Postseason Pendulums - Part 1 - The Greatest Comebacks

May 31, 2017
One of the ideas for an article that I’ve been toying with is to outline all of the things that make baseball different from other sports.  Obviously, each sport is unique, but baseball has so many things that distinguishes it from most other team sports.  The most iconic work on this has to be George Carlin’s classic Football vs. Baseball routine.  I can’t improve on that (no one can). 
 
Anyway, this isn’t that article….. maybe I’ll still write that someday.  But one of the things inherent in baseball that I find very appealing is that there is no clock.  Rather than minutes and seconds guiding the game, baseball uses innings and outs.  In basketball, if you’re down by 10 points with 1 second to go, the game is over.  It’s impossible to close that gap.    Same thing in football or hockey….there are deficits that, based on the score and the time remaining, are literally impossible to overcome, because time won’t allow it. 
 
However, in baseball, as long as a team avoids getting that 3rd out, it gets to keep batting.  In baseball, there’s always a chance.  A team could be down by 10 runs with two outs in the ninth inning, and it still has a chance.  Not a good one, of course….but a chance.  Just avoid making that last out, and you can keep trying.   As the great philosopher Yogi Berra said…."It ain’t over ‘til it’s over" (he was talking about pennant races, but it applies here as well).
 
Lately I’ve been messing around in what I call the Seamheads.com sandbox.  I love wallowing on that site…they have so many handy tools that are fun to just play around with and see what kind of results you get. 
 
Today, I’m looking at postseason results…..in particular, some of the greatest series comebacks ever.  I’ll cover those in part 1, and then take a look in a few days at some of the most pivotal plays in part 2, and then some individual postseason player results later on in part 3.   The articles will lean heavily (but not exclusively) on Win Probability Added (WPA) and other similar calculations that, based on historical evidence, arrive at probabilities of outcomes and the impact that players and events had on "flipping the odds" from defeat to victory.
 
Now, I’m not going to get into a primer of WPA and these other tools.  If you’re not familiar with them, there are plenty of resources on the Internet that can explain them better than I can.   In short, WPA captures the change in win expectancy from one play to another based on how much that play affects the team’s probability of winning based on historical outcomes of similar game states.   But, I think some things about these types of win expectancy metrics are important to remember as we review:
 
  1. These metrics are generally not predictive.  

    They really don’t tell us much about the inherent skill of a player.  When a player generates a game-pivoting event, it doesn’t necessarily reveal intestinal fortitude or great clutch ability.  These metrics describe what happened….not what is likely to repeat.  But, as Bill James said recently in response to a "Hey Bill" question, "Not everything is intended to predict."  Sometimes, we just want to take a look at what actually happened.
 
  1. These metrics tell the story of a game or a series

    WPA and similar metrics reveal a lot about context.  A two-run home run in the 9th inning when your team is trailing by a run may count the same in the statistics on your baseball card as a two-run home run in the 2nd inning of a blowout.  However, in the context of a specific game, the former has much greater impact on the probability of winning the game (or an entire series) than the latter does.   The game situation, the series situation…..it all matters.  These metrics are useful for storytelling, for capturing what happened, and how important events were in context of the games being played.
 
  1. These metrics can isolate key events

    WPA and other similar metrics help identify key shifts or turning points in a game or a series.
 
OK….onto the results.  What do probabilities and win expectancies tell us about some of the greatest postseason comebacks in history?
 
The Greatest Postseason Comebacks (in a series or single-game round)
 
The metric used here is called "Comeback" on Seamheads.com.  It captures the losing team’s highest win probability during the series in question.   For example, a figure of 97% means that, at some point in the series, the team that eventually lost had a 97% probability of winning the series (implying the winning team only had a 3% probability at that point in time).
 
