Beyond OPS: filling in the gaps
I wanted to talk a little about OPS. I love OPS. I use it all the
time when I need to know, say, whether Nick Markakis had a good year
in 2008*, or if I’m wondering which was Vlad Guerrero’s best year.**
OPS will not give you a definitive answer, but it’ll get most
of the job done.
*Markakis’ OPS last season was .897, a very good number (seventh in the
American League). I’m beginning to really like Nick Markakis: fine
bat, solid defense, rifle arm, decent base runner. The 25-year-old
Oriole right fielder doesn’t seem to have any holes in his game.
**Vlad’s best year with the bat was probably 2000 or 2003 or 2004. He
topped 1.000 OPS in each of those years.
The value of OPS derives from its ability to measure offensive
production pretty well, while being super easy to calculate. Actually, you don’t
need to calculate it anymore, since it’s found on most mainstream
baseball web sites nowadays. By now we have a feeling for what’s a
good OPS: .750-.770 is about average, and 1.000 is a
league-leading-type number.
OPS’s biggest shortcoming is probably its failure to correct for park
effects: a 1.000 OPS in Colorado is not as impressive as the same
number achieved in San Diego. This defect is remedied by OPS’s close
relative: OPS+, which takes into account overall league offensive
level in addition to park effects and puts everything on a nicely
calibrated scale where 100 is average, 150 is 50 percent better than
average, and so on.*
*A typical league-leading value for OPS+ is around 170 or so. In 2008,
Albert Pujols led MLB with an OPS+ of 190. Albert, though, was
way ahead of the pack—only one other player topped 165
(Chipper Jones at 174).
So, is OPS+ the best way to judge a player? First of all, let’s not
forget that hitting is just one aspect of baseball. When estimating
a player’s value or comparing two players such as,
say, Manny Ramirez and Carl Yastrzemski you need to
consider defense (range and arm), position played, base running and
maybe some other stuff as well. So, when I said that Markakis had a
good year last year, I really don’t know that. I need to check his
defense and base running. (I just checked: average range and
base running, super arm—overall a very good season.)
The neglected double play
Even putting aside defense and base running, OPS has some
shortcomings, one of which is the subject of this article.
Now, let me say right off the bat that this isn’t going
to be one of those articles that compares a bunch of different run
estimators, complete with correlation coefficients, citations of RMSE
and the like.* Instead, I’m going to focus on one aspect of hitting
that OPS doesn’t take into account: hitting into double plays or, more
specifically, grounding into double plays (known as GIDP or GDP). Sometimes a player’s
tendency to ground into double plays or his skill in avoiding the
double play can make a real difference in how we view his
production. At least, it should make a real difference.
*I don’t want to disparage the “theoretical” work that goes into
finding the best run estimator. That kind of research is essential
for moving sabermetrics forward, but this article is going to be
smaller in scope.
I don’t mean to pick on OPS here. As far as I know, no statistic that
purports to evaluate offensive production includes GDP. Batting runs,
base runs, runs created, wOBA, VORP—none of these incorporate
double plays. And for a good reason—not all players have the
same opportunities for hitting into a double play, so to include
double plays in a stat like OPS, you’d need to consider how often a
player comes to bat with a runner on first base and fewer than two
outs. And that requires a analysis of play-by-play data, not just
season totals. Taking GDPs into account fairly is a lot of
work. Work that, in many cases, is probably not justified. But not all
cases, not all.
My own double-play philosophy
My views on the GDP have changed over the years. I used to
disparage the slow-footed oafs who seemed to ground into double plays
with monotonous consistency. I can remember cringing whenever George
“Boomer” Scott came to bat with Rice or Yaz on first base,
knowing, just knowing that a nifty 4-6-3 double play was
imminent. I can remember once standing up in the right field bleachers at
Fenway and yelling, “Strike out, ya’ bum!” Boomer hit a two-hopper to
the right side. I hated Boomer.
Many years later, but still several years ago, my brother and I were
talking about Mike Piazza.
“What do you think of Piazza?” he asks me.
“What do I think of him? He’s a future Hall of Famer and the best guy
on the team, that’s what I think of him! How can you not love
Piazza?”“Doesn’t he ground into a lot
of double plays?” asks my brother. The Mets star catcher, despite being the best
hitter on the team (as measured by OPS, heh), was being bad-mouthed in
the New York press for grounding into a lot of double
plays.“Yes, I suppose he does, but hey, he’s a big strong guy who
hits the ball hard and doesn’t run that well. Plus he’s always got
Alfonzo on first base.” Then I asked a question: “Do you know who the
career leader* in grounding into double plays is?”“Who?”
“Hank Aaron*.” Ah.
*Cal Ripken has since passed Aaron on the career GDP list. Cal was a pretty good player, too.
This was the point I was trying to make: excellent hitter, hits with lots of guys
on base, doesn’t run well (and not many catchers do)—of course, he’s going to hit into a lot of double plays. Don’t
worry about it, it’s part of the package. Don’t hold it against him.
