Do September Streaks Correlate to Playoff Success or Failure? Don't Fall for Recency Bias

The Indians are on a historic winning streak while the Dodgers are having their most dreadful month in years. Will that matter once the postseason starts?

Thus far in September, streaks have dominated the baseball news, namely the Indians' record-tying 21 consecutive wins and the Dodgers’ 11 straight losses, with the Diamondbacks’ franchise record 13-game winning streak not to be overlooked. All three teams are on track for the postseason, leading to plenty of speculation as to what the past few weeks will mean for their October fates.

However, according to a 2009 study I did at Baseball Prospectus that has been frequently cited—including, coincidentally, during ESPN's Giants-Dodgers telecast this week just as I was putting this updated version together—there is virtually no correlation between the performance of wild-card era playoff teams in September and October. 

That study examined 112 playoff-bound teams from 1995 to 2008 and how each performed in their final seven, 14 and 21 regular season games as well as those from Aug. 31 to the end of the season, which sometimes trickles into the first week of October (for the rest of this piece, I'll refer to this as September records or September winning percentage despite the spillover). At each of those levels, what I found was an essentially random relationship between the winning percentages in those fragments and the results of the Division Series.

Thanks to @BoogSciambi for shoutout on Wed.'s ESPN Dodgers-Giants 'cast. Update of my September study today @si_mlb https://t.co/Kzc4S3CKWR

— Jay Jaffe (@jay_jaffe) September 14, 2017

The largest positive correlation was .019 between the final seven games and the first series. The largest negative was -.042 between the final 21 games and the division series results. That's a whole lot of nothing, and when I expanded to include the postseason as a whole, as measured by number of series won, there was a small but consistent inverse correlation between September winning percentages and October success; for the post-Aug. 31 records and the number of series won, the correlation reached -.112. That's still virtually meaningless, but it did tell us that that the higher a team's winning percentage was after Aug. 31, the less likely they were to go all the way.

MLB Playoff Tiebreaker Scenarios: What Would A Five-Team (Or More) Tiebreak Look Like?

Obviously, that's a counterintuitive finding. At the time, the 2007 Rockies' run was a fresh memory; that team had won 13 of its final 14 games, including a Game 163 tiebreaker, and stormed all the way to the National League pennant before being swept by the Red Sox in the World Series. What had been forgotten was that the 2006 Cardinals had dropped nine of their final 12 regular season games before going on to win the World Series. The 2000 Yankees had gone 3-15 over their final 18 games yet still won their third straight championship. The lesson of the study was in the article's title: "The Perils of Relying on Short-Term Memory," which is to say that it's a fool's errand to assume that what happens in September predicts what we'll see in October.

It’s been eight years since I ran that study, and between the time elapsed, the expansion of the playoffs and the aforementioned streaks, I decided that it was worth revisiting. Spoiler alert: there's still virtually no correlation, which is to say no predictive relationship, between September and October.

Category	With Two Wild-Cards	Without TWo Wild-cards
Full Season % to Postseason%	0.24	0.21
Full Season % to Postseason Win Total	0.17	0.14
Full Season % to PSS	0.15	0.13
September % to Postseason %	0.09	0.08
September % to PSW	0.03	0.02
September % to PSS	-0.01	-0.02

There's a lot to unpack in this table, so bear with me. The first rows compares the correlations between a team's full season winning percentage over 162 games (give or take one due to tiebreakers and rainouts), abbreviated as Full%, and its postseason winning percentage (PS%). The second looks at the correlation between the full season winning percentage and its postseason win total (PSW; it's all about getting to 11 wins in the Division Series, League Championship Series and World Series, no matter how many times a team loses). The third shows the correlation between Full% and a club's postseason series win total (PSS).

For each, there are positive but small correlations between full season records and the postseason. The take-home is that the teams with better regular season records tend to do slightly better in October, whether I include the 10 teams that lost in the 2012-16 wild card games (referred to here as WC2, whether or not they had worse records than the teams that beat them in those winner-take-all games) or consider only that contest's winner and thus a uniform eight teams per season from 1995-2016.

The second set of three columns show the correlations between "September" winning percentages (remember, that's after August 31, perhaps including a few games in October) and postseason success, again via the three different measures of postseason winning percentage, postseason win totals, and postseason series win totals. They're all much smaller than the full season correlations, and very close to zero. If we had correlations of .7 or even .5 in either direction, we might be on to something; here, we emphatically are not. There is no predictive relationship to be found; if all you know about a playoff team is that they played .650 ball in September, you have no cue as to whether they were more likely to be swept in the Division Series or to win it all, and the same goes for if they played .450 ball.

Note that there is a selection bias at work here, in that at a certain point, a playoff-caliber team might play so badly that they miss the postseason. The 2011 Red Sox went 7-20 in September, for a .259 winning percentage, and were ousted from the AL wild card race on the final day of the season. The lowest September winning percentage of any wild card-era playoff team was the 1998 Padres' .375 (9-15)—yet they made it all the way to the World Series before being swept by the Yankees. This year's Dodgers, 3-11 (.214) so far in September, clinched a postseason berth via the Tuesday night win over the Giants that snapped their 11-game streak. Unless they win nine of their final 16 games, they'll supplant the 1998 Padres as the low team here, but that doesn't mean that we know what they’ll do next month.

