There’s something wrong with a voting system where a majority of voters can vote for a candidate, and he can still lose.
By Devin McCarthy and Matt Dewilde
Cabrera vs. Trout / ESPN
Major League Baseball decides its MVP using a ranked-choice system, of sorts. It is a point-based system known as a “Borda count” – a method already discussed on this blog in the context of FIFA Ballon D’Or voting. It’s a pretty good voting system for baseball, as it allows voters to fully express their MVP preferences without any fear of vote splitting or wasting a vote on a “spoiler” player.
In the MVP Borda count, each voter receives ten votes. A first place vote gives a player 14 points, a second place vote gives 9 points, a third place vote 8, and so on, down to one point for a tenth place vote.
The effectiveness of the Borda count is predicated on voters being honest on their ballots. The system is very susceptible to strategic voting, as a voter with a strong preference that one player win the MVP over another serious contender could rank his or her favorite player first and not rank the second at all, so that he receives zero points.
Fortunately, the members of the Baseball Writers’ Association of America have very little reason to be dishonest in their voting. MVP voting is not an election of a player to an office, but rather an evaluation of his performance in the past year. The results of the election have no effect on the voter. Baseball players also do not have political parties, so voters do not have an incentive to boost the vote totals of one party at the expense of another.
Most importantly, MVP ballots are public, not secret, and there are only 30 total voters. If a voter casts an absurd ballot that leaves off an obviously great player, there’s a good chance that ballot will be noticed and thoroughly mocked on the internet.
Polarized MVP Voting
Nonetheless, there have been indications of some factionalism emerging among MVP voters. Two relatively distinct “parties” have developed: voters who prefer to use traditional measures of player value, like batting average, RBIs, and the success of the player’s team, and voters whose decisions are based on more recently developed “sabermetric” measures of value, such as WAR (wins above replacement).
The MVP award has become a focal point of the debate over how big a role sabermetrics – that is, advanced statistical methods of interpreting baseball – should play in evaluating the quality of players. During the 2012 season, Miguel Cabrera made baseball history by becoming the first player to win the Triple Crown (leading the league in the traditional measures of batting average, home runs, and runs batted in) since 1967. In the very recent past, a Triple Crown winner would have been a unanimous choice for MVP. But sabermetrics have become increasingly popular among some Hall voters, and these metrics – which better account for things like speed and defense – strongly suggested that Mike Trout was more valuable than Miguel Cabrera in 2012.
Cabrera still won the MVP in 2012 with a large majority, garnering 22 first place votes to Trout’s six. There was no evidence of either “sabermetrician” or “traditionalist” voters attempting to sabotage the other side’s candidate by ranking him low or leaving him off the ballot entirely – all voters ranked both Trout and Cabrera either first or second except one, who ranked Trout third behind Adrian Beltre.
MVP voters were given a nearly identical question this past year, after Cabrera and Trout once again dominated the league. Cabrera didn’t win the Triple Crown in 2013, but was actually a better offensive performer than in 2012. Trout had another stellar year, and while his defense slipped slightly from his extraordinary rookie year, most sabermetric measures again showed Trout to be the most valuable player.
Unsurprisingly, Cabrera won again, with 23 first place votes compared to just five for Trout. This time, the voting was less straightforward; while every voter ranked Cabrera first or second, there were ballots that ranked Trout fourth, fifth, and seventh.
Borda Count Breakdown
The man who ranked Trout seventh, Bill Ballou, explained himself by arguing that no player on a team that doesn’t contend for the playoffs has any value, and Trout’s Angels were a last-place team in 2013. His argument was somewhat undermined by the fact that he ranked David Ortiz, a member of the World Series-winning Red Sox, below Trout; if he had fully followed through on that principle, he would not have ranked Trout at all. The other voters who ranked Trout so low likely had similar rationales.
Even if these voters were not strategically ranking Trout lower to increase the chances that Cabrera would win, it is problematic that a player’s point total could be significantly hurt because a small group of voters believe that the definition of the award disqualifies the player from contention. In a Borda count system, ranking one of the two main contenders very low will hurt him much more than simply not voting for him in a plurality system.
In fact, a Borda count system gives a rogue voter almost three times as much power to influence the outcome as a “normal” voter who ranks the two main candidates first and second. That’s because the normal voter could only create a 5 point swing between Cabrera and Trout (giving one player 14 points and the other 9). But a “rogue” voter – one who is either strategically voting to help one of the two candidates win or believes that one of the two does not qualify as “valuable” at all – could create a 14 point swing (giving one player 14 points and the other 0).
Thus, the system gives disproportionate power to voters who are either dishonest or crazy.
