June 7, 2006
Lies, Damned Lies
Tony LaRussa and the Terrible, Horrible, No-Good, Very Long Game
Each year, there are a handful of relatively ordinary baseball games that forever etch themselves in my memory. Last Friday's Cubs-Cardinals game is one such contest.
For one thing, I happened to catch the end of the game at a small rock venue, where it provided some much-needed relief from the most obnoxious indie pop act that I've heard in the past decade--names will be withheld to prevent any contrarian iTunes downloads. For another, the game provided a remarkable display of inefficiency (even for squads managed by Dusty Baker and Tony La Russa): 35 hits between the two clubs but just 9 runs, and 30 men left on base. Finally, there was the drama related to La Russa's bad LOOGY karma. After Sidney Ponson was removed, La Russa employed a sequence of six relievers who managed just three innings between them. This meant that Josh Hancock, the last man standing, had to hit for himself in a key situation in the 12th inning (bases loaded with just one out), and that when Hancock was finally pinch hit for in the 14th, backup catcher Gary Bennett was prepared to enter the game as a reliever if the Cardinals managed to re-tie the game.
Perhaps this wasn't such an ordinary game at all. But in any event, it got me to thinking about just what happens to teams after they play VLGs--Very Long Games that consist of 14 innings or more. This VLG, for example, had some residual effects the next day: La Russa let a struggling Mark Mulder hit for himself in the bottom of the 5th inning rather than turn to the depleted bullpen, even though Mulder had already yielded seven runs and had struck out just one man. This was also the game in which Albert Pujols was injured. Coincidence? Probably, but a day game after a night game is tough enough as it is, and especially so when the night game ran for more than five hours.
Is there a persistent hangover effect from playing in a VLG? If so, how long does it last? Is it severe enough to have an impact on in-game strategy? (For example, a manager could pinch hit for his ace reliever in the bottom of the 10th after just one inning of work, a strategy which would tend to make him more likely to win the game immediately, but perhaps less likely to win the game overall). And does it matter at all whether a team won or lost the VLG?
I examined ten years worth of data on VLGs: 1991-1996, 1999-2000, 2002, and 2005. Why these seasons? Because they were the ten most recent seasons of schedule data on baseball-reference.com that contained detail on extra-inning contests. This represented 126 VLGs in all--14-inning games are rare, occurring at a rate of about once per season per team.
The next step was to eliminate any results from contests in which the team played the same opponent that it battled in the VLG. We're trying to examine whether there are non-zero sum effects from VLGs--if both the Cardinals and the Cubs lose two hours of sleep from playing in a VLG, presumably that will affect the teams about equally if they're playing one another again the next day. Even if it doesn't, one of those teams needs to end up with a victory and the other with a loss, so these games won't be of any help to our study. With that restriction in place, here is how teams fared in the five days following a VLG:
Table 1: Results in five days following a VLG (14+ innings):
VLG +1 22-30 (.423) VLG +2 71-61 (.538) VLG +3 96-116 (.453) VLG +4 105-106 (.498) VLG +5 103-109 (.486) TOTAL 397-422 (.485) WEIGHTED 995-1063 (.483)Note that the sample size increases as we move further away from the date of the VLG, as the team is less likely to be playing the original opponent. The "weighted" total assigns weights in inverse proportion to the number of days since the VLG: a multiplier of 5 for the next day's game, 4 points for a game played two days later, and so forth. (This is arbitrary, admittedly, but the individual data is there for you to play with as you see fit).
The result comes down on the "right" side of the ledger--teams appear to incur a slight penalty from playing in a VLG. But the penalty is very small. Even if we accepted as gospel that a team would play at a .485 winning percentage over the course of the next five days (instead of a .500 winning percentage), and even if this portion of their schedule included no off-days and no games against the original opponent, this would translate to an average penalty of about 1/10th of a win for playing in a VLG. That is hardly worth worrying about in terms of in-game strategy, unless a decision was otherwise very close.
Speaking of off-days, perhaps they reset the deleterious effects of a VLG? After all, this is a chance to recharge the bullpen, to give the starting catcher's knees a day's worth of rest, and so forth. I re-ran the analysis, but this time stopped counting games once a team had a rest day.
