Sabermetricians are often accused of not enjoying the game of baseball and instead just caring about the numbers. But it’s entirely possible to love both. And in the best case scenario, the numbers can help us even further appreciate our enjoyment of the game.
A great way to experience that best case scenario is to look at a memorable baseball moment and see what the numbers show us that might give us more insight into the game. For me, the most exciting World Series I ever watched was also the first World Series I was old enough to stay up and watch: the 1991 matchup between the first two teams to ever win the pennant after finishing in last place the year before, the Atlanta Braves and the Minnesota Twins. Even then-commissioner Fay Vincent remarked, “It was I think probably the greatest World Series ever!”
Using Win Expectancy (WE), a metric that tells us the likelihood a team will win a game at any moment during that game, I decided to re-evaluate that World Series to see if the plays and moments I found most memorable were actually the most important ones in the Series. I created a new metric, Championship Expectancy (CE). CE takes the current in-game WE and combines that, using weighted averaging, with the likelihood of winning enough of the remaining games to win the series, assuming that every game later in the series has a 50% chance of going either way. The resulting percentage is the likelihood, at a particular point during the Series, that a team will win the championship.
Below, I walk through each of the games of this exciting Series and look at the major shifts in Win Expectancy and Championship Expectancy. You will find after major plays that I write in parenthesis the change in WE and CE, such as (WE: 50% to 60%, CE: 50% to 53%). I have also included graphs of the Twins’ CE after each game, with the major plays labeled. By looking at these numbers, we can better appreciate all the excitement – the obvious and the non-obvious – that occurred during a thrilling World Series.
GAME ONE
The Twins sent Jack Morris to the hill against the Braves’ Charlie Liebrandt. I remember watching that night and thinking that Greg Gagne‘s 3-run home run in the fifth, that changed a 1-0 Twins lead into a 4-0 lead, put the entire Series in the Twins’ control. But instead, that three run blast (WE: 82% to 93% CE: 60% to 63.4%) only added 3.4% to the Twins’ chances of winning it all. The Twins did go on to win the game, and now had a 65.6% CE.
GAME TWO
Tom Glavine pitched against Kevin Tapani the following night. The game was tied going into the bottom of the eighth when Scott Leius, with a career .353 SLG, homered to give the Twins the lead (WE: 61% to 87%, CE: 69.1% to 77.2%). This was the largest jump at that point in the Series. The Twins went on to win and now had a commanding 81.2% CE.
GAME THREE
Down two games to none, The Braves sent out NLCS MVP Steve Avery against Scott Erickson, who had a 134 ERA+ for the Twins in 1991. The Braves took a 4-2 lead into the top of the eighth, when Chili Davis hit a two-run homerun to retie the game (WE: 20% to 50%, CE: 73.7% to 81.2%) and eventually send it into extra innings, for the first of three times in the Series.
In the top of the twelfth, with Dan Gladden on first and no one out, the Twins’ Chuck Knoblauch reached on an error by Braves’ Second Baseman Mark Lemke, sending Gladden to third. The Twins now had a 65% WE and an 86% CE. Oddly enough, even though the Twins wound up not scoring because they had run out of hitters and their pitcher, Rich Aguilera, was forced to hit, this moment had the highest CE odds that they would have until they actually won it all.
The Braves turned the Series around in the bottom of that inning. Aguilera allowed two runners on with two outs when Mark Lemke batted. Lemke had a career .617 OPS, but in the 1991 WS, he turned into Albert Pujols, mounting a 1.170 OPS. His game-winning single lowered the Twins’ CE from 78.5% to 68.75%, which was by far the biggest change in CE in the entire series up until that point.
GAME FOUR
John Smoltz pitched Game Four for the Braves, and Jack Morris pitched his second game of the Series on just three days rest for the Twins.
