Can We See Double Plays Coming?
In a playoff atmosphere, everything gets magnified a bit. Situations are always more tense, every little thing seems to carry more weight. It seems that, if you ask enough people, almost every play could be labeled as a “game-changer” by someone. That’s because there are only so many plays to go around. Outs are finite. In the playoffs, they seem astronomically finite. So when a home run is hit, or a great defensive play is made, it seems to matter more. An approving eyebrow raise in the regular season turns into a full-on shout in the postseason. A minimal eye roll on a random Wednesday in June morphs into an audible “UGH” sound in October.
Nothing can produce an audible reaction faster than a double play. The double play seems to be the only situation where both sides react in an equal and opposite way. Home runs are one-sided. Run-scoring singles are one-sided. Double plays raise sounds out of both sides. Someone is quite happy and someone is quite upset. This is because the twin killing is seen as such a momentum changer — it’s two outs for the price of one. The idea of momentum is something some in the statistical community scoff at a bit. They scoff because they scoff at things they can’t measure, at least accurately. But I’ll come right out and say it, I believe in momentum. I believe that when I’m in a good mood, I do better work. I pay more attention, I get things done quicker, and the overall product is of higher quality. I don’t know how to measure this on a baseball field, but I believe it exists there, too. To think that everything on a baseball field is static and has no affect on anything else in the entire game seems shortsighted. Double plays take wind out of sails, so to speak. They bother people.
The current Tigers roster can do a lot of things well. They can certainly pitch. They can hit. They have at least one person who can field. Something they don’t do very well is run the bases. From 2012 to 2013 — the current Prince Fielder era — the Tigers have ranked dead last in BsR, and it’s not even close. They are a team of basecloggers, if you will. They make up for their lumbering ways by ranking second overall in wRC+ over that span. So the Tigers are slow, but they are good at hitting. And now they are one game away from getting eliminated. Why? Their pitching has been quite good, but the bats aren’t there. A hobbled Miguel Cabrera certainly doesn’t help. Some of their hitting woes may be explained by the fact that they have hit into 12 double plays so far in the playoffs. That’s a lot. Too many, if I may say so. They hit three in their last game, and it cost them dearly, as they all ranked within the top four worst outcomes of the game.
The only team with more double plays this postseason is the recently-departed Dodgers, who also rank poorly when it comes to base running. So the question is: does a reputation of poor base running and slow players lead to a high rate of double plays in the postseason? Could we have seen this coming?
To start, I looked at the teams with the highest GDP rates in the postseason (again, this is in the expansion era).
| Year | Team | GDPRate(Po) |
|---|---|---|
| 2005 | SDN | 5.83% |
| 2001 | HOU | 4.85% |
| 2000 | CHA | 4.42% |
| 2007 | CHN | 4.39% |
| 2007 | BOS | 4.06% |
| 2007 | NYA | 3.92% |
| 2002 | NYA | 3.80% |
| 1995 | CIN | 3.64% |
| 1998 | HOU | 3.55% |
| 2013 | LAN | 3.54% |
| 2006 | OAK | 3.47% |
| 2013 | DET | 3.31% |
| 1997 | BAL | 3.24% |
| 2000 | OAK | 3.13% |
| 2008 | LAN | 3.12% |
| 2005 | ATL | 3.09% |
| 1995 | COL | 3.09% |
| 1997 | NYA | 3.08% |
| 2007 | ARI | 3.01% |
| 1999 | TEX | 2.97% |
| 2010 | TEX | 2.84% |
| 2007 | PHI | 2.83% |
| 2010 | MIN | 2.78% |
| 2011 | DET | 3.20% |
Here we see both the Tigers and the Dodgers as having some of the highest rates in recent history. If the teams in this list all suffer from similar base running ineptitude, we should be able to see a correlation between a high GDP rate in the postseason and a poor BsR in the respective regular season.