Rank
Year
Series
Winner
Result
Loser
Comeback
1
1986
World Series
New York Mets
4-3
Boston Red Sox
99.4%
2
1986
ALCS
Boston Red Sox
4-3
California Angels
99.2%
3
2015
ALDS
Kansas City Royals
3-2
Houston Astros
99.1%
4
2002
World Series
Anaheim Angels
4-3
San Francisco Giants
98.4%
5
2004
ALCS
Boston Red Sox
4-3
New York Yankees
98.2%
6
2011
World Series
St. Louis Cardinals
4-3
Texas Rangers
97.9%
7
2003
NLCS
Florida Marlins
4-3
Chicago Cubs
97.8%
8
1968
World Series
Detroit Tigers
4-3
St. Louis Cardinals
97.3%
9
2014
ALWC
Kansas City Royals
1-0
Oakland Athletics
96.7%
10
1980
NLCS
Philadelphia Phillies
3-2
Houston Astros
96.3%
 
A lot of these are ones you would certainly expect to appear.   Quick note – the Bobby Thomson/Ralph Branca "Shot Heard ‘Round the World" game was not listed on the web site under this section, I suspect because it’s not classified there as "postseason".   However, in locating the game on baseball-reference.com, it looks like, at the end of 8 innings, Brooklyn had a 97% chance of winning.  So, if you want to include it, it would be included in the top 10 by this measure.  Of course, as much as Thomson’s blast was memorable all by itself, it was made even more so by the fact that it was the culmination of the Giants having come back from 13 games behind the Dodgers as late as mid-August.  Those two things combine to make it such a remarkable and memorable achievement.
 
The #1 reversal of fortune is the 1986 Mets/Red Sox World Series.  As I’m sure you recall, the Red Sox were up 3 games to 2, and had taken a 5-3 lead in the top of the 10th of game 6, and the first two Mets hitters were retired in the bottom of the 10th.  The Mets were down 2 runs, down 3 games to 2, and down to their last out in the series.  The probabilities state that, at that point in time, Boston had a 99.4% probability of winning the series.  Keith Hernandez reportedly went into the clubhouse and undressed, and the scoreboard briefly flashed "Congratulations Boston Red Sox, 1986 World Champions".   I was watching that game.  It certainly seemed over.
 
Then….Carter single.  Mitchell single.  Knight single.  Wild pitch.  Dribbler to Buckner.  On to game 7.
 
Interestingly enough, this comeback was only slightly more improbable than the one that happened in the round just prior to this, where the Red Sox pulled off their own unbelievable comeback against the Angels, having come back from being down 3 games to 1, and down 5-2 in the top of the 9th, with the highlight being Dave Henderson’s 2-out, 2-run HR to take the lead.
 
So, 1986 saw, by this measure, the 2 greatest postseason comebacks ever.  Quite a year.
 
Other notes:
 
The #3 comeback hasn’t had enough time, I don’t think, to work itself into baseball lore yet, as it only happened 2 seasons ago.  The best-of-five 2015 ALDS between the Astros and the Royals had Houston up 2 games to 1, and up 6-2 heading into the 8th inning of game 4, with the Astros having just tallied 3 runs in the bottom of the 7th to seemingly put the game (and the series) out of reach.  The Royals then put together an unbelievable 5-run 8th inning, plus 2 more in the 9th, to shock the Astros, forcing a game 5 that KC won handily on the way to eventually winning the championship.
 
The #4 comeback is the 2002 World Series between the Angels and the Giants.  San Francisco looked to have this locked up with a 3 games to 2 lead and a 5-0 lead heading into the bottom of the 7th of game 6.  However, the bullpen couldn’t hold off the Angels who tallied 3 runs in the 7th and 3 more in the 8th to force game 7 and eventually win the title.
 
The #5 comeback is the 2004 ALCS where the Red Sox were able to come back from a 3 games to none deficit over the Yankees.  I think many would consider that the greatest series comeback ever.  The math only slightly disagrees.
 
