Thursday, June 27, 2013

Clutch Hitting Revisited (Part 1)

I first studied clutch hitting in a piece that was published over two years ago in the Community Research section of FanGraphs (FG)--talk about beginning by jumping into the deep end. I was mildly surprised at the reaction it received and I always intended to return to the subject and investigate it further. The Cubs and White Sox aren't going anywhere--why not now?

Let me be absolutely clear up front--clutch hitting EXISTS.

Allow me to repeat that--clutch hitting EXISTS.

Have I made myself clear? When Joe Carter came to the plate in the bottom of the 9th in Game 6 of the 1993 World Series with Rickey Henderson on 2nd, Paul Molitor on 1st and on the wrong end of a 6-5 score and hit the game-ending and Series-ending home run off Mitch Williams, that was absolutely clutch in any way you wish to define the term--it was in an important game, an important situation, etc. There were runners in scoring position, the game was tight, the Blue Jays already had one out and Carter came through in the quintessential clutch situation. NOBODY ARGUES THIS.

What we DO argue about is whether clutch hitting is a skill, i.e., something that can be replicated. In other words, are there players in baseball history who are better at performing in the clutch--do they seem to rise to the occasion more often than others? If so, what made them special, and can we tease it out, and even more importantly, can it be cultivated and taught? These are all valid questions that I touched on in that FG piece but will look at further. In addition, advances in the data available from Baseball-Reference (B-R) allow me to separate out even more what is important and what isn't and I'll introduce metrics that I was either unaware of or didn't have access to even as recently as 2011.

What is a clutch situation? This will be a somewhat fluid definition as I write this but I'll explain as I add situations or permutations. Initially, my definition is very simple--a clutch situation is any one in which a runner is in scoring position, period. I don't care how many outs there are, what inning it is or the score of the game--I firmly believe that any major league hitter (shoot, any HITTER at any level) batting with the opportunity to drive in a run is keenly interested in doing so. I contend that had the Blue Jays been beating the Phillies 18-2 at the time of Joe Carter's at-bat, first, he wouldn't be batting (there would have been no bottom of the 9th), and second, he would STILL want to do his best to drive the runners in. Would it be EXACTLY the same--probably not, but my point is clear and I will proceed from that foundational understanding.

This chart shows the difference in batting average with the bases empty and with runners in scoring position from 1947-2012:

Not all opportunities to hit in the clutch are the same, and I'll introduce this chart knowing full well it might confuse the issue--it shows how often base runners are driven in depending on the given base situation:

To explain, with a runner on 1st, hitters drove in the runner 5% of the time in 1950, with the range between 3.7-6.7%. In 2012, that improved, and depending on how you want to cut the numbers, an argument can be made that the improvement was rather dramatic, about a 14% increase from 5.0 to 5.7%. Be my guest, but I think that's overstating--I would argue that over time, with runners on 1st, they'll be driven in about 5% of the time, runners on 2nd around 15% and so on. When I created this table I don't remember if I only used players with a certain number of plate appearances, but even so, the ranges aren't dramatic. Taken all together, the average hitter drives in about 15% of the base runners, and in cases like last year with Miguel Cabrera, can move that number as high as 22%. That's about a 50% difference, and that's something to talk about. I wrote about this in a prior post if you wish to read further on it--it's an important topic, but not exactly what I'm writing about here, but it is related.

I don't use advanced math but I do utilize college-level statistical methods that I will explain to anyone interested--email me at the link to the right and I'll answer any questions you have. In my initial FG piece, I suggested a 10% increase in batting average as a marker of clutch hitting and saw nothing anywhere near that, making me smugly assume that I had conclusively proven once and for all that clutch hitting didn't exist. As I thought about that, that was a ridiculous standard--a 10% increase in ANYTHING in life is usually something momentous and a fairly arbitrary and rigid standard.

