Dedicated (or even casual!) readers of our NFL, NBA, tennis or soccer coverage probably know that FiveThirtyEight loves it some Elo ratings. The Elo system, originally designed by a Hungarian-American physicist to rank competitive chess players, can be easily adapted to assess just about any form of competition, telling us who’s good and bad at any given moment. And with the MLB playoffs set to commence Tuesday night, we thought it was the right time to (re-)introduce Elo rankings for baseball teams.
How does Elo work for MLB? The basic formula is the same as it is for pro football: Both teams come into a given game with initial ratings (on a scale where league average is 1500) that generate an expected win probability for the matchup, and they leave the game with some of the loser’s rating points re-allocated to the winner. If the game plays out as expected, the transfer of Elo points is small; if it’s a shocking upset, more points are transferred between loser and winner.1 The ultimate goal is for the Elo ratings to adjust themselves so that when the two teams play again, we’ll have an accurate sense of which is stronger and by how much.
Of course, baseball is a different game than football. For instance, something Elo-ologists call the “K-factor” (which determines how quickly the ratings react to new information) is much smaller in baseball than in football,2 because luck influences the outcome of any single game in MLB far more than it does in the NFL. The ratings also revert further toward the mean between seasons3 in baseball than they do in football, perhaps reflecting the lack of certainty we should have about any baseball team’s talent at the beginning of the year. Finally, mimicking FiveThirtyEight editor-in-chief Nate Silver’s original formulation of Elo for baseball, I assigned playoff games an extra 50 percent boost in terms of the degree to which they change each team’s rating. (This doesn’t have any real effect on the predictive quality of the ratings either way, but it’s a nod to the increased importance and focus that the postseason takes on.)
This season, Elo — like other predictive systems — loves the Toronto Blue Jays, assigning them roughly a 19 percent probability of winning the World Series. It’s also especially fond of the two teams facing off in Wednesday night’s National League wild-card game, the Pittsburgh Pirates and Chicago Cubs. Although one of the two will be lost to the whims of a single-elimination play-in game, Elo thinks the Cubs and Pirates are two of the three best teams in baseball right now. (By contrast, Elo thinks Tuesday night’s AL wild-card game between the New York Yankees and Houston Astros is a snoozer featuring the two worst teams in the playoff field, both of whom are worse, according to Elo, than the 81-80 Cleveland Indians.)
This wasn’t a season of all-time dominance by any team. The Blue Jays rank 53rd in regular-season Elo since baseball’s expansion era began4 — in no small part because the team’s Elo was exactly average as recently as June 2. It’s a common refrain that MLB’s playoffs are a crapshoot, and that notion is certainly not dispelled by Elo: Compare the Blue Jays’ 19 percent probability of winning it all to the 48 percent probability we gave the eventual NBA champion Golden State Warriors before their playoffs began in the spring.
But maybe that’s a good thing. This year’s playoff field may not have any historic standouts, but it does feature an abundance of good, solid teams, any one of which would be an interesting and worthy champion. In an age in which the best teams in baseball aren’t always the ones left standing come World Series time, that might be all we can ask for out of an MLB postseason.
Check out our live coverage of the NL Wild Card game.