As a quick aside, I do think it’s an interesting comparison between the Red Sox overcoming a 3-0 deficit (98.2%) and the Mets overcoming a 3-2 deficit in games and being down 2 runs and down to their last out in game 6 (99.4%).  They’re both extremely unlikely comebacks, of course, but the Mets’ situation is empirically a little tougher to come back from.  I’m not sure that everyone would have guessed that.  I think part of that is psychological.  Perhaps it’s splitting hairs, but I think, for many, the thought of overcoming a 3 games to 0 deficit seems so much more daunting  and improbable than coming back from 2 runs down, even if you’re down to your last out.   The probabilities ever so slightly tilt towards the latter being a little more unlikely.
 
#6 is the 2011 World Series between the Cardinals and Rangers, which earned the right to be an "instant classic".  I watched in amazement as that unfolded before my eyes.  So many unbelievable moments.  As is often the case in these rankings, game 6 was the one with all the fireworks, the one that catapulted David Freese into his place in St. Louis lore.  The Rangers led 7-5 in the 7th inning, and 9-7 in the 10th, and twice they had the Cardinals down to their final out, but they couldn’t put the game away.  
 
The #7 comeback is the 2003 NLCS between the Marlins and the Cubs, highlighted by the "Bartman" incident in game 6.  Enough said.
 
The #8 comeback is the first series I ever remember hearing about, the 1968 clash between Detroit and St. Louis.  The Cardinals were up 3 games to 1 and jumped out to a 3-0 lead in game 5.  In addition, they also had their ace-in-the-hole, Bob Gibson (who was looking quite unhittable), lurking in the background if needed.  However, the Tigers fought back, forced a game 7, and were able to complete the comeback.
 
The #9 comeback is a bit of an oddball in that it was just a one-game, winner-take-all wild card, the 2014 contest between the Royals and the A’s.  Oakland was up 7-3 heading into the 8th inning, but the Royals kept pecking away, tallying 3 in the 8th and then 1 more in the 9th to force extra innings.   The A’s took the lead in the top of the 12th, but the Royals countered with 2 more in the bottom of the inning to take the game and advance to the next round.
 
The #10 comeback is the 1980 NCLS between Philadelphia and Houston, which I think has kind of gotten overlooked in the passage of time.  This was a classic.  Houston was up 2 games to 1, and had a 2-0 lead heading into the top of the 8th inning in game 4, and a 4-2 lead heading into the top of the 8th inning of game 5.  In both games, however, the Phillies mounted comebacks, on the road, in the top of the 8th inning, scoring 3 runs in game 4 and 5 runs in game 5 to take the lead in both of those games, only to have Houston come back and force extra innings.  4 of the 5 games in this series went extra innings.  A great, classic, and underrated series.
 
 
Others that just missed (but were still well over 90%):
 
2012 NLDS – Cardinals over Nationals 3-2 (Cardinals come back from 6-0 and 7-5 deficits in game 5 by scoring 4 in the 9th inning)
1995 ALDS – Mariners over Yankees 3-2 (Joey Cora and Ken Griffey Jr. score on Edgar Martinez double)
1960 World Series – Pirates over Yankees 4-3 (Hal Smith and Bill Mazeroski HR’s)
2003 ALCS – Yankees over Red Sox 4-3 (Aaron Boone HR)
1992 NLCS – Braves over Pirates 4-3 (Francisco Cabrera hit)
 
By the way, last year’s World Series ranked 18th with a 92.3% figure, as the Indians were up 3 games to 1 and (briefly) had the lead in game 5 before the series started to turn towards the Cubs.
 
In the Part 2, we’ll look at isolating some of the top, pivotal, game-changing postseason plays.
 
Thanks for reading.
 
Dan
    
 
 

COMMENTS (18 Comments, most recent shown first)

Marc Schneider
Would you put any weight on the quality of the player involved? In other words, how bizarre was it that someone like Francisco Cabrera in 1992 would get the hit with two outs to win the game? I guess I'm asking is the win probability affected by who is at bat or on the mound? The Mets were down to their final out but had the middle of the order coming up. I'm not arguing that the Braves win was more improbable than the Mets-the Braves already had the bases loaded, were only one run behind, etc-but the Mets at least had good hitters coming up.
12:22 PM Jun 16th
 
DMBBHF
Hi Julesig,

Thanks for the kind words.

