How to Pull Off a World Series Upsetby Chris Jaffe
October 15, 2007
It's that time of the year again. The leaves are changing colors, the weather is getting colder, and the Cubs are no longer playing. Clearly the World Series is upon us. This sparks some classic debate among experts around every water cooler in the country about what sort of team should be expected to really perform at its best at this time of year.
Last year, with the stunning success of the barely .500 St Louis Cardinals, one particular question gained prominence among baseball fans: what sort of team is most likely to pull off a massive upset in the Fall Classic? There are several famous bits of received wisdom on this issue.
- Most famously, good pitching beats good hitting. A team that leads with its arms should be expected to do better at this time of the years.
- Veterans should beat kids. The thinking goes the older guys should be more used to the pressure and be less likely to wilt than the young pups.
There are numerous other theories out there, but I propose to take a look and figure out just what makes a team more likely to shock the world. Rather than just aimless pontification, I'll look at the historic precedents and see what sorts of common attributes have been held by the game's greatest underdogs, as well as what sort of teams were the most likely to choke on the highest stage. Dang it, I intend to aimfully pontificate!
Plan of action
Find the 12 biggest upsets in World Series history based on win-loss record of the winner vs. the loser. This ain't perfect as it assumes no difference in league quality, but it works. Due to the rise of divisional play and the wild cards, most of these upsets come from more recent times, and the underdog dozen represent the 1906, 1945, 1954, 1969, 1974, 1985, 1987, 1988, 1990, 1995, 2003 and 2006 contests.
I have a database that can determine what attributes the teams have. Simply put, I raided Baseball Reference and Retrosheet to add in dozens of stats on each team in baseball history. I looked at everything from basic things like batting average and saves, to more esoteric concerns like DER and average pitcher age. I even added in hitting, pitching and fielding win shares.
From this, I can figure out how a particular team ranks among its brethren in any given league. All I have to do is sort by the category. One little wrinkle: to minimize ties in a given category, whenever possible I convert into some sort of rate stat. Thus instead of looking at batter strikeouts, I look at K/AB. Instead of saves, I use saves divided by win (because you can't have a save unless you win). Heck, even for sac hits, I use SH/(H+W+HB+ROE-HR) to ensure there are no ties and account for opportunity.
A second wrinkle is comparing across eras. (Warning: the following is the most math-intensive paragraph in the article. If that number stuff makes your eyes glaze over, just skip to the next paragraph to see the results summarized.) Coming in third in an eight-team league is very different from coming in third in a 14-team league. I can't just solve this by dividing rank by number of teams because they're not all centered at the same place. An eight-team league averages out at .5625 while a 16-team league centers at .53135. Check and see for yourself: 1/8+2/8+3.8+4/8+5/8+6/8+7/8+8/8 equals 36/8, which is 4.5. That averages to .5625. To solve this, you take a team's rank in a given category, divide by the number of teams, and then divide by the league's average (which is .5625 for an eight-team league, and .53125 for one with 16 teams). An average team in any era will end up around one.
Short version: the problem of league size can be solved. An average score for a given stat: any stat is one. Lower scores mean teams ranked higher at a given category.
With this database I've tracked the following few dozen stats for the winners and losers of the biggest October upsets of all-time: batting average, on-base percentage, slugging average, isolated power, strikeout rate, homer rate, walk rate, likelihood to hit sac bunts, likelihood to hit into double plays, singles per at-bat, average hitter age, runs scored adjusted by park factor, triples per at-bat, and stolen base likelihood (SB+CS)/(H+W+HB+ROE-HR).
Impressive, eh? That's just the hitting stats. For pitchers and fielders I have: walks per nine innings pitched, Ks/9IP, HR/9IP, strikeout-walk ratio, fielding percentage, complete game percentage, saves/wins, defense efficiency ratio, double plays turned, opponent batting average, opposing OBP, average pitcher age, how much they rely on their top three pitchers (innings pitched of top pitchers divided by team IP), runs allowed adjusted by park factor, relief pitchers used, and stolen base likelihood (SB+CS)/(H+W+HB+ROE-HR).
Last but not least, at the team level I have: hitting win shares, pitching win shares, fielding win shares, record in one-run games, deviation from the team's pythagenpat win-loss record, and win-loss record against .500 or better teams divided by their overall record. About three dozen stats altogether.
Fun part: the making of a World Series upset
So, based on all that, how does the conventional wisdom hold up?
First off, underdogs are by and large better at preventing runs than at scoring them. In fact, the best score the underdogs have in any of the above stats is runs allowed adjusted by park factor. The 12 teams average out at .412. That may not sound too impressive until you think it through. Since these are the biggest dogs, they are also the least impressive teams to win the World Series. The 1987 Twins were outscored in the season, and last year's Cards of course barely played .500. One or two mediocre teams can sink the entire bunch.
Half of them (1954 Giants, 1969 Mets, 1974 A's, 1988 Dodgers, 1990 Reds, and 1995 Braves) topped their league in adjusted run prevention while the 1906 White Sox, 1945 Tigers, and 1985 Royals came in second. In comparison, only the 1974 A's led the league in runs scored adjusted for park factor, with only one other coming in second.
However, the upset teams score even better with RA+, averaging out to .253. Again, thinking it through, it makes some sense. After all, the runners-up won far more games in the regular season so they should do better. Teams like the 1906 Cubs, 1954 Indians, and 1969 Orioles are going to score high in every phase. The losers also had better pitching than hitting, but it's a virtual toss up for them. The victorious underdogs, however, had offenses barely better than average. They were extremely pitcher-orientated.