This game featured a pitcher’s duel and two seventh inning solo home runs, one by Twins’ third baseman Mike Pagliarulo (WE: 46% to 68%, CE: 67.3% to 75.5%), and the other by Lonnie Smith (WE: 73% to 48%, CE: 72.8% to 65%). The game was tied at two in the bottom of the ninth when Mark Lemke, the Game Three hero, batted with one out in the bottom of the ninth and tripled to left-centerfield (WE: 42% to 18%, CE: 60.1% to 55.6%). One IBB later, pinch-hitter Jerry Willard batted for Atlanta against Steve Bedrosian and hit a flyball to right field. Jack Buck announced the play, “The runners tags at third, here’s the throw from [Shane] Mack, here’s Lemke…he is out…safe, safe, safe! They called him safe! Atlanta wins and they’re going to say Harper did not tag him!” The series was tied at two. And what seemed like a huge play was indeed a huge play, lowering the Twins’ CE from 56.8% down to 50%.
GAME FIVE
The Braves won Game Five handily, by a score of 14-5. This was the only game of the entire series that was decided by more than three runs, and the only game other than Game One that was decided by more than one run. The Braves needed to take only one of two games in the Metrodome.
GAME SIX
Steve Avery and Scott Erickson both pitched on three days rest in game six. The game went into extra innings thanks to Kirby Puckett and Terry Pendleton theatrics. In the top of the eleventh, Rick Aguilera allowed a single to Braves’ first baseman Sid Bream. After that single, the Twins had a 42% WE and 21% CE, the highest the Braves’ odds would get at any point during the series. Kevin Mitchell pinch ran for Bream, was caught stealing (WE: 42% to 56%, CE: 21% to 28%) and Aguilera escaped the 11th (WE: 64%, CE: 32%).
Kirby Puckett led off the bottom of the eleventh against Game One starter Charlie Liebrandt who now made his only appearance other than Game One. On a 2-1 count, Jack Buck made his famous homerun call, “Into deep left center, for Mitchell…and we’ll see you…tomorrow night!” Here’s where numbers and pure fandom converge. Everyone knew that was a huge hit, and the numbers agree. That hit was the biggest of the World Series at that point, raising the Twins’ CE from 32% to 50%.
GAME SEVEN
The biggest change in Championship Expectancy going into this game had been Puckett’s Game Six walk-off, but Game Seven featured four plays that had an even larger effect on the teams’ odds of winning it all. Smoltz pitched again for the Braves, and Morris pitched his third game of the Series for the Twins.
Through the first seven innings, neither team scored. By the time the last few innings of Game Seven rolled around, plays repeatedly changed the CE by more than almost any event in the first five games. Lonnie Smith led off the top of the eighth with a single. Terry Pendleton followed with the largest play of the Series so far, a double to send Smith to third (CE: 43% to 24%).
The Braves’ chances didn’t get any better that inning; instead, the Twins pulled off the biggest play of the Series, a play that is probably not remembered as one of the top five moments from the Series, if it’s remembered at all. Morris got Ron Gant to ground out weakly to first, and intentionally walked David Justice. Then Sid Bream batted with bases loaded and one out. He hit a groundball to first baseman Kent Hrbek who threw home to Brian Harper who threw back to Hrbek for the 3-2-3 double play. This play had a larger effect on CE than all four of the Series’ walk-off hits, from 32% to 61%. For reference, that 29% gain in CE was larger than Kirk Gibson‘s 1988 WS HR (+27% CE), Bill Buckner‘s 1986 WS error (-20% CE), and Carlton Fisk‘s 1975 WS HR (+18% CE). In fact, this was the biggest change in CE on a run-less play in major league baseball history.
The Twins threatened in the bottom of the eighth. Chuck Knoblauch’s one out single sent Al Newman to third with one out (CE: 61% to 75%). After Mike Stanton intentionally walked Puckett, Kent Hrbek lined into a double play to end the threat (CE: 77% to 50%). Like the Twins’ double play in the top of the inning, this double play was the biggest play for the Braves’ CE in the Series.
Morris sent the Braves down in order in the ninth. In the bottom of the inning, Brian Harper’s bunt single put two runners on with no one out (CE: 71% to 82%), but Shane Mack grounded into a double play (CE: 82% to 63%), another play that hurt the Twins more than Puckett’s Game Six walk off helped them. After an IBB, Braves reliever Alejandro Pena struck out Paul Sorrento (CE: 64% to 50%).