I believe WPA is measured at the per at bat level.

Also, the Cabrera play doesn't rank higher because the probabilities are a little more in the Braves' favor than some of the other situations, when looking at the series as a whole (not just the one game). For most of that series, the Braves were up (3 games to 1, then 3 games to 2, before being down most of game 7). When Cabrera came up, the Braves were down a run, but the bases were loaded, and all he needed was a single to win.

There is another related metric, Series Win Probability Added (SWPA), which is the change in SERIES (not game) Win Probability Added on a single play. The Cabrera play actually ranks #1 by that measure.

Thanks,
Dan
8:48 AM Jun 11th
 
DMBBHF
Hi Brock,

I suppose it is. I'm not real well versed in Markov chains, but I think it may be in that category.

Also, I replied to an earlier observation that was made about the lack of early seasons on the list, and I think it's mostly a matter of there being so many more postseason games each year nowadays, where as pre-1969 you only had the 2 teams that matched up in the World Series. If you run the same analysis but only include World Series matchups, you start to see more of the older seasons emerge.

Thanks,
Dan
8:29 AM Jun 11th
 
julesig
This is a super interesting article, thanks for writing it.

I'm surprised the Francisco Cabrera series didn't rank higher. Is the WPA measured at a per at bat level or at the per pitch level? Like the 86 Mets, Cabrera got two strikes on him, fouled on off, then hit the single. I've always been struck at how dramatic those pitches were because one swing changes the entire outcome. If he misses, Pirates win, if he gets a hit, Braves win because two runs score to overcome the 1-run deficit. Usually the big hits either tie (e.g Rajai Davis HR last year) or break a tie, but not flip the outcome entirely.
10:42 AM Jun 6th
 
Brock Hanke
Dan - Is it safe to assume that WPA, like all the win probability methods, is essentially a Markov Chain analysis? I don't see how it could be anything else.

Also, I noted the lack of early, much less REALLY early seasons on the lists. My guess would be that the difference in quality between teams increases the further you go back intake, which increases the probability that the better team will win game 1 and decreases the probability that the weaker team can stage a comeback.

This is a factoid just a bit off-topic, but one of the WS I looked at was the 1929, because of the 8-0 lead being lost. However, the team that lost that lead was only down 2-1 in games, and would lose the WS in five games. The factoid, which seems like it should have been monstrously important, was that Lefty Grove did not start even one game. He pitched 6.1 innings, gave up no runs, and got one save, but did not start, even once. Were they really THAT afraid to start a lefty against Rogers Hornsby? What would be the Starting Win Probability be if you factor in that Lefty Grove will only pitch 6.1 innings?
10:25 PM Jun 4th
 
DMBBHF
Steve,

Your computer should have crashed just as easily with the Mets' situation as it did with the Red Sox being down 3-0. As far as I know, no one else other than the Mets were down 3-2 and 2 runs down and down to their last out with no one on base. Both are historically unique. Why are you so convinced that the Red Sox situation is the more unlikely comeback scenario?

The Cardinals-Rangers setting was close to that situation, but a key difference (enough to explain a 1.5% difference) is that, before the Cardinals made their 2nd out in the bottom of the 9th, they had runners on 1st and 2nd. The Mets had 2 outs with no one on before they started their comeback. That's enough to explain a 1.5% difference in probability. And that's all these figures are reflecting, just the little nuances between different situations.

Look at it another way. What was the probability that the Red Sox would win a single game against that Yankees team? 50%? 45%? Which figure would you use? If it's 50%, the probability of winning 4 in a row would be .5x.5x.5x.5, which would be about 6%. If you use 45%, it's .45 to the 4th power, or about 4%. If you use 40%, it's about 2.6% of the time. So, winning 4 in a row is low probability, but it is possible.

And, I will concede (and said so in the article) that being down 3-0 in a playoff carries a psychological challenge as well. But, these are just probabilities, after all. They don't capture everything....they merely try to re-state the odds. Take them as you wish.