Well, makes sense. When great pitching is on, it's almost impossible to defeat it without some sort of voodoo curse. Unfortunately, great pitching isn't always on. Pitching's a double-edged sword causing great, unexpected victories and leading to unforeseen losses. Conventional wisdom has this one half right.
When you look at win shares, the scene changes a little. The favorites again are very well balanced (hey, you try winning 116 games without a balanced attack), but they do the best at hitting win shares (.270 average) followed by pitching (.282), and fielding in the rear (.411). With the underdogs it's reversed, with fielding (.413), followed by pitching (.433) and hitting well in the back (.796). Every single one underdog came in the top half of its league in fielding win shares, and four led their circuits. They unexpectedly are barely average in things like DER, and double plays, which is adjusted for opportunities to turn two. They do very well in that boring old standby, fielding percentage (.590).
What about player age? Here, conventional wisdom has problems. Upset winners have younger pitching staffs and line-ups than the squads they vanquish. Admittedly, the unexpected champs have hitters slightly older than league average (1.088), and are only a hair younger than the opposing bats (1.115), but still, they had younger offenses.
With pitching it's a very different story. Underdogs are faaaaar younger (.754) than their opponents (1.473). The striking bit here is how old the losing team's arms are. That 1.473 average is the highest score they muster in any category. Four of them had the oldest rotations in the game. The youngest of any of them were squads just barely in the youngest half; the 1906 Cubs had the fourth-youngest staff out of eight, and Whitey Herzog's 1987 Cards were sixth out of 12. Those old arms tired out in the seventh month. Pitcher age is the second-biggest discrepancy between upset winners and losers in any of these categories.
Second biggest? Well, that leads to a new question, now doesn't it? The biggest discrepancy is...drum roll, please...every stathead's favorite offensive stat, OBP.
The victorious squads are downright lousy when it comes to OBP, averaging at 1.236. That's one of the highest scores they have in any category. Meanwhile, the loser squads naturally did very well here. Every single upset featured a team with better OBP skills fall except for 2006. With only one or two other exceptions it wasn't even close. The 1985 Cards who led the league in OBP lost to a Royals squad dead last in it. The same thing nearly happened in 1969 (the Mets had better OBP than only two other teams).
Almost all the other major discrepancies fall among similar categories; underdogs were much worse at batting average, getting singles and drawing walks than their rivals. Come October, it didn't seem to matter. No wonder Billy Beane's #### doesn't work in the postseason. It fits in a little with what we found earlier. As long as their pitching stayed lights out, they had a shot, especially if the opposing arms faltered.
The underdogs close the gap some with brute force. They score almost as well with isolated average as the losers, which is notable because they do worse in almost every category.
The only categories the winners do better in, aside from younger players, are: they bunt more, hit into ever so slightly fewer double plays, their pitchers complete a bit more games, their pitchers strike out more batters (which also gives them a slightly better strikeout/walk ratio), they exceed their pythag projection by more games, hit a few more triples and use more relief pitchers per game.
I don't know what to make of most of that, other than to note one potentially surprising fact: the pythag deviation. You'd think that the upset specialists would have undershot their expected wins while their rivals overshot. It would explain the upsets if runs scored and allowed said these teams were closer than their records suggest. In fact, just the opposite happens. Neat, huh?
Does it mean anything in 2007?
[ed note: This article was written as the League Championship Series began. It hasn't been changed other than this comment because the parallel still holds true].
Well, that's all nice and fun, but does it have any application to the 2007 postseason?
Turns out it just might. You see, there's one team still out there, one squad in the mix, that does a rather nice job fitting into the characteristics described for upset specialists. Added bonus: if they get to the World Series, they won't be the favorite.
I'm talking about the 2007 Arizona Diamondbacks. Young pitching? They have the youngest pitching in the postseason. (OK, fine, they're only 0.1 years younger than Colorado. But they're also a lot younger than the Boston Red Sox). Upsetters usually have younger hitters than their rivals? The Snakes have the youngest line-up in the postseason. Admittedly, they're just plain young, while upset squads are generally average age, but no team fits the profile perfectly.
An upset squad is supposed to be carried by its pitching? Oh man, is that ever the D-backs. Low batting? Check: worst of any team in the postseason, 18 points worse than the runners-up Indians. Hell, only the White Sox were worse in the regular season. They combined that with the fewest walks of the final four for easily the worst OBP. The only good thing about their offense on the year was their power, which was OK, if not great. That all comes right out of the playbook for How to Win the World Series Without Being Favored.
Solid pitching? Check. Though Baseball Reference hasn't calculated 2007 park factors as of this writing, and thus all OPS+ and ERA+ are unofficial, the D-backs currently have the best ERA+ in the NL. One of their secrets was strikeouts, as only the Red Sox pitchers whiffed more batters of the teams still out there. Just like an underdog should be. They also have a pretty good defense in the desert.
Anything else? Oh yeah, pythag deviation. As many (including myself) have noted, the D-backs' win-loss records mocks Pythagoras and all his theorems. They led the league in wins despite being outscored by their opponents. Again, neat, huh? Well, I don't want to get into the debate on if that's skill or luck, but I will say the underdogs historically exceed their projected records by a greater degree than their opponents do. That will certainly be the case with Arizona.
This doesn't mean the D-backs will win, but of all the teams to advance to the October semis, they're the ones most built like the game's Autumn Surprises.
History instructor by day, statnerd by night, Chris Jaffe leads one of the most exciting double lives imaginable; with the exception of every other double life possible to imagine. Despite his lack of comic-book-hero-worthiness, Chris enjoys farting around with this stuff. His new book, Evaluating Baseball's Managers is available for order. Chris welcomes responses to his articles via e-mail. Oh, and now he's on twitter.
<< Return to Article