The Twins sent Jack Morris out for the tenth inning, and he put down the Braves in order. Alejandro Pena allowed a leadoff double to Dan Gladden in the bottom of the tenth (CE: 64% to 81%). After a sacrifice bunt and two intentional walks, Gene Larkin pinch hit with the bases loaded and one out against Alejandro Pena. The Twins’ CE was 83% at this point; Larkin swung at the first pitch. Vin Scully called the play on the radio, “Pena, right foot on the rubber. You can taste the pressure here in the ‘Dome as Alejandro straightens up. And the pitch to Larkin. Swung on, a high flyball to left center, the run will score, the ball will bounce for a single, and the Minnesota Twins are champions of the world!” The odds went from 83%…up to 100%.
TAKEAWAYS
What I found amazing about this exercise was how many of the biggest plays were not necessarily the most memorable ones. I remember Mark Lemke scoring on a sacrifice fly in Game Four, Puckett’s walk-off in Game Six, and Morris shutting down the Braves in the top of the tenth of Game Seven, but apparently the bigger plays were double plays that ended threats in Game Seven.
What’s interesting is that my five most memorable plays were not the ones that had the biggest change in CE. In the table below, I list my top five most memorable plays in the left column with the CE next to it, and the top five largest plays overall in the third column with the CE next to it.
As analysts have started looking more at Win Expectancy and value added in terms of Wins in general, it is important to keep in mind that the ultimate goal of baseball is to do what the Twins did in 1991-win the World Series. Considering the effect of brief moments on the Championship Expectancy is a more direct method to learn what we are curious about, and it is also is quite an exciting way to let the numbers behind the ballgames enhance your enjoyment of it.
Thank you for reading
This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.
Subscribe now
I found this piece difficult, because of the central irony unintentionally created by Matt when he opened with a line saying that statheads aren't all about the numbers, then proceeding to do a piece that's all about the numbers. It's a dry recitation of of moments and stats, with no character, no drama, no evocation of the scene or the people involved, and history is a STORY. The piece did come to life a bit in the section on Game Seven, and I was interested to learn about the way various big plays swung the odds of their respective Series. This would have been a
neat topic to pursue more broadly in October, just prior to the start of the playoffs or this year's World Series. In fact, this would be more interesting DURING a postseason series in the way that the chances of winning a hand is shown in broadcast poker games -- after each play you could see the way the win expectancy has changed.
I think this piece weakest because it is a historical perspective. Implementing a better version of CE and using it to report on the playoffs would be a great feature.
Thumbs up from me, despite the weaknesses. :)
There were some problems with the CE metric itself... I like the idea of CE% but I kind of got confused which team's perspective the CE% was from. I kept thinking that the Braves CE and the Twins CE should add up to 100% but as I was actually reading, I wasn't sure if that was a correct premise or not... It also seems that CE should be equal to WE in Game Seven, but since it isn't explicitly stated, some readers might miss the implication.
There were also lines like this in the Game Six Recap: "After that single, the Twins had a 42% WE and 21% CE, the highest the Braves' odds would get at any point during the series" which led me baffled... if the Twins and Braves after Game 4 were tied 2-2, and after Game 6, tied 3-3, wouldn't the Braves's CE be 50% at each instance, and thus, the highest odds for the series?
As usual, when I read Matt's articles, there are a lot of neat ideas that come to my mind for future research. How much does a bunting decision, a stolen base or a caught stealing factor into CE? If double plays affect CE so drastically, is that an indication that a team's defensive ability matters more in the playoffs? How much does being the "home team" affect CE?
To reiterate, I like the overall idea and the concept is thought-provoking even though WE (which forms the basis of CE) has been used before. Still, the presentation was a bit confusing.
Thumbs up.
The CE was the Twins the whole time at least in an effort to avoid confusion. And you're correct that WE=CE for Game 7.