It's difficult to look at meaningful trends without having samples of adequate size, so I chose to use any player with at least 3,000 plate appearances through the 2012 season going back to 1980, and for players whose last year was 1980, going back for their entire career. I used the B-R Event Finder for each  player to amass all their plate appearances, and in the end had approximately 5 million. I also incorporated any Hall of Famer from 1947 on--it unfortunately leaves out players like Babe Ruth, Lou Gehrig, Ty Cobb, etc. but does include players like Hank Aaron, Mickey Mantle, Willie Mays and so on.

"But Scott, that's not FAIR--you cherry-picked the best hitters in baseball history to make your point!" Absolutely I did--I would argue that if clutch hitting exists, this is precisely where to find it, in the creme de la creme of baseball hitters. The number of plate appearances in a player's career serves as an excellent marker of player success--it's very difficult to find historical players who had fewer than 3,000 plate appearances that are considered among the best in baseball history (absent injuries or Negro League players, of course). If there's clutch hitting to be found, it should be in this group.

I'll present the data in three different ways. In this first method, I'll show differences in batting average (BA) and on-base + slugging (OPS) over three different scenarios:
1. Bases empty (e)
2. Runners in scoring position (risp)--(-2, -3, 1-2, 1-3, -23, 123)
3. Runners in scoring position, less than 2 outs (rispl2o)
This table shows the first results:

Just so we're clear, this sample of 838 hitters, arguably the best hitters of the past 30-odd years and inclusive of the best since 1950, has a collective batting average of .268 and OPS of .756. With batters in scoring position, that average moves to .276, an increase of 2.9% and so on. I'll be the first to state I'm intrigued by the increase from the bases empty to runners in scoring position with less than two outs, because that's huge, but I'll leave that aside for the moment.

What DOES an increase in average of 8 points really amount to? In a typical season of 600 at-bat, that would translate into about 5 hits. Of course, when these hits occur and what kind of hits they are is tremendously important, but it's the rare season that one state hinged on one hit in a given situation--to suggest that is to believe that games are played in a vacuum and that seasons aren't cumulative experiences. Having said all this, I must state that the difference in batting average between the three situations is statistically significant. I don't dispute or underplay that, I merely suggest that there is a huge difference between statistical significance and real-world impact, and I believe that's the case here, because the statistical significance translates into about one more hit a MONTH.

The argument I make is a simple one--if clutch hitting is a skill, it can be replicated andutilized on demand. Tiger Woods hits his driver 300+ yards--it may not go straight, it might end up in the rough, woods, water or sand, but it will go 300 yards, and it's in the fairway 62.2% of the time. If a hitter had similar numbers, he'd be batting .620 with runners in scoring position, and of course no one does that. Tiger Woods is in control of his golf club--he can't control the weather or the course, but he can certainly adjust to these and any other factors. A hitter doesn't have that luxury--he can't ask for a fat fastball right where spray charts show is his power zone. While the hitter wants the ball chest high right down the middle, chances are the pitcher is throwing inside or at his knees, making adjustments difficult.

But we can make some distinctions. Early on in the data collection process I noticed that players with fewer career plate appearances were showing more dramatic increases in batting average. This is to be expected--the lower the batting average, the easier it is to have a higher percent increase. A .220 hitter can increase his batting average 5% and still only be hitting .232--better than .220, but still nothing to be celebrated. The .300 hitter, however, might only increase his batting average 2.5%, but this still translate into a .308 average, an increase only four points less than the .220 hitter. That's the eternal battle between absolute numbers and percentages--I heard it best explained to me in my former life as a pharmaceutical sales representative. I was discussing relative profit margins with a pharmacist, stating that while a branded drug costs more, a generic has a significantly higher profit margin, to which he replied "I pay my people in dollars, not margin." Pretty much ended that discussion. The hitter that hits .270 gets paid, the .220 hitter than increases his batting average 10% gets a ticket to Triple-A (or released).

