What were the best (and worst) World Series, and World Series games, ever?

Two days ago, the Giants won the World Series. Congratulations to them (I guess; I was rooting, fairly stridently, for the Royals to beat them). While I wish the result had been different, there's no denying that the Series was a ton of fun to watch -- it was a back-and-forth affair between two very good teams. The Royals, in the playoffs for the first time in 29 years, rode blazing speed, terrific defense, and a brilliant (and almost totally anonymous) bullpen to the brink of a world championship, while the Giants, uh, have Madison Bumgarner. (And Hunter Pence, who annoys the shit out of me with how many hits he gets from two-strike counts.)

As often happens in the days after the Fall Classic (especially when it's won by a team I dislike, which has happened every single year for the last nine goddamn years), I find myself refusing to accept the prospect of five months without baseball. I also love data, not just for its own sake but for its ability to provide an answer (though not necessarily the answer) to seemingly intractable questions. And so, while thinking about the joy, anxiety, and ultimately disappointment of the World Series just ended, and reading a fascinating article by Fangraphs' Jeff Sullivan, I started to wonder: just how good was the 2014 World Series compared to its predecessors? Which was the best World Series? The worst? If I could pick one World Series game from the past to attend in person, which one would it be? (I was pretty certain it would not be the one I actually did attend -- Game Three of the 2007 Series, which blew horse cock, and not just because my Rockies lost or because it was cold or because Red Sox fans are the absolute fucking worst, though none of those things helped). Four hours and some judicious use of Microsoft Excel later, I had my answers.

I am, by undergraduate education, an engineer and economist, so my first step in attempting to solve any problem is always the same: define the parameters. In this particular case, what makes a World Series "good"? That seems pretty easy -- a good World Series is one that is exciting, hard-fought, close throughout (tense, to put it simply). All else being equal, a World Series that goes the maximum seven games* is generally preferable to a four-game sweep, not just because it means more baseball but because it means the two teams are probably pretty evenly matched. Okay, but how do we measure excitement/tension in baseball? That's trickier, but if you're even the slightest bit versed in "advanced" baseball metrics, you know the answer to that as well: Leverage Index (LI). (TL;DR version of LI for those of you who don't care about the nitty-gritty: LI essentially measures how important a particular moment in a baseball game is, based on how the probability that one team or the other will win the game would change given the outcome of the situation. The average LI across all situations is 1.0; a plate appearance with two outs and no runners on base in the ninth inning of an 11-0 game has an LI near zero, because no matter what happens, the win probabilities won't change much; a plate appearance with two outs and runners on second and third in the bottom of the ninth with the home team trailing by a run has an LI much, much greater than 1.0, because an out wins the game for the away team while a hit (probably) wins it for the home team.)

The missing piece was how to account not just for how exciting each game was, but how the outcomes of the games combined to make the Series, as a whole, exciting -- close games are good, but if the same team keeps winning, it doesn't make for a classic World Series. (Take the 2005 World Series as an example -- each of the Series' four games was very competitive, but by the time Game Four rolled around, it felt pretty anti-climactic, because the White Sox had won each of the first three games and were almost certain to win the Series regardless of what happened in Game Four.) The order of the games also matters; a seven-game Series that features six close games and one blowout will be much more exciting if the blowout occurs in Game One rather than Game Seven. But ordinary LI is blind to these considerations, because it only measures each team's probability of winning an individual game, not a larger series of games. We need a way to measure not just the excitement level of the in-game situation, but the excitement level of each game within the larger context of the Series.

As it turns out, a few years ago Sky Andrecheck (now employed by the Cleveland Indians) came up with a metric, Championship Leverage Index ("ChampLI"), that measures how much a particular game impacts a team's chances of winning a championship, in almost exactly the same way that the conventional LI measures how much a particular situation impacts a team's chances of winning an individual game. (Thanks to Dave Studeman and his excellent 2012 article for pointing me in ChampLI's direction.) As we can see, Game Seven of the World Series has a ChampLI of 166.7, or, in other words, is 166.7 times as critical as an average regular season game! So, if we can combine LI with ChampLI, we should be able to see how competitive each World Series, and each game within each Series, was, and thus get a good idea of the quality of the Series as a whole.

The rest was time-consuming but pretty straightforward, thanks to the Play Index at Baseball Reference (which might be my favorite site on the Internet), which allows you to see basically any data you want for tens of thousands of Major League games, including every World Series game. Figure out the average LI of all 641 World Series games played to this point, and then multiply that by the ChampLI of that game (52.1 for the first or second game of a Series, 166.7 for a Game Seven, &c.), and we get how good each World Series game was. Take the average of all of the games in a given Series, and hey presto, we can figure out which was the best World Series ever.

Enough chatter. Here are the nine (because there are nine innings) best World Series ever, listed together with what I call the game-adjusted Series Championship Leverage Index, or GAS-ChampLI:

1. 1924 (110.6)
2. 1991 (105.9)
3. 1975 (103.0)
4. 1952 (100.0)
5. 1926 (94.0)
6. 1925 (93.4)
7. 2011 (93.0)
8. 1997 (92.6)
9. 1972 (92.5)

And, for completeness, the nine worst Series:

1. 1989 (24.2)
2. 1928 (32.2)
3. 1963 (32.8)
4. 1937 (32.8)
5. 1919 (35.0)
6. 1910 (35.3)
7. 2007 (35.4)**
8. 1999 (35.9)
9. 2004 (36.8)

Think of GAS-ChampLI this way: an average plate appearance in the 1924 World Series was 110 times as crucial as an average regular-season plate appearance, and almost five times as crucial as an average plate appearance in the 1989 Series! The GAS-ChampLI of the 2014 Series was 75.6, a solid number that makes it the 24th-best Series played to date.

Taking one step back, we can also use this to figure out which have been the best and worst individual games in World Series history. The nine best:

1. Game 7, 1924 (314.8)
2. Game 7, 1991 (298.8)
3. Game 8, 1912 (286.6)
4. Game 7, 1997 (284.2)
5. Game 7, 2001 (259.8)
6. Game 7, 1975 (248.1)
7. Game 7, 1946 (242.6)
8. Game 7, 1972 (241.1)
9. Game 7, 1926 (234.2)

And the nine worst:

1. Game 4, 1989 (10.6)
2. Game 4, 1937 (12.0)
3. Game 5, 1961 (14.0)
4. Game 4, 1950 (14.5)
5. Game 3, 1960 (14.5)
6. Game 4, 1954 (15.3)
7. Game 1, 1959 (15.5)
8. Game 5, 1977 (16.0)
9. Game 4, 1938 (16.4)

(The full dataset is available here for those interested.)

Not all World Series games are created equal -- the players of Game 7 of the 1924 Series felt roughly 26 times the pressure that the players of Game 4 of the 1989 Series did. Fun (at least for me) fact: the best Series, the best game, the worst Series, and the worst game were all lost by the New York/San Francisco Giants.

This is as much an interesting (I hope?) thought experiment as anything else, but it gives us a good sense of how the various high dramas of October compare to one another. Baseball lends itself to endless narratives; it's nice to see how each one fits into the long, grand history of the game. I'll update this post with more numbers, and some stories about the best and worst games and Series, later.

* On four occasions -- in 1903, 1912, 1919, and 1921 -- the World Series actually featured a Game Eight. In 1912, prior to the advent of lighted stadiums, Game Two was declared a tie after 11 innings due to darkness, and on the other three occasions, under the rules of the day, the World Series was best-of-nine. As we shall see, 1912's Game Eight was one of the best World Series games ever.

** God damn it.