I did take the Championship Expectancy concept to other World Series, but I figured that I would leave that for another article. The Sid Bream DP was the highest Championship Percentage Added moment of any defensive play in history. It was tied for 11th overall. At some point, I'll do a Top Ten Moments in Championship Expectancy article, but I loved the story of the Series, especially Game Seven, and just went with that.
One thing I'm wondering though... isn't Game 7 (which naturally involves a tied series with only one game left to play) always going to involve the highest CE shifts since both teams start at a 50% CE and one team has to end with a 100% CE (the winning team) and one team has to be the losing team with 0% CE? I guess then I also wonder if CEs for Game 7 of the 1991 series should only be compared against other Game 7s... same thing for series that end in six games, five games, and four games.
I do think the metric is interesting in and of itself, but comparing plays across games is probably an exercise that doesn't contain a whole lot of information. Instead, discussing plays in a given game (or a given context, i.e. game 5s where teams are 2-2 coming in) is more useful.
Game 7 might not matter at all (so much so that it doesn't happen in 4-0 4-1 and 4-2 Series) most of the time, but when it does, it's huge.
If the Twins' CE after that single was only 21%, then the Braves' CE would be 79%. This would be higher than the 50% mark after the 2-2 and 3-3 ties, and their highest of the series.
2000 words was too little for such a broad topic as a seven games series. The game-by-game "non-stats" descriptions were lacking in emotion throughout. Matt didn't quite succeed in capturing the feel of that fantastic '91 series.
Also, the limit forced Matt to shorthand the conclusions section, which I feel was the most interesting part. We are still continuing to learn how important outs are, and his conclusion that double plays, not homeruns, were some of the biggest turning points was a great reminder of the criticality of conserving a teams' greatest asset - outs.
Good job, Matt.
Interestingly, the homeruns and extra base hits were usually the biggest plays in World Series history. This Series was the exception, which made it all the more interesting. I think I've always loved good defense and pitching duels as a baseball fan, which is maybe why this Series has always stood out in my mind so much. That, and I find home field advantage fascinating, and the Twins going 4-0 at home and 0-3 on the road in both the 1991 World Series and the 1987 World Series makes it that much cooler.
(Of course, the other thing it shows is that nothing that happens in Game 1 is going to affect CE much. That's mostly because CE is partly a measure of how complete our information is -- Game 1 counts as much as any game, but we don't have enough info yet to know how important it will turn out to be.)
Overall, yeah, a bit dry, but good concept and reasonable execution. I've been stingy this round, but I think this gets one of my 2 thumbs up.
"Still, one of the problems I had with Matt's piece right off the bat was the subject matter of the '91 World Series. I'm not sure why I need to be reading about that in June, 2009. "
Um, maybe because that is when this competition is being held. Seriously, if you want people to do good work and show in-depth analysis, how could you even make any sort of criticism like that.
I would have loved to see Matt just break down game 7 and really pull the drama strings, but that obviously was not the scope of his article. At the very least this could have been like a 3 parter (3 games, 3 games and 1 game) but Matt appeared to do what he could within the word contraints.
But the main thing is Matt appreciates that the task of all the number-crunching is to help make the game more enjoyable. Excellent work.
Content B+ - "In fact, this was the biggest change in CE on a run-less play in major league baseball history." That is a bold declaration. It demands back-up. Did you check every World Series or did that info come from another source?
That the double plays were mathematically bigger deals than the walk off hits is interesting.
Writing B- - Otherwise this is a pretty basic sub-Goldman recount of the 1991 Series. Not at all dull, but could have been more exciting in the hands of someone who was more familiar with the nuances that could have brought the drama back to life.
I, too, was wondering what the Puckett and Pendleton theatrics were. I assumed incorrectly that you were referring to amazing fielding plays.
I remember an analysis of clutch done a few years ago by BP where they compared WE before and after certain batters with runners in scoring position over a season. By adding the differences in WE, they used that to determine who was and wasn't clutch. I would imagine something similar could be done with CE and might be interesting to look at, even with the sample size problems.
Baseball-Reference does list the biggest WPA (Win Percentage Added) for every World Series game so I went through all of them a little more quickly than I could have otherwise.