These next tables segment players by relative success. This first one breaks down the data by career plate appearances:

It should come as no surprise that career batting averages increase directly with career plate appearances--if a player isn't hitting enough, he'll no longer have a career, and even that success needs to be sustained, or else he's done. In baseball history there have been 268 players with at least 8,000 plate appearances, of which 152 are included in this sample. To  do that a player has to be not merely good, but among the best that ever played the game. To illustrate, I sorted this group  by WAR using the B-R Play Index feature, and the WORST players by WAR with at least 8,000 PA are Doc Cramer, Don Kessinger, Kid Gleason and Charlie Grimm, but we understand how that occurs--Gleason played in the Dead Ball Era, Cramer was a light-hitting center fielder that still gets suggested (wrongly) for the Hall of Fame, Kessinger was a middle infielder who played during the Mini-Dead Ball Era of the 1960s and Grimm gets dinged in WAR calculations for being a first baseman. Having said all that, none of these guys are considered bums or players unworthy of lengthy careers, but instead help prove my overall point--you don't have a long career unless you have skills.

Let's slice it differently--here is the data by All-Star Game appearances:

Now we start to see something. I don't argue that All-Star selections are perfect, particularly with modern rosters numbering in the 30s, but it's rare to be an All-Star more than once without being considered among the best players in the game. We do see some separation, and also the expected lower increase in average with less than two outs as the players improve, because their ability to improve reaches natural limits. 

I've made the decision to split this into two posts, so I'll finish with some tables showing some players that did have success--here are the top 20 players in terms of increased average with runners in scoring position:

There are some intriguing numbers in here, but generally speaking, these are players who have tremendous increases because they had low batting averages to start with. Those Kevin Youkilis numbers are impressive, but not so much for the Yankees this year. Do any of these players jump out at you and roll off your tongue when you discuss clutch hitting? Of course not. So, who's at the bottom?

So  I guess the Yankees should trade Robinson Cano and get whatever they can for him--he's the perfect illustration of the shortcomings of this exercise, in that when you bat .322 in any situation, there's typically only one direction in which to go.

No matter how we view the data, the difference between the bases empty and runners in scoring position with less than two outs is downright dramatic--even players who had career averages increased from .313 to .324, a 3.5% increase. The one upside of watching Cubs games this year is listening to Len Kasper and Jim Deshaies call the games on TV, and they'll have plenty of time to discuss the phenomenon over the long Chicago summer. It can be many things--throwing from the stretch instead of a windup, nerves, maybe even hitters trying harder, but the numbers are there and irrefutable. I plan on doing a clutch PITCHING analysis as well (I'm still gathering the data) that will show slightly different values from what is seen here since that sample will include only pitchers who faced at least 5000 batters. In this sample, the hitters were facing everyone from Roger Clemens and Greg Maddux to bums who got one start and were never heard from again (believe it or not, there have been 12 pitchers like this since 2000)--there HAS to be some difference between facing bums and facing aces, but that's for another day. 

I'll end this section with stating that it's best to view clutch situations independently and not as a pool of instances to be grouped, sliced and diced. Every single plate appearance is different, from the big (pitcher, score, outs, time of season) to small (time, date, day/night, feeling good/bad and many other factors), and extrapolating is difficult. This table shows it best, breaking down these numbers by batting average with bases empty:

The higher the average, the smaller the increase with runners in scoring position--perfectly reasonable. Think about it--who would you rather have, the .270 or above hitter or the .240 hitter who may (or may not) show some propensity to perform in the clutch? Most managers and GMs don't think too hard about this.

And in the next post, I'll show why none of this matters--how's THAT for a teaser? I'll also tell you how you can contact me and receive the data that I used--not every one of the five million or so plate appearances, of course, but the amalgamated data that you can look at yourself and reach your own conclusions. Clutch situations exists, and it's very real that hitters have more success with runners in scoring position and less than two outs, but it's not a skill, it's a simple binary process that each time the hitter bats, he is either successful or he isn't. There are NO clutch hitters, just hitters that bat in clutch situations--sometimes the worst hitter hits the game-winning home run, sometimes Yasiel Puig strikes out (just not very often), but what happens in one instance has no carryover effect to the next. Clutch hitting is an opportunity, and the better hitters will perform better, not because they're better clutch hitters, but better hitters.

No comments :

Post a Comment