That’s wrong, though. Mike Piazza is what he is (or was, I guess), a
.300 hitter, decent plate discipline, exceptional power—all of
it exceedingly rare in a catcher. But, you know what? Piazza did hit
into a lot of double plays; that’s part of his package, too. And if
you’re going to evaluate a player, you need to evaluate the whole
package, or at least as much of it as you can.
Figuring out double-play performance
You have probably already guessed that the main reason that Ripken and
Aaron have hit into more double plays than anybody else is that they
had more chances to hit into twin killings than most. In fact, they
are nos. 3 and 2 on the all-time (actually, since 1953, when the
Retrosheet play-by-play data kicks in) most GDP opportunities list.
Obviously the thing to do is to normalize a player’s GDP by his GDP
opportunities. “Normalize” is just a fancy word that means “divide,”
and don’t ask me why I used it at all. Even when you normalize the
GDP number, though, do we know how to judge the result? For example, Albert
Pujols has GDP’d in about 12.4 percent of his opportunities. Is that good?
Bad? Average? You need a reference point, say the average value for
all players, which happens
to be 11.5 percent, more or less.* So, Pujols grounds into double plays at a
slightly higher rate than the average batter.
Now, Pujols ranks third in double plays grounded into since 2001, so
if you didn’t consider his opportunities, you’d conclude that he was
one of the worst GDPers of his time, instead of merely slightly below
average.
*As mentioned, the average GDP rate in recent years is 11.5 percent.**
It varies slightly from year to year, but not much. Note that I’m only
including groundball double plays. About 2 percent of double-play situations
result in non-groundball double plays, some of them line drives, some fly outs and
not a few of them strike ’em out/throw ’em out caught stealing double
plays. I’m only considering GDP’s—I don’t think batters have
much control over the other kinds. Or, at least, they have less
control.
**Just for fun, I checked the Retrosheet data for the year
1911. That’s not a typo—Retrosheet has play-by-play data from 1911,
and in that Dead Ball year, the average GBP rate was around 6 percent, about
half the modern value. Whether the difference is due to different
strategies (lots of bunting, hit-and-runs, stolen base attempts) or
inferior defense is an interesting question. Maybe for a future
article.
Ok, so we can compare a player’s GDP rate and compare with league
average, but an even better measure is the number of GDP’s above or
below average. That is, above or below the average number of GDP given
a certain number of opportunities. Let’s go back to Pujols and look at
his 2008 numbers: he
came up in 173 double-play opportunities and grounded into 16
DPs. His rate is 9.25 percent, which is actually better than average. An
average batter with 173 opportunities would have grounded into about 20 double
plays (173 times .114, which was the average GDP rate in 2008),
so let’s call Pujols +4 in DPA (double plays avoided*). Make sense?
*Feel free to forget the acronym DPA as soon as you have finished
reading this article. Despite being on the geeky side, I have a real aversion to the
proliferation of acronyms to label arcane baseball stats. Right now, I need to
give this stat a name, so I can refer to the thing I’m writing about, but you
can safely forget it after finishing this article. I’m going to.**
**UPDATE: This article will probably have a follow-up, so don’t forget what DPA is for another week
or so.
The best and the worst
Now that we’ve decided to evaluated GDP tendencies with DPA (right, Double Plays Avoided), here are
the leaders over the last three seasons (2006-2008):
+-----------------+------+------+------+ | name | Opps | GDP | DPA | +-----------------+------+------+------+ | Sizemore_Grady | 308 | 10 | 25.4 | | Beltran_Carlos | 406 | 25 | 21.8 | | Patterson_Corey | 233 | 6 | 20.9 | | Abreu_Bob | 505 | 38 | 20.2 | | Suzuki_Ichiro | 313 | 17 | 19.0 | | Utley_Chase | 376 | 25 | 18.3 | | Giambi_Jason | 300 | 17 | 17.6 | | Damon_Johnny | 261 | 13 | 17.1 | | Howard_Ryan | 415 | 31 | 16.8 | | Hawpe_Brad | 387 | 28 | 16.6 | +-----------------+------+------+------+
This list makes sense; most of these guys have good speed and can get
down the line fast to avoid the double play. Wait, what are Giambi and
Howard doing on this list? What, do they get exceptional jumps out of
the box or something? Jim Thome is 17th on this list (14.6 DPA) and
David Ortiz is 20th (+12.7). Is anybody else surprised that these
sluggardly sluggers are above average at not hitting into double
plays? What’s this about?
In other words,
what makes a hitter good at avoiding the double play? The obvious
answer, foot speed, is only a partial answer. Otherwise Giambi and
Howard
would not be on the list. These big slow guys avoid double plays with
another technique: they avoid hitting ground balls. Actually, a fair fraction of the
time, these guys avoid hitting the ball at all. For example, in 2008, 37 percent of Giambi’s
plate appearances resulted in a strikeout, walk or hit-by-pitch. And
when he did put the ball in play it usually wasn’t a ground ball.
In fact, both Giambi and Howard hit a ground ball in only 20 percent of their
GDP opportunities. Compare that to Ichiro, who hit a grounder in over
40 percent of his GDP opps.