I plan to write a follow up article in October (at BP if I win or maybe at StatSpeak if I lose), given that people got as much of a kick out of the concept as I did, in which I will go through the best moments in CPA (Championship Percentage Added) in history. Some famous and some not so famous ones in the mix. Mostly hitting ones, but the 1991 big CPA plays were defense, which is another reason this Series was cool.
One of the things I would like to see if you ever expanded this article is a discussion of some dramatic plays that don't show up in win expectancy calculations. For example, Puckett's catch at the wall in game 6 looks like another out in win expectancy, but the fact that it wasn't a double is surely a huge shift. Similarly, I find myself wondering how much change in win expectancy results from Lonnie Smith getting decoyed and not scoring on Pendleton's double in game 7.
calculates Win Expectancy from historical data-- actual games event that occurred between 1977 and 2006. Using that, the difference is this:
23.5% odds of the Twins winning at that point with runners on 2nd and 3rd and 0 outs in the Top of the 8th with a tie game.
15.2% odds of the Twins winning at that point with a runner on 2nd and 0 outs in the Top of the 8th with the road team up 1-0.
Wow. I would not have expected that. The 23.5% has a small sample size issue (only 85 total games like that during 1977-2006), but still a best approximation. This is not the same way as doing Win Expectancy using the Markov Model that I believe is used on Baseball-Reference.com's listed WE's.
My only quibble about CE would be the home team should be given ~53% WE for the future games.
I don't mean to be snarky, but the "why" is that this was the topic given for this week. Matt picked an exciting moment from baseball's history and gave us a fresh analysis of it that will be useful as we watch future championships. Isn't that what good history, which was the assignment for this week, is supposed to do?
The criticism above is like saying last week "I'm not sure why Matt chose one player to focus on and write nothing else about."
My apologies to Matt, whose work has been terrific, for my putting this comment here. This is really an issue with all of the contributions this week.
Mr. Swartz's articles have thus far been very consistent and very good, but this may have been his weakest yet. Not to say it won't get my vote, but even the greatest writers have their lesser achievements.
So yes, the 2004 Series qualifies (although you are right that it was a dud), because it ended a long run of futility. But it might also be seen as the culmination of some changes in the way the Red Sox make decisions and allocate resources, and thus the beginning of a period of continued success.
The 1991 WS was the type of series that you can do that with. Game 7 was just an incredible event, one that you could watch in full over and over again on MLB Network.
Setting that criticism aside, this is still not my favorite article from Matt. This is the first article on my randomization this week, so I don't know how it compares to the others yet.
My biggest problem on readability is that I got confused by the basic construct of the WE and CE always being from the Twins' perspective. When writing about good plays by the Braves, I expected an increase in the WE and CE that immediately followed in parentheses. That disconnect made me stop reading the article and start scanning around trying to understand what was going on. Eventually, I jumped to the end and found the explanation in the comments.
The competition is getting tighter, and that was Not a Good Sign (NGS), and not a thumbs up on the first reading. This is the first week that Matt stumbles, so maybe a thumbs up will be forthcoming based on previous performance... we'll see when I'm done reading the rest of the articles.
The indicator for me is whether enough people flag it to mark it as spam/inappropriate and the comment (and every other comment in that thread) gets minimized.
(Nor, should I mention, does it seem a requirement for you in judging, as the Babe Ruth piece was not connected to a current issue, but you raved about it. Now, it certainly could be, in your opinion, a better piece because of other things. That's fine. But, it has the same fault you jumped on Matt for, which you just ignore in the Babe Ruth piece.)
Matt accomplished that, this was an article I enjoyed reading right through to the end.
There were only 10 plays in major league history that changed the odds of one team winning a World Series by a larger amount and all of them were run-scoring plays, hitting accomplishments. Even LCS games could not be more significant unless they ended in a bases loaded triple play by the road team in the bottom of the 9th or extra innings. I couldn't find an instance of that ever happening, which means that the 1991 World Series' historical significance is the biggest defensive play in the history of Major League Baseball.
