications—if you trust them—to see that Hernandez gave up eight ground balls, five fly ball and two line drives. Again, velocity off the bat does matter, but most line drives end up as hits, so there is a good chance there was some luck involved in both line drives being caught. Still, eight ground balls and two line drives out of 15 balls in play represents relatively poor contact.
Hernandez’ outing looked perfect in the box score. His outing still looked very good under a more analytical lens. He struck out more than 40 percent of the Rays hitters he faced, he generated loads of swinging strikes, induced what appears to be weakish contact, and stayed around the strike zone all night. The venue was to his benefit, but the opposition was neutral. And the game-time conditions couldn’t have swayed very far in either direction.
If we were able to look at the HITf/x data and at each pitch’s location in relation to each hitter’s hot and cold zones, and if we were able to quantify the value of both, then we could better analyze this, or any start.
Despite the fact that Hernandez was brilliant in his dismantling of the Rays lineup, labeling his outing as perfect would be sensationalizing what was simply an excellent or exceedingly excellent outing.
Was this one of the 23 best single-game pitching performances of all time? Possibly, but we simply do not have the requisite tools to make an objective conclusion about that.
It can be exciting to watch historic performances, but the arbitrary filters for what is and isn’t considered historic are not the best measurements of true performance. They measure the result, but not the process, the pitcher who had the best outcomes, but not the pitcher who pitched the best.
Is a perfect game better than a one-hit, no walk game? Sure, usually, but is the difference between a perfect game and a one-base runner game so many magnitudes larger than the difference between a one-base runner and a two-base runner game that it requires accentuation and cajolery on such a historic level? No, it just isn't.
Unfortunately, until we can quantify the expected value of every individual action on a baseball field, there will be no sterling measurement of single-game performances.