(The following is being syndicated from The Captain’s Blog).
The Red Sox’ bandwagon has slowly begun to empty as the team struggles through a nightmarish stretch during which it has lost six games to Tampa in the standings. Fortunately, those abandoning the Boston ship have found passage on the Tigers’ pennant drive. Over the last three weeks, Detroit has won 16 of 20 games and built an insurmountable lead in the Central.
If the season ended today, the surging Tigers would meet the slumping Red Sox in the ALDS. Judging by the trending conventional wisdom, this clash of teams going in opposite directions would have Red Sox Nation tuning into the Patriots a lot earlier than expected, but should we rely on September records when handicapping post season series in October?
In a 2009 Baseball Prospectus post, Jay Jaffe looked at winning percentages over defined periods in September and found no meaningful correlation to various measures of post season success. However, the study did not distinguish between the relative records of opponents, nor any relationship between overall winning percentage and the rates compiled over the defined periods. In addition, Jaffe’s study was limited to post season series played since 1995. So, as a supplement to his work, the following incorporates the two aforemetioned variables into a breakdown of every first round post season series since 1969.
First Round Post Season Outcomes, Based on Relative Opponent Records (click to enlarge)
Note: Includes all first round post season series since 1969. Ranges refer to the winning percentage disparity between teams. A negative range indicates that a team with a lower record won the series.
Source: Baseball-reference.com and proprietary calculations
There have been 116 first round post season series since 1969 (LCS from 1969 to 1993 and LDS in 1981 as well as from 1995 to 2010). Of that total, 65 series (or 56%) were won by teams with the better regular season record, while in 58 (or 50%), the victor was the team with a better winning percentage in the final month. According to this data, the hottest team’s chances of winning in October were no better than a coin flip. However, if we assume those teams tend to be inferior (i.e., they have to win just to make the playoffs, while better teams have the luxury of resting their players down the stretch), then perhaps a 50/50 chance actually represents an enhancement?
One way to test this theory is by combining our comparison of winning percentages for the season and final month. In 69 first round post season series, the team with the better regular season record also entered October as the hotter side. However, only 38 teams of those teams actually won the series, a percentage that is almost equal to the rate associated with having a better regular season record. This implies that being hot doesn’t provide a marginal advantage. What’s more, of the 47 occasions when the better regular season and September records were split, the team with the 162-game advantage won the series 27 times (57%).
Because a significant number of post season series take place between teams with relatively comparable records, the data can become skewed. In order to overcome that distortion, the sample of post season series was refined based on degrees of disparity in winning percentage over both the full season and final month.
First Round Post Season Outcomes, Based on Range of Relative Opponent Records (click to enlarge)
Note: Includes all first round post season series since 1969. Ranges refer to the winning percentage disparity between teams. A negative range indicates that a team with a lower record won the series.
Source: Baseball-reference.com and proprietary calculations
As evidenced by the charts above, a greater disparity in regular season record increasingly correlates to a first round victory by the better team. However, being hot once again seems to have no impact. Not only was the cumulative split between “hot” winners and losers close to 50/50, but the even distribution persists at the most extreme parameters. Based on this evidence, it seems as if being hot really does have little-to-no impact on success in October.
Before dismissing the benefit of entering October on a hot streak, one more data segment was considered: those series in which there was a less than 25 point difference between the opponents’ full season winning percentages. Among this subset of 54 series, only 24 were won by the “hotter” team. What’s more, when increasing the granularity in this segment to teams with a September winning percentage disparity of at least 100 points (i.e., series with a small full season disparity, but significant September divergence), the ratio only returns back to 50%. In other orders, being hot doesn’t even make a difference among teams with very similar full season records.
Top-Five Upsets* in a First Round Post Season Series, Since 1969
| Year | Series | Winner | W% | Loser | W% | Full Diff | Sept. Diff |
| 1973 | NLCS | Mets | 0.509 | Reds | 0.611 | -0.102 | -0.138 |
| 2008 | NLDS | Dodgers | 0.519 | Cubs | 0.602 | -0.084 | -0.103 |
| 1987 | ALCS | Twins | 0.525 | Tigers | 0.605 | -0.080 | 0.165 |
| 2003 | NLDS | Cubs | 0.543 | Braves | 0.623 | -0.080 | 0.180 |
| 1996 | ALDS | Orioles | 0.543 | Indians | 0.615 | -0.072 | 0.011 |
*Based on greatest disparity between opponents’ full season records.
Source: Baseball-reference.com and proprietary calculations
Finally, if you look at the top-five upsets based on disparity in regular season records, only two involved an underdog with a significantly better winning percentage in the final month. Even more confounding, in the 10 most extreme instances in which the team with the lesser September record won the first round series, all 10 also had the weaker regular season record. For example, when the Athletics met the Yankees in the 2000 ALDS, not only did Oakland have a better regular season record (.565 to .540), but the team also compiled the largest September winning percentage differential (.759 to .419) since the advent of divisional play. Despite these two advantages, which extended to winning the first game of the series, the Yankees were the team that advanced to the ALCS.
If there is an advantage to entering the post season on a roll, it isn’t evident in the series outcomes. Contrary to popular misconception, baseball isn’t a “what have you done for me lately” sport, and momentum only exists if defined by the entire 162-game schedule. So, if the Red Sox and Tigers do meet in October, don’t be fooled by September when making your prediction. A coin flip would probably serve you just as well.