Obviously, Miguel Cabrera would have won the 2012 and 2013 MVP awards regardless of what voting system was used. But in a closer race, the use of Borda count could absolutely have made the difference.
Here's a hypothetical example in which the Borda count system would be a problem, sticking with the Cabrera/Trout debate. Let’s say that of 30 voters, 16 of them ranked Trout first, and all 16 of those voters ranked Cabrera second. That’s a total of 224 points from first place votes for Trout. Meanwhile, 14 voters ranked Cabrera first, giving him 196 points from first place votes. Two of those 14 writers, however, decided that Trout was ineligible for the award because the Angels were not in playoff contention, and did not rank him at all. The other 12 ranked Trout second. Trout would then get 108 points from his second place rankings, while Cabrera would get 144 from his.
Add that all up, and Miguel Cabrera has defeated Mike Trout 340 to 326, even though a majority of voters thought Trout was the Most Valuable Player.
It’s an unlikely scenario, but not an implausible one. In fact, it very nearly occurred in 2006, when Justin Morneau defeated Derek Jeter by a slim 14 point margin for the AL MVP, in part because Joe Cowley ranked Jeter sixth. Cowley’s logic: Jeter could not be the most valuable because the Yankees had several other good players that year.
The Change: Ranked Choice Voting
There’s a way to address this problem without changing the MVP ballot: ranked choice voting. Instead of a point-based preference system, ranked choice voting determines the player with majority support by transferring votes from defeated players to their voters’ second choices. The player with the fewest first choice rankings is eliminated first, and transfers continue until one player has a majority of votes.
In the hypothetical example above, Mike Trout would have won outright without any transfers due to his majority of first choice votes. In a race with more than two serious candidates – say, the 2011 AL MVP race, where Justin Verlander received only a plurality of first choice votes – some voters’ ballots would be transferred to their second choices, in that case starting with the lone voter who ranked Michael Young first.
Ranked choice voting maintains all the advantages of the Borda count – allowing voters to express their full preferences and eliminating spoilers and vote-splitting issues – without giving voters the temptation of dishonest strategic voting or the ability to distort the election result with outlying interpretations of the definition of “valuable.”
Of course, baseball’s MVP is far from the institution most in need of ranked choice voting – that’d be the House of Representatives. The Borda count method has never been a real problem for baseball, only a potential one. Still, the example of the Cabrera/Trout MVP battle is instructive for why a Borda count (and other systems that give voters incentives to dishonestly hurt a candidate, such as Range Voting) might be a bad idea for political elections. In an election with secret ballots in which voters actually have a vested interest in the outcome, there would be a lot more strategic voting. It’s hard to imagine too many Obama voters giving Romney nine points when they could give him zero, and vice versa.
So, should baseball change its MVP voting system? It’s well worth thinking about.
In the MVP Borda count, each voter receives ten votes. A first place vote gives a player 14 points, a second place vote gives 9 points, a third place vote 8, and so on, down to one point for a tenth place vote.
The effectiveness of the Borda count is predicated on voters being honest on their ballots. The system is very susceptible to strategic voting, as a voter with a strong preference that one player win the MVP over another serious contender could rank his or her favorite player first and not rank the second at all, so that he receives zero points.
Fortunately, the members of the Baseball Writers’ Association of America have very little reason to be dishonest in their voting. MVP voting is not an election of a player to an office, but rather an evaluation of his performance in the past year. The results of the election have no effect on the voter. Baseball players also do not have political parties, so voters do not have an incentive to boost the vote totals of one party at the expense of another.
Most importantly, MVP ballots are public, not secret, and there are only 30 total voters. If a voter casts an absurd ballot that leaves off an obviously great player, there’s a good chance that ballot will be noticed and thoroughly mocked on the internet.
Polarized MVP Voting
Nonetheless, there have been indications of some factionalism emerging among MVP voters. Two relatively distinct “parties” have developed: voters who prefer to use traditional measures of player value, like batting average, RBIs, and the success of the player’s team, and voters whose decisions are based on more recently developed “sabermetric” measures of value, such as WAR (wins above replacement).
The MVP award has become a focal point of the debate over how big a role sabermetrics – that is, advanced statistical methods of interpreting baseball – should play in evaluating the quality of players. During the 2012 season, Miguel Cabrera made baseball history by becoming the first player to win the Triple Crown (leading the league in the traditional measures of batting average, home runs, and runs batted in) since 1967. In the very recent past, a Triple Crown winner would have been a unanimous choice for MVP. But sabermetrics have become increasingly popular among some Hall voters, and these metrics – which better account for things like speed and defense – strongly suggested that Mike Trout was more valuable than Miguel Cabrera in 2012.