Thanks,
Dan


10:43 AM Jun 4th
 
steve161
Dan, I take your point but I think you're placing too much trust in an algorithm the details of which I suspect no one knows outside of its inventors. How do you compare what the Red Sox did with anything that had happened historically? Sure, plenty of teams have won four games in a row, but none of them did it in a championship series against an opponent of comparable ability that had already beaten them three times. The denominator of that fraction is zero, and I don't have to tell you what happened to my software when it attempted to divide by zero.

Likewise, do you really take seriously the 1.5 percent that separate Mets-Boston from St Louis-Texas? The Cardinals were down to their last strike and TWICE trailed with two outs in the final at bat. This list is ordered by the highest probability of losing, but does not take into account that there were, in this instance, two such occurrences within the same game. If you're trying to declare something the greatest comeback ever, it seems to me that that ought to count for something.

In other words, the conclusion is driven by the methodology. It's fun to assemble a list like this, but to think that it really tells us anything worth believing is naive.
4:43 PM Jun 3rd
 
DMBBHF
Hi Steve161,

Thanks for the comments.

You stated that: i"Intuitively, does anybody buy that coming from two runs down in the ninth (which has happened countless times) is less probable than coming from 3-0 down in a best of seven (which had never happened)?"....

Well, that's not exactly the situation. In game 6, the Mets were 2 runs down in the 10th, not that 9th, but that's just me being picky, and that's not the important point anyway. :)

The really important points, though, are that they weren't just down 2 runs.....they were:

a) down 3 games to 2
b) 2 runs down
c) down to their last out

When Hernandez was retired for the 2nd out in the bottom of the 10th, Boston had a 99% chance to win (not just the game, but the series), because teams in that situation in the past (not just in postseason, but in all games) had won that often when in that situation. That's the empirical evidence on the matter.

The Mets not only had to come back from that deficit to win that game, but they also had to win the next game as well. The 99.4% probability captures the odds of both of those things happening (not just the comeback in game 6, but coming back to win game 7 as well. The comeback % is a series win probability, not just the isolated game.

Plus, it's not really accurate to say that it has happened countless times. I think, at least in the postseason (which is what counts, since you're comparing it to the 3-0 deficit Boston faced), that's the only comeback like that. The '86 Red Sox over the Angels is close, but Boston had 1 out in the 9th when Baylor hit a 2-run HR to start to close the gap.

The %'s just are what they are, and they conclude that coming back from the Mets' situation to win both that game AND the next one is just a little less likely, based on all the historical data we have to conclude such things, than coming from 3-0 games down to win 4 in a row. It's not much of a difference at all.

Oh, and I presented it with the first decimal point because that's how Seamheads.com presented it, and I didn't see any reason to alter it.

Thanks,
Dan


9:16 AM Jun 3rd
 
DMBBHF
Hi Owen & Steve161,

I think the fact that most of the series comebacks on this list are fairly recent is distorted by the fact that there are so many additional rounds of playoffs that exist now, so there are more opportunities for great "comebacks". 6 of the 10 listed are either wild card, LDS, or LCS rounds. Only 4 of the 10 were World Series.

If we limit the comebacks to World Series only, so that all years would have equal opportunities, the 10 top would be:

1986-99.4%-New York Mets over Boston Red Sox
2002-98.4%-Anaheim Angels over San Francisco Giants
2011-97.9%-St. Louis Cardinals over Texas Rangers
1968-97.3%-Detroit Tigers over St. Louis Cardinals
1960-94.1%-Pittsburgh Pirates over New York Yankees
1979-93.0%-Pittsburgh Pirates over Baltimore Orioles
2016-92.5%-Chicago Cubs over Cleveland Indians
1985-90.9%-Kansas City Royals over St. Louis Cardinals
1958-90.9%-New York Yankees over Milwaukee Braves
1925-89.7%-Pittsburgh Pirates Washington Senators

If you extend the list even further, then you get even more series that date back a ways:

1997
1987
1924
1975
1952
1981
1921
1903

So, I do think there's a recency bias in terms of the # of playoff rounds.