| Year | Team | GDPRate(Po) | BsR(Rg) |
|---|---|---|---|
| 2005 | SDN | 5.83% | -2.6 |
| 2001 | HOU | 4.85% | -8.8 |
| 2000 | CHA | 4.42% | 5.3 |
| 2007 | CHN | 4.39% | -7.3 |
| 2007 | BOS | 4.06% | -3.8 |
| 2007 | NYA | 3.92% | 10 |
| 2002 | NYA | 3.80% | 0.3 |
| 1995 | CIN | 3.64% | 7 |
| 1998 | HOU | 3.55% | 8.7 |
| 2013 | LAN | 3.54% | -11.5 |
| 2006 | OAK | 3.47% | -11.3 |
| 2013 | DET | 3.31% | -19.4 |
| 1997 | BAL | 3.24% | 1.9 |
| 2000 | OAK | 3.13% | 1.9 |
| 2008 | LAN | 3.12% | 3.2 |
| 2005 | ATL | 3.09% | 2.1 |
| 1995 | COL | 3.09% | -2.1 |
| 1997 | NYA | 3.08% | -4.3 |
| 2007 | ARI | 3.01% | 9.2 |
| 1999 | TEX | 2.97% | -2.1 |
| 2010 | TEX | 2.84% | 8.2 |
| 2007 | PHI | 2.83% | 16 |
| 2010 | MIN | 2.78% | -4.5 |
| 2011 | DET | 3.20% | -10.5 |
That doesn’t really flush out, it seems. Here’s a graph with the same data.
OK, so if base running value, or lack thereof, doesn’t correlate, perhaps we can predict playoff rates by just simply looking at the regular season rates. I haven’t seen any studies done of GDP rate normalization, but perhaps the sample doesn’t need to be that big for things to match up.
To be totally honest, I thought this one would be a lot more close. But it seems that the playoffs just aren’t long enough for these rates to level out.
This was my last attempt to find something of substance. I took the rate at which the team surpassed the league rate to see if that matched up at all. It’s the closest of the bunch, but still way far from being a reliable comparison.
There is no one quick litmus test for double play rates in the postseason, it seems. It’s most certainly a mixture of player speed and ground ball rates and a good amount of randomness thrown in. There is still work to be done in this, more metrics to test, but I’m not sure what could be gleaned from it. Would managers construct their lineups to avoid plays that happen a little over 3% of the time, at most? Probably not. The Tigers may hit into more double plays tonight, and it may be because they’re slow, or it could be for all sorts of reasons. At this point, it doesn’t really matter, I guess. If there is such a thing as momentum, any more double plays — with elimination looming — will certainly help kill it for the Tigers.
David G. Temple is the Managing Editor of TechGraphs and a contributor to FanGraphs, NotGraphs and The Hardball Times. He hosts the award-eligible podcast Stealing Home. Dayn Perry once called him a "Bible Made of Lasers." Follow him on Twitter @davidgtemple.
GIDP’s are really a factor of luck, right? In a small sample size such as the postseason, there is probably a lot of luck involved. I would imagine that singles+walks and GB% would be the two biggest factors that can be easily measured (speed is harder to measure). There must be a way to figure out what an expected GIDP rate would be, but I see no point in applying that to the playoffs because a) not much will change for the regular season and b) it’s a really small sample size.
Unless you link it to pitcher, I can’t see the worth of even looking. It might be interesting to look at FIP and ERA differences and GDP rates.
I don’t know how much of a link there would be, because groundball/single pitchers have good FIPs in addition to good ERAs (if they get a lot of DPs)
The last two graphs are the same.
Fixed that. Thanks, Ben.
The playoff sample size seems to be the main factor. Maybe you could combine playoff teams above and below a certain rate and see if there is any more correlation.
GDP is mostly a function of being right handed, hitting ground balls, and opportunities (runner on first, less than 2 outs). Speed is secondary. Obviously if your GDP rate stat is “per opportunity” then it is a function of handedness, GB rate, and less importantly speed.
Base running, as FG measures it (using my methodology), is mostly speed. So to start with, there is a tenuous relationship between base running and GDP, since the former is mostly speed and the latter is mostly handedness and GB rate and only partially speed.
