A Better Way to Choose a World Series MVP

In the grand scheme of things, the World Series MVP is little more than a historical footnote (though the award presentation has been rather amusing in recent years). This year Salvador Perez seemed to be an appropriate winner, with a World Series traditional stat line of .364/0/2 (batting average, home runs, RBIs), and he also picked up the game-tying RBI in Game 5.

At the same time, aside from his traditional outstanding defense that should not be discounted, he didn’t have THAT great of a World Series.  He had a high batting average, but just two extra base hits, two RBIs (and the game-tying RBI in game 5 was all because of Hosmer’s baserunning, not because Perez produced a particularly positive result with the bat).

Generally speaking, there weren’t a whole lot of attractive options for MVP.  Mike Moustakas was the only other starting with a batting average above .300. Edinson Volquez was the only pitcher with two starts, and while he pitched well he didn’t pick up any victories. The bullpen was lights out, but traditionally only a closer will win MVP, and Wade Davis was unable to rack up enough saves to make him a viable candidate.

When looking at another statistic, Win Probability Added (WPA), there are some more attractive candidates. WPA calculates the likelihood that a player’s team will win the game as a result of each outcome. For example, in Game 5 Curtis Granderson led off the bottom half of the ninth inning with a home run. According to Baseball Reference, before Granderson’s home run the Mets had a 55% chance of winning the game.  After the home run they had a 65% chance of winning, so Granderson was credited with 0.1 WPA for that play alone.

When looking at WPA for the Royals for the whole series for their offensive players, Perez definitely isn’t anywhere close to the top of the list:

WPA is not a perfect statistic (is there one?). Its primary shortcoming is that the batter is given all “credit” for a batting outcome, even if he did nothing to deserve that credit.  For example, if a batter reaches on an error he is in all likelihood given a positive WPA. Hosmer tops the list above in part because of Daniel Murphy’s eighth inning error in Game 4.  On the other side, if a runner on second is thrown out at home on a single, the batter is the one who is given negative WPA. And pitchers are given a negative WPA if their defense commits an error (for example, in Game 4 Jeurys Familia was given -.31 WPA as a result of Murphy’s error).

On the pitching side for the Royals, here were the WPA leaders for the World Series:

Luke Hochevar threw just five innings total in the series, but they were high leverage innings.  He pitched the top of the ninth inning in Game 1 when the Royals were down a run (and would go on to win).  He pitched the fifth inning in Game 3 when the Royals were down by two and already lost their starter Yordano Ventura (this was the lowest leverage appearance Hochevar would make).  He pitched the sixth inning in Game 4 when the Royals were down by a run.  And most importantly he pitched allowed no runs (just one baserunner) in the tenth and eleventh innings in Game 5 (and picked up his second win of the series when the Royals scored 5 times in the twelfth inning).

Of course, it’s one thing to produce a positive WPA in the World Series when your team is already up 3-0 in the series versus coming through in the clutch in a Game 7.  Multiple websites have determined how important particular games are in a 7 game series. Fangraphs recently used this to determine that Murphy’s Game 4 error was the sixth costliest error in World Series history in that it made it 13% less likely that his team would win the World Series (Bill Buckner’s error was the costliest in World Series history as it made it 20% less likely the Red Sox would win the 1986 World Series).

If we use this chart over at Basketball Reference (written by Neil Paine, now at fivethirtyeight.com) we find that each of the games in the World Series had the following Leverage Index:

• Game 1: 0.16
• Game 2: 0.166
• Game 3: 0.125
• Game 4: 0.176
• Game 5: 0.08

In other words, Game 4 of the Series was the most important, as there’s a huge difference in the likelihood of a team winning the series whether they’re tied 2-2 or down 3-1 (which is why Murphy’s error was so costly). As entertaining as Game 5 was, it was the least important game of the series given the Royals had three chances to close out the series.

So if we were to combine the Leverage Index with a player’s WPA for every single game, we’d calculate the win probability a player added to his team winning the series (not just the game). In Game 5 Luke Hochevar produced 0.25 WPA in a game with 0.08 Leverage Index.  Multiplying them together would give us 0.02, which in plain English could be described by saying “Luke Hochevar’s performance in Game 4 made it 2% more likely that his team would win the World Series.”

For all Kansas City players this is what we would come up with:

And after all that it was Ben Zobrist whose performance made it most likely that the Royals would win the World Series. And Salvador Perez is a lot closer to the bottom of the list than the top.  Zobrist was at his best in Games 1, 2, and 4, which were the three most important games in the series:

Perez’s defense is not accounted for in any of these calculations, and certainly he should get some credit as well for catching all 14 innings of Game 1.  And he’s something of the face of the Royal’s franchise, whereas Zobrist is likely little more than a three month rental for Kansas City.  Essentially Perez was a perfectly fine choice for World Series MVP. But many of his teammates, especially Ben Zobrist, had performances that improved their team’s chance of winning the World Series much more.