Thanks,
Dan
8:38 AM Jun 3rd
 
OwenH
Steve, you could certainly be right. Intuitively, it does seem like there have been a lot of amazing postseason comebacks in the past couple/few decades. I haven't been watching baseball long enough to say if that's significantly more true now than, say, the 50s or the 60s, but it does seem like there have been a lot in recent years. Which is good and exciting, of course.
5:04 PM Jun 2nd
 
steve161
Owen, my decidedly non-random speculation is that the problem is with the metric. On the other hand, if I'm wrong and this is indeed a relatively recent phenomenon, then how much more interesting has baseball become, where late leads are less likely to be sure things.
11:57 AM Jun 2nd
 
OwenH
How interesting that almost all of these highly improbable comebacks took place in recent decades. There are only a couple that are even from the 1960s, nothing before that, at least at the top of this list (in the high 90%s).

Random speculation, I wonder if the trend has anything to do with the advent of modern bullpens? It seems rather counter-intuitive; bringing in your top relievers surely improves your chances to win, rather than leaving in a tired starter, but who knows? Not always.
1:19 PM Jun 1st
 
bhalbleib
Mauimike, I am sure that somewhere in the past we can find a player who, at least in part, took his life because of some failing on the field. The only similar story that comes to mind to me is Willard Hershberger of the 1940 Reds. He was their backup catcher and then Ernie Lombardi got hurt. Here is what his player bio from baseball reference.com says:

"An injury to Lombardi and a slump in late July, while not dislodging the Reds from first place, caused concern. With Lombardi out of action, Hershberger caught nine of the next 11 games, five of them losses, and blamed himself for every loss. Particularly galling to the sensitive young man was a loss to the Giants on July 31. Reds’ pitcher Bucky Walters had a 4-1 lead going into the ninth inning, but couldn’t hold it. Harry Danning won the game for the Giants on a home run off a fastball called by Hershberger. “I called for the wrong pitch to Danning,” the despondent catcher insisted. “It was my fault.”

There's more at his bio, but the story ends with his lifeless body in a bathtub.

The Reds ended up winning the pennant and the Series too that year (their only clean Series win until the Big Red Machine in 1975)
11:45 AM Jun 1st
 
Steven Goldleaf
Gotta be a few extra points for the 1986 Mets being down not only to their final out, but their final strike several times in Game Six.
10:05 PM May 31st
 
steve161
Very entertaining article, Dan, which is not going to stop me from expressing a couple of quibbles.

1) Rendering the percentages to the first decimal point is what photographers call 'empty magnification'. I don't even buy the accuracy of the ordering, much less a distinction of two tenths of a percent. Intuitively, does anybody buy that coming from two runs down in the ninth (which has happened countless times) is less probable than coming from 3-0 down in a best of seven (which had never happened)? If that were my site, I'd rethink the algorithm. Since it isn't, I'll be content with my grain of salt.

2) If a comeback doesn't enter the lore of the game immediately, it never will. Unfair to KC-Houston, no doubt.
3:10 PM May 31st
 
Brian
That should read the Mets "went down" 3-0 in game 7
7:29 AM May 31st
 
Brian
As I recall, some of those teams got into trouble again, after escaping their lowest point. In 1986, the Mets went 3-0 in game 7. Also in 1986,after the Henderson HR, I believe the Angels got a runner to 3rd with less than 2 outs with a chance to win in walk off fashion. As you mentioned, in 2011, the Cardinals were down to their last out not once but twice in their series.
7:28 AM May 31st
 
mauimike
I was very involved as an Angel fan in two of those series. 1986. Henderson - Moore. Donnie never got over it. Suicide. Has there been another player who took his life after a failure. 2002. Down 5 zip in the 6th game, I never had a doubt that the Angels would win the game and the World Series. I was 51 at the time and should not have believed in miracles.
1:34 AM May 31st
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy