Does momentum matter going into the postseason?
The Yankees won four in a row and 16 of their final 21 to capture another AL East flag. The Tigers won eight of their final 10 games, and 15 of 22 down the stretch to claim the AL Central. The A's won their last six games of the regular season, and went 17-8 in their final 25 games — including a trouncing of the rival Rangers on the season's final day — to win the AL West. Clearly, all of those teams are carrying some amount of momentum into the playoffs, and can be expected to fare better than the Rangers, who lost seven of their final nine games and went 15-16 from Sept. 1 onward, blowing what was thought to be a sure-thing division title to wind up in the wild-card game. Right?
Not so fast. Conventional wisdom may suggest that teams who steamroll into the playoffs will continue chugging along and destroying anything in their paths, but the reality is that recent performance tells you very little about how a team will fare once it reaches the postseason. That's the result of a study I did a few years ago for Baseball Prospectus, slicing up segments of playoff-bound teams' final months in search of telltale signs of who held the advantage. Somewhat surprisingly, I found none.
Intuitively, that should be at least somewhat obvious given that the aforementioned teams are on collision courses. Which immovable force will yield in an A's-Tigers matchup? Who can say offhand whether the Rangers are at a certain disadvantage in Friday's wild-card game against the Orioles, who went 5-2 over their final seven, and 20-11 since Sept. 1? After all, the game will be played in Texas, and home field advantage persists in the postseason to more or less the same degree as in the regular season.
In general, we tend to remember extremes such as the 2007 Rockies storming down the stretch by winning 13 of their final 14 scheduled games, then taking a Game 163 play-in and sweeping both the Division and League Championship Series en route to the NL pennant. We forget that just a year earlier, the Cardinals lost nine of their final 12 games and finished 83-78 before recovering to trample the Padres, Mets and Tigers to win the World Series. Or that in 2010, the Rangers went 16-14 from Sept. 1 onward before flipping the switch to defeat the Rays and Yankees (who themselves had gone 13-17 in that span) before falling to the Giants. Or that in 2000, the Yankees lost 15 of their final 18 games while being thrashed by a combined score of 148-59, leading historians to dust off comparisons to the 1899 Cleveland Spiders, the ultimate doormat; the Yankees eked out a win in the Division Series against the A's and then mowed down the Mariners in the ALCS and the Mets in the World Series en route to their third straight championship.
The data shows that from a predictive standpoint, a playoff spot is like a parking space: you can enter front first, or you can back into it, and in the grand scheme, it makes little difference. What I did in this 2009 study was examine the records of all playoff-bound teams from 1995-2008, searching for correlations between the records over their final seven, 14, 21 and 28 games — "weeks," essentially, though the schedule didn't always line up as such, and I wanted to ensure uniformity in my sample sizes. The correlations — which is to say, the predictive values — that I found between those segments and Division Series records were negligible and clustered around zero, ranging from .02 to -.04; keep in mind that a perfectly positive correlation is 1.0, a perfectly negative one is −1.0. Extending the latter to include the subsequent rounds, the largest correlation I found in either direction was -.12, between a team's record over the final 21 games and their postseason won-loss record. Even then, the correlation was negative, which is to say that the ever-so-slight advantage was held by teams that actually fared slightly worse, perhaps because many had sewn up playoff berths and could afford to rest their regulars down the stretch.
It's important to note that "worse" in this context still means playing roughly .600 ball over the segment in question — these are good teams, by and large — and it's worth noting that there's a selection bias at work here, in that teams that fare particularly poorly in those intervals, such as the 2011 Red Sox (8-20 over their final 28 games) or 2012 White Sox (4-10 over their final 14 games) often do miss the playoffs. Furthermore, even good teams will eventually post losing records in the playoffs, because they're facing other good teams. As strong as they may be, they can't all remain above average, like the children in Lake Wobegon.
The larger point, which I also illustrated in that 2009 BP article, is that short-term results aren't very good predictors of other short-term results. Calling upon a quick-and-dirty study I had done the year before, and again using 28-game intervals ("months"), I tested the correlation between those results and the next interval, and found correlations in the vicinity of .2 — still not much in the way of predictive value. Year-to-date winning percentages, or Pythagorean records, were somewhat more predictive, in the .3 range. That's still not much on which to hang your hat.
As the postseason begins, it's important to remember that while recent performance isn't a good predictor of what happens going forward, that doesn't mean the information is without value. The loss of key players due to injuries can have an outsized effect on a short series. Does anyone believe that the A's wouldn't be better off with a playoff rotation that features Brandon McCarthy (who actually does retain some hope of a return later this month even after a skull fracture that necessitated emergency surgery), or that the Orioles, if they make it through the wild-card game, wouldn't be better off with Nick Markakis than without him, Chris Davis' recent heroics notwithstanding?
On that note, it is worth noting that the new one-and-done wild-card format has the potential to change the equation a bit over time, because the teams that wind up there may include ones that did struggle down the stretch and lose out on division titles, such as this year's Rangers. Whereas previous wild-card teams were essentially on level ground relative to division winners, once the new format's wild-card winners expend their top starters in such "coin toss" games, they will be at significant disadvantages going into the Division Series. Still, that remains to be seen, and in a small sample of one-game playoffs, we can read just about anything we want into the results.