Rather than open by a discussion of sabermetrics and numbers vs. enjoying the game, perhaps start us off with the oft repeated cliche that pitching and defense wins championships. You can even toss in the recent evaluations of the re-emergence of the value of defense in building winning ballclubs, like the White Sox a few years ago, or last year's Rays. And then you drop the bomb shell on us - because you're going to tell us the story of *the most important defensive play in the history of the world series*.
I'm not suggesting that the topic isn't appropriate, just that I feel you could have found a better way to engage me in it. As Goldman said, history is a story. Pretend you're Mark Twain for a few minutes and you can really make this defensive play into amazing a story.
You could've removed the markers on each line and only added them to the points you were highlighting. Also, it would have been nice to see each game's WE plotted on the graph.
I understand that these guys are doing all the work on their own and they might not be an excel charting whiz, but I think BP needs to provide SOME help prettying up the articles. We're right in the middle of the competition, isn't it at that point when American Idol starts giving their contestants new wardrobes?
If you've checked every single World Series, maybe you have this answer available.
What have been the biggest chokes/comebacks in terms of Championship expectancy? Has any team ever been over 98% and gone on to lose?
I'm just wondering how the percentage line up to the BP article a couple of years back on the biggest regular season collapses in terms of playoff probabilities.
As for Championship-thwarting defensive plays/miscues, I wonder where Mariano Rivera's wide throw on the bunt in the 9th inning of 2001 Game Seven would rank.
Great work Matt, thumbs up.
The highest ever was the 1968 World Series-- the very last before Divisional Play began. The St. Louis Cardinals had a commanding 3-1 lead over the Detroit Tigers and an 87% WE with a 3-0 lead in the Top of the 4th with men on 1st and 2nd and 1 out. They failed to score and went on to lose the game 5-3 and the Series 4-3 after Detroit beat them 13-1 and 4-1 in Games 6 & 7. The Cardinals' CE% at its highest was 96.75%.
1. No conclusions drawn. Sure, we numbers people are fans, too, but I would like some analysis, not just a numbers-assisted trip down memory lane. Some suggestions made by other posters above (it's all about the outs, or the missed opportunities). Or at least discussion of what CE adds to our understanding of the game.
2. I'm sure many will disagree, but - Too linear. Maybe I'm too accustomed to reading Christina, but there were a thousand interesting side trips to take here. Obviously you can't take 'em all, but I'd have a better idea of how you write and think if you took one or two of them: it's not a term paper. You mention that you're going to use a 50% WE for each following game, and your comments here show you're aware of the benefits and drawbacks of doing that...isn't that worth a few words? And, again, maybe I'm addicted to snark, but are you really going to leave Fay Vincent's quote unsmirked upon? It's core to BP (in my view) that such meaningless marketing drivel isn't taken at face value. Or was the smirk implicit?
3. I agree with an above commenter - you could have used twice the words to address this fully. Which suggests you might have bitten off more than you should have for this assignment.
Summary - I enjoyed it a ton, but so many questions are raised that I want more. So, pretty good.
Uhh, Will - doesn't Fangraphs do this for every game every single day?
http://www.fangraphs.com/livescoreboard.aspx
To prove I was reading critically, I will offer a few criticisms. First, you demonstrate but don't address the fact that CE is highly leveraged. A huge play in the 5th inning of game 1 just ain't like a huge play while tied in the eighth inning of game 7. Two, the graphs don't really add much to presentation. Three, your obvious Twins fanhood doesn't enhance the story for (most) anybody who isn't also a Twins fan.
Mr. Goldman's criticisms I find just plain silly. A history piece needs to tie directly to today? Does that mean I have to discard my David Halberstam books? As for this being an "October article" and not a "June article", I haven't seen a seasonal theme among the other entries (so far), and I also have not seen the same criticism leveled.
Yes, there are some flaws, both in conception and execution, but this is still, overall, a very good piece.
Go Phillies! :)
I love WE charts, and I routinely used them to follow games, but they're no longer sufficiently novel to carry an article.
It was actually Keith Mitchell.