If you enjoyed this article, please consider sharing it!
Does this cover the question?
Does “record in September” sufficiently deal with “being hot”? I don’t think that when people say “hot down the stretch” they necessarily want to include teams that have played .540 or .550 ball over 30 games, which would cover teams that go 17-13. Is there enough data to isolate, say, teams who have played .700 ball over any stretch of last games, say 10 or more?
One of the charts above covers a range of every winning percentage in September, so you are free to consider hot by whatever winning percentage you choose. What’s more, there is a link provided that looks at the correlation between winning in the post season and shorter periods before the end of the season.
Of course, I don’t see value in looking at how a team plays over a sample as small as 10 games. Not only is the W% skewed by just one game, but there is a greater chance that playoff teams have clinched by that time. For that reason, I think looking at the entire month of September makes more sense.
That’s a little like the old joke of looking for something under a lamp because that’s where the light is best. It’s true that sample sizes of 10 games aren’t very meaningful–but that’s what we’re looking for. We’re looking for what the public perception of hot is, and I don’t think it’s necessarily the amount of winning percentage better than your eventual playoff opponent in the month of September.
Sorry, that was kind of snotty and didn’t really get at what I mean. Probably statistically what I’m thinking of just isn’t valid, which would mean it isn’t really appropriate. Let me see if I can formulate what I’m saying better.
No offense taken at all, but between the study Jay Jaffe did and the different ways I looked at the relationship, there just doesn’t seem to be any correlation between how you play late and performance in the 1st round.
If you can think of a test, I’d be happy to run the numbers if it falls within the data I already compiled(it took several hours to build the data base, so I can’t make any promises).
I just hit a Jayson Stark column from Tuesday that did a similar thing to what I was asking–.7000 record, but not over just 10 games, over 30 or 40 games–incidental to an article on the Tigers recent hot streak. They came to the same conclusion you did.
http://espn.go.com/mlb/story/_/page/rumblings110913/detroit-tigers-roaring-postseason
Here are the three relevant paragraphs:
• Since the dawn of the wild-card era, 14 postseason-bound teams have finished a season by playing at least .700 baseball (28-12 or better) over their last 40 games. Exactly one of them — the 2004 Red Sox (28-12) — wound up winning the World Series. But five of them lost in the first round. And of the 27 postseason series those teams ended up playing, they barely won more series (14) than they lost (13).
• Meanwhile, 22 postseason-bound teams in the same era have finished a season by playing at least .700 baseball (21-9 or better) over their final 30 games. Just two of those 22 went on to win the World Series — the 2002 Angels (21-9) and 2003 Marlins (21-9). But eight of them lost in the first round. And those teams went just 21-20 in their 41 postseason series.
• On the other hand, seven teams stumbled into the postseason after playing sub-.500 baseball in their final 40 games. You’d think that would be a formula for disaster, right? Guess again. There were more World Series won by those teams (two) than by teams that played .700 baseball over the same stretch. The two sub-.500 teams that won it all: The 2000 Yankees (19-21) and 2006 Cardinals (18-22). And the overall record of those bumblers and stumblers in their 16 series (11-5) was also better than the record of those “hot” teams.
I find this a more convincing “coverage” of the question–that’s probably just because I don’t have a feel for winning percentage of one team vs their opponent–but I also notice something else.
To say that 5 of 14 teams playing ,700 baseball over 40 games lost in the first round is to say that their first-round record is 9-5.That’s a VERY good record in baseball terms, especially for playing only other good teams. Similarly, for teams playing .700 over 30 games, 8 of 22 lost in the first round, but once again, that’s a terrific record: 14-8. The problem, of course, is that it’s a very small sample size, so it doesn’t prove anything. But it also doesn’t disprove that maybe being hot helps in the first round.
Anyway, between your article and Stark’s, I’m pretty much going to have to give up another cherished myth.
It was a good piece, but at the risk of being immodest, I think mine is “more convincing” because it goes all the way back to 1969, and not only looks at September records (so about 25-35 games), but compares it against both regular season records and other teams Sept/regular season records. That is much more comprehensive than the more anecdotal look taken by Stark.
The reason I think it makes more sense to do a comparative analysis is because there are lots of examples when two teams have very high winning percentages over a defined stretch. In such cases, one will lose and one will win, so if looked in isolation, the results become distorted.
Either way, I think it seems clear that being on a role heading into the post season is more or less a myth.
There’s always that statistical thing, as I understand it: the more data you get, the the less the chance of randomness, but the less specific the situation you’re covering. To me, and I’m no statistician, yours is a more definitive answer, but to a more specific question than I, for one, have in mind.
It’s like the question of bunting. There is just no question that on average, the bunt is a bad play. But that doesn’t prove that there aren’t specific situation in which it’s a good play.
I agree with you, and I was trying to find even a very specific situation when “being hot” correlated to first round success. I just couldn’t locate one.
[...] to three games, and the AL East rivals will square off one final time Sunday in Fenway Park. …Does Success in September Carry Into October? The Yankee AnalystsBeckett helps Sox defeat Rays, extend wildcard lead Suffolk Voiceall 3,039 [...]