Table 2: Results in five days following a VLG, without an off-day:
VLG +1 22-30 (.423) VLG +2 59-49 (.542) VLG +3 69-86 (.445) VLG +4 73-63 (.537) VLG +5 53-62 (.461) TOTAL 397-422 (.487) WEIGHTED 748-792 (.486)This doesn't appear to be illuminating at all. Looking at things like getaway and travel days and combos might be more revealing, but I didn't compile that data.
The next question is whether the length of the VLG matters. What if the game isn't merely a VLG, but a Very Very Long Game (VVLG), consisting of at least 16 innings? VVLGs are very very rare--only 36 were played in the ten years of our sample.
Table 3: Results in five days following a VVLG (16+ innings)
VVLG +1 5-10 (.333) VVLG +2 17-15 (.531) VVLG +3 29-33 (.468) VVLG +4 25-37 (.403) VVLG +5 32-24 (.571) TOTAL 108-119 (.476) WEIGHTED 262-307 (.460)Although we have to contend with some very small sample sizes, there does appear to be some progressive effect related to the length of the VLG, particularly if we look at the weighted version of winning percentage.
Finally, does it matter whether a team was victorious in the VLG? These are the results for all teams that played in and won a VLG. For this version of the analysis, by the way, I am re-inserting the games played against the same opponent.
Table 4: Results in five days following a VLG win
Vs Same Opponent Vs Other Oppt's Total VLG +1 49-34 (.590) 11-15 (.423) 60-49 (.550) VLG +2 16-25 (.390) 41-28 (.594) 57-53 (.518) VLG +3 1-3 (.250) 61-48 (.560) 62-51 (.549) VLG +4 2-1 (.667) 61-44 (.581) 63-45 (.583) VLG +5 3-0 (1.000) 53-55 (.491) 56-55 (.505) TOTAL 71-63 (.529) 227-190 (.544) 298-253 (.541) WEIGHTED 319-281 (.532) 577-474 (.549) 896-755 (.543)Conversely, this is what happened to the teams that lost a VLG:
Table 5: Results in five days following a VLG loss
Vs Same Opponent Vs Other Oppt's Total VLG +1 34-49 (.410) 11-15 (.423) 45-64 (.413) VLG +2 25-16 (.610) 30-33 (.476) 55-49 (.529) VLG +3 3-1 (.750) 35-68 (.340) 38-69 (.355) VLG +4 1-2 (.333) 44-62 (.415) 45-64 (.413) VLG +5 0-3 (.000) 50-54 (.481) 50-57 (.467) TOTAL 63-71 (.471) 170-232 (.423) 233-303 (.435) WEIGHTED 281-319 (.468) 418-589 (.415) 699-908 (.435)Well, this is pretty interesting. Naturally, we'd expect the team that won their VLG to perform slightly better going forward, because winning a game as opposed to losing it is some small evidence of team quality. However, it isn't much evidence of team quality, particularly since games that go 14 or more innings are almost by definition random affairs. And this isn't a small effect--teams that win a VLG play at the equivalent of an 88-74 pace over their next five days, while teams that lose one play 70-92 ball.
It might be thought that this is also the result of the home-field advantage, which could conceivably be especially important in a VLG scenario. A team, of course, is a favorite to play at home tomorrow if it played at home today. However, our results suggest that the hangover effect holds (in fact, it's a bit stronger) even if a team begins a new series against a new opponent. A two-series homestand or two-series roadtrip is by far the most common arrangement under today's scheduling preferences, meaning that a team is about as likely to play its next series in a different venue as it is in the same one.
Perhaps the endorphin rush of stealing a game from your biggest rival is a miraculous antidote for the fatiguing effects of an extra-inning game? Whatever it is, I think we need to look more seriously at the impacts of a demoralizing loss--losing a game in extra innings, after holding a big lead, when blowing a save, and so forth. Baseball players are human beings, after all, and when a loss carries with it both physical and psychological after-effects, their impact can be profound.