Other ways of examining the data yield their share of surprises. For example, as a group, the teams that won pennants or World Series did collectively worse in September than the group of playoff teams as a whole.

Group	#	Thru Aug 31	Sept.
All PS teams	186	.577	.595
WS win	22	.589	.585
Pennant win	44	.588	.587
Non-pennant	142	.576	.597

Collectively, the 186 teams (including the WC2s) won at a .577 clip through August and did a bit better in September; the 18-point difference in winning percentage equates to half a win over the course of 30 games. But the World Series and pennant winners didn't experience such gains in September; in fact, they did slightly worse relative both to their own pre-September performances and to those teams eliminated before the World Series.

What's more, a total of 26 teams made the postseason in this span despite sub-.500 records in September; collectively, they went from .595 through August 31—higher than even the group of World Series winners—to .447 for September. Yet of those 26, nine won pennants, a 35% success rate at reaching the World Series, compared to a 22% success rate for the field as a whole. Cross the three WC2s in that sample off the list, and all 10 from the field as a whole, and the percentages are 39% and 23%. Four of those teams, namely the 1997 Marlins (.444 in Septmeber), 2000 Yankees (.419), 2006 Cardinals (.414) and 2015 Royals (.469), won the World Series. Excluding the WC2s, that’s a 17% success rate versus 11% for the field as a whole.

Looking at it another way, and excluding the results of the 2012-16 wild card games, the 23 sub-.500 teams won a total of 25 postseason series, an average of 1.09 per team (note that I'm not counting winning the wild card game as a series win). For the 176-team group, there were 154 series won (88 LDS, 44 LCS, 22 WS), an average of 0.88 per team. The 23 sub-.500 teams went 109-99 in the postseason overall, for a .551 winning percentage, compared to the automatic .500 as a group.

MLB Playoff Tiebreaker Scenarios: What Would A Five-Team (Or More) Tiebreak Look Like?

On the other side of the coin, there were 24 wild card-era postseason teams that posted winning percentages of at least .700 in September, one of which was the 2013 AL wild card game loser, the Indians (.778 in September). The remaining 23 teams combined to win only 17 postseason series (0.74 per team) and three pennants. None of them—repeat, none of them—won the World Series. The top two teams, the 2001 A's (.852) and 1995 Yankees (.786), lost in the Division Series, while the 2002 Cardinals (.778), lost the NLCS. The top September team to reach the World Series, the 2011 Rangers (.760), ranked seventh among the 176 teams, meaning that the six above them all fell short of the Fall Classic.

Without examining each team more closely, we don't know what circumstances caused their red-hot, mediocre, or terrible Septembers. Considering the former, it's possible that teams peak too early and tire themselves out. Conversely, for the laggards, one explanation may be that they may have eased up in the season's final weeks, resting their regulars with an eye towards keeping them fresh for the playoffs. That certainly seems to be at least part of the explanation for the Dodgers' slide. Manager Dave Roberts, whose team at one point looked as though it could challenge the 2001 Mariners' 116 wins, has the luxury of experimenting with lineups, rotations and bullpens right now while avoiding pushing his banged-up players too hard.

Here it's worth noting that regardless of their September records, what this study won't catch is changes in player availability. The Dodgers' September record is virtually meaningless, but it would be foolish to say that their chances at winning the World Series would be the same if they were to lose Clayton Kershaw (who recently missed five weeks with a lower back strain) or Corey Seager (who was limited to pinch-hitting duty for nearly two weeks due to loose bodies in his right elbow) for October. That Alex Wood and Yu Darvish, both of whom have had recent stays on the disabled list, have scuffled is important information as well, as is Tom Verducci's observation of their offense's ongoing vulnerability to high fastballs. But all of those things are specific to their case, and flipping the results of a few of the games during their 1-16 slide wouldn't change that.

Team Entropy Arrives: The American League Playoff Picture Is Wonderfully Cluttered

As for the Indians, it's trivial that neither of the other teams to win at least 20 straight games, the 1935 Cubs or 2002 A's, won the World Series, and just two of the six with win streaks of at least 18 games—the 1947 and '53 Yankees—did so (the 1904 Giants refused to play the World Series, which had been introduced just the year before). That doesn't mean that the Indians haven't boosted their confidence when it comes to another postseason run, or that manager Terry Francona hasn't learned anything from the tremendous performances of Trevor Bauer, Carlos Carrasco and Mike Clevinger during the streak. But even if they'd sprinkled a few losses into their last 21 games, their shot at winning it all would be unchanged.

The Indians and Dodgers could contradict my findings if the former won the World Series and the latter were eliminated in the first round. Even so, those results would be outliers. With 16 games left to play for both teams, I plugged in a 30-0 September record and 11-0 postseason record for the Indians, then did the opposite for the Dodgers, 3-27 for September and then 0-3 in the Division Series. Together, those moved the correlation between September winning percentage and postseason wins from .03 to .09, and between September winning percentage and postseason series wins from -0.01 to .06. That's still essentially random, and those are at the absolute extremes. Even if the two teams fulfilled such destinies, my inclusion of the noise-inducing performances of the eight other postseason teams would reduce the impact of those two outliers.

There’s little doubt that we’ll still be talking about the Indians and Dodgers as October approaches. The extremes of their September team performances may tell us something about their strengths and weaknesses, but the underlying lesson of the playoffs is that in a short series of five or seven games, anything can happen. And that’s why we tune in.