That’s how Giambi and Ryan Howard
and Jim Thome manage to be above average when it comes to avoiding the
double play—they don’t hit many ground balls.
Here’s a list of the trailers, the guys who are grounding into more than
their share of double plays (2006-2008):
+-----------------+------+------+-------+ | name | Opps | GDP | DPA | +-----------------+------+------+-------+ | Tejada_Miguel | 438 | 82 | -31.5 | | Molina_Yadier | 247 | 53 | -24.6 | | Hudson_Orlando | 301 | 55 | -20.3 | | Konerko_Paul | 373 | 63 | -20.0 | | Castillo_Jose | 257 | 49 | -19.3 | | Berroa_Angel M. | 134 | 34 | -18.5 | | Martinez_Victor | 350 | 58 | -17.5 | | Molina_Ben | 292 | 51 | -17.4 | | Lo Duca_Paul | 222 | 43 | -17.3 | | Young_Mike | 432 | 67 | -17.2 | +-----------------+------+------+-------+
I guess there aren’t many surprises here: catchers, other slow guys,
guys who don’t strike out or walk too much. Some are good hitters,
but when you evaluate their worth, better take into account all these
rally-killers they are hitting into. Which raises the question: how
much are these double plays costing the team, anyway?
The price you pay
Let’s try to put a run value on the double play. Now, a generic out
costs the team about .3 runs, more or less. So, we might estimate
that a double play is worth, well, double that: minus .6 runs.
The thing is, though, the double play seems more costly than just two
outs, doesn’t it? It just seems to hurt more* than two generic outs. And
you know what? The double play is more costly than two regular
outs. That’s because it always happens with fewer than two outs and
runners on base. These are often situations where the offense is expected
to score multiple runs; the double play drastically reduces that
run scoring potential.
*The worst moment of my Little League career (which admittedly had
quite a few bad moments): runners on the corners, one out, we’re down
by two in the bottom of the seventh (final) inning. I step to the
plate and hit it right at the third baseman. I run hard down the
line, I think I can beat it out. I see the relay to second base from
the corner of my eye. I’m starting to worry as the ball is on its way
to first. I’m out by a full step. Grounded into a double play to
snuff out our comeback and end the game! Do you know how rare it is
for 11-year-olds to turn the double play? I cried as our manager
gathered up the bats.
So, I went through the Retrosheet play-by-play data to figure out the
run value of the GDP: I get -.85 runs, pretty much constant over
the last 55 years or so. So, now we can attach a run value to these
extra double plays (above or below average). We do that by taking the
difference between the value of the double play (minus .85) and the
generic out. In other words, how much did the double play
cost relative to making a single out (e.g. striking out or popping out
or grounding into a 6-4 forceout, but beating the throw to first).
Now an average generic out is, as mentioned above, worth -.3
runs. But in a double-play situation, the generic out is somewhat
more costly—it’s roughly -.38 runs. So the difference in
value between hitting into a double play and making a generic out is
-.5 runs*.
The actual value is -.47 runs, but to be honest, I don’t really know if the real value is -.47 or -.45
or -.51 runs. It’s best to round off to -.5 runs. Let’s not
pretend we know more than we really do.
I won’t reproduce the above tables with the run values; it’s enough to
divide the DPA number by two to get the runs saved (or cost). So
Grady Sizemore was worth an extra 11 runs (about one win) to the Indians over the
last three years. At the other extreme, Miguel Tejada cost the
Orioles nearly 18 runs (almost -2 wins) over the same time period.
These are not large numbers, but as I like to point out, in today’s
game a win on the free agent market is valued at around $4-4.5
million.
What have we learned?
For one, we’ve seen that for some players it’s important to take into
account their performance in double-play situations when estimating
their value. Not for the majority of players, but for some. Here’s a
list of players, from the 2006 to 2008 time period, whose double-play performance changes their overall single-season value by at least a half-win:
+-------------------+------+------+---------+ | Name | Year | Team | DP_runs | +-------------------+------+------+---------+ | Thome_Jim | 2006 | CHA | 6.1 | | Johnson_Kelly | 2008 | ATL | 5.3 | | Beltran_Carlos | 2006 | NYN | 5.0 | | Glaus_Troy | 2006 | TOR | -5.0 | | Molina_Yadier | 2008 | SLN | -5.0 | | Peralta_Jhonny | 2008 | CLE | -5.1 | | Berroa_Angel M. | 2006 | KCA | -5.3 | | Butler_Billy | 2008 | KCA | -5.5 | | Escobar_Yunel | 2008 | ATL | -5.5 | | Ordonez_Magglio | 2008 | DET | -5.5 | | Guerrero_Vladimir | 2008 | ANA | -6.2 | | Tejada_Miguel | 2008 | HOU | -7.1 | +-------------------+------+------+---------+
We’ve also learned that avoiding double plays is not just a question
of foot speed. You can also avoid the DP by not hitting ground
balls. Indeed, the correlation between speed and double play
tendencies is not all that strong. In a future piece I will
investigate this issue a bit more. I also want to have a look at
historical double-play performances: the best and the worst all-time
at avoiding the double play.