On top of that, any time you correlate two variables, if one variable has a small underlying sample size, like a team’s post season GDP rate, you will always get a low “r.” Try correlating two of the exact same metrics but with one or both having a small underlying sample size and you will see what I mean for example a player’s BA in one 10 game period with his BA in another 10 game period. Even though they both have roughly the same population mean (the player’s “true” BA), the correlation will be small, maybe something like .05.
Or another thing… since said teams are playing the same opponent multiple times… is there any correlation between opponents ability to turn double plays and team’s hitting into them?
This. The Cardinals led the league in DPs, and it’s not particularly close. Makes sense they’d turn a bunch in the playoffs too.
It seems arbitrary to look for correlations in postseason data. I know it’s the postseason now, but why limit yourself? If the correlation is really there it will jump out in the regular season data. If it does you can bet it will carry over into the postseason.
Not necessarily. In a 7-game series, a team can face a particular starting pitcher 3 times. If he happens to have a high GB rate, you’d expect a lot of singles and a lot of double plays. This can really skew small-number statistics.
That’s exactly my point though. Why are we focusing on the skewed numbers when we have perfectly good, large N data? If you’re interested in something that changes regular vs. postseason (which I don’t have any reason to think this would given large N) then we should see *both* fits. It makes no sense to focus entirely on the low N data.
As a side note, I’ve been enjoying this site for several months now and I’ve never run into a single confidence interval or standard deviation calculation. That is somewhat surprising and I think it would find productive use in this article.
Momentum almost certainly exists, simply because it is a psychological effect and the baseball players are human: I don’t have the links on me but, IIRC, it’s been proven(or, at the least, data heavily suggests) that people work better when placed in a work enviroment that makes them feel comfortable and happy compared to one that makes them feel uncomfortable or angry/sad, even if the perameters (IE work hours, etc) are the same. Likewise, it’s been shown and suggests that breaks of certain times (IIRC, roughly 20 minutes) produce more work then if they worked all that time straight (even if it was non-physical tasks). And so on. Basically, we KNOW that people’s mental state affects what they are doing.
The hard part, of course, is measuring it, because we can’t ever know what a player’s exact mental state is and seperating it from every other factor that goes into a baseball game is hard. I think that psychologists, psycho-analysis and basically digging into that will, in the future, become a big part of how we look at players…even if it isn’t as big a deal as announcers will make it out to be. I also think that medical science will eventually change the way we look at players: Some people appear to have genetic predisposition to certain injuries and once they find ones that corrolate to common baseball injuries, there’s gonna be a lot of testing. I firmly believe that, while injuries obviously have some luck factor (Like Greinke this year), being injury prone or not is a genetic skill, like the ability to throw 100 MPH gas.
Looking into these kind of things, even if the results show nothing, is an excellent idea. Enjoyed the article and the frankness of believing in momentum.
“Home runs are one-sided. Run-scoring singles are one-sided. Double plays raise sounds out of both sides. Someone is quite happy and someone is quite upset.”
Literally every one of these events evoke similar but opposite reactions out of fans…
“Astronomically finite” is the best oxymoron I’ve ever read in my life. Well played.
You’ve obviously missed ‘creatively conservative’ over thae last few weeks.
Yeah, hadn’t seen that one. It’s still a close second, though.
Well, watching playoffs (perhaps more carefully in these last days of play), I certainly suspect that pitchers are setting up DPs for difficult batters when the right type of batter is up next. Sometimes it even works, and that generally seems to be due to a slow-footer headed for first.
How about analyzing some of this from the other end? Are the Red Sox pitchers ground ball pitchers and with Drew/Pedroia/Napoli, you’ve got great infield defense. Maybe combine the bad base running with a heightened ability to induce double plays gives more predictive information?
“They scoff because they scoff at things they can’t measure”
Doesn’t WPA somewhat measure the concept of momentum? Or at least “potential energy”? Often, a double play reduces WPA by a lot, even though none of those runners were assured to score. This is particularly true for late-inning situations. I’d say “Ugh” is the sound of getting the WPA knocked out of you.