Cabrera still won the MVP in 2012 with a large majority, garnering 22 first place votes to Trout’s six. There was no evidence of either “sabermetrician” or “traditionalist” voters attempting to sabotage the other side’s candidate by ranking him low or leaving him off the ballot entirely – all voters ranked both Trout and Cabrera either first or second except one, who ranked Trout third behind Adrian Beltre.
MVP voters were given a nearly identical question this past year, after Cabrera and Trout once again dominated the league. Cabrera didn’t win the Triple Crown in 2013, but was actually a better offensive performer than in 2012. Trout had another stellar year, and while his defense slipped slightly from his extraordinary rookie year, most sabermetric measures again showed Trout to be the most valuable player.
Unsurprisingly, Cabrera won again, with 23 first place votes compared to just five for Trout. This time, the voting was less straightforward; while every voter ranked Cabrera first or second, there were ballots that ranked Trout fourth, fifth, and seventh.
Borda Count Breakdown
The man who ranked Trout seventh, Bill Ballou, explained himself by arguing that no player on a team that doesn’t contend for the playoffs has any value, and Trout’s Angels were a last-place team in 2013. His argument was somewhat undermined by the fact that he ranked David Ortiz, a member of the World Series-winning Red Sox, below Trout; if he had fully followed through on that principle, he would not have ranked Trout at all. The other voters who ranked Trout so low likely had similar rationales.
Even if these voters were not strategically ranking Trout lower to increase the chances that Cabrera would win, it is problematic that a player’s point total could be significantly hurt because a small group of voters believe that the definition of the award disqualifies the player from contention. In a Borda count system, ranking one of the two main contenders very low will hurt him much more than simply not voting for him in a plurality system.
In fact, a Borda count system gives a rogue voter almost three times as much power to influence the outcome as a “normal” voter who ranks the two main candidates first and second. That’s because the normal voter could only create a 5 point swing between Cabrera and Trout (giving one player 14 points and the other 9). But a “rogue” voter – one who is either strategically voting to help one of the two candidates win or believes that one of the two does not qualify as “valuable” at all – could create a 14 point swing (giving one player 14 points and the other 0).
Thus, the system gives disproportionate power to voters who are either dishonest or crazy.
Obviously, Miguel Cabrera would have won the 2012 and 2013 MVP awards regardless of what voting system was used. But in a closer race, the use of Borda count could absolutely have made the difference.
Here's a hypothetical example in which the Borda count system would be a problem, sticking with the Cabrera/Trout debate. Let’s say that of 30 voters, 16 of them ranked Trout first, and all 16 of those voters ranked Cabrera second. That’s a total of 224 points from first place votes for Trout. Meanwhile, 14 voters ranked Cabrera first, giving him 196 points from first place votes. Two of those 14 writers, however, decided that Trout was ineligible for the award because the Angels were not in playoff contention, and did not rank him at all. The other 12 ranked Trout second. Trout would then get 108 points from his second place rankings, while Cabrera would get 144 from his.
Add that all up, and Miguel Cabrera has defeated Mike Trout 340 to 326, even though a majority of voters thought Trout was the Most Valuable Player.
It’s an unlikely scenario, but not an implausible one. In fact, it very nearly occurred in 2006, when Justin Morneau defeated Derek Jeter by a slim 14 point margin for the AL MVP, in part because Joe Cowley ranked Jeter sixth. Cowley’s logic: Jeter could not be the most valuable because the Yankees had several other good players that year.
The Change: Ranked Choice Voting
There’s a way to address this problem without changing the MVP ballot: ranked choice voting. Instead of a point-based preference system, ranked choice voting determines the player with majority support by transferring votes from defeated players to their voters’ second choices. The player with the fewest first choice rankings is eliminated first, and transfers continue until one player has a majority of votes.
In the hypothetical example above, Mike Trout would have won outright without any transfers due to his majority of first choice votes. In a race with more than two serious candidates – say, the 2011 AL MVP race, where Justin Verlander received only a plurality of first choice votes – some voters’ ballots would be transferred to their second choices, in that case starting with the lone voter who ranked Michael Young first.
Ranked choice voting maintains all the advantages of the Borda count – allowing voters to express their full preferences and eliminating spoilers and vote-splitting issues – without giving voters the temptation of dishonest strategic voting or the ability to distort the election result with outlying interpretations of the definition of “valuable.”
Of course, baseball’s MVP is far from the institution most in need of ranked choice voting – that’d be the House of Representatives. The Borda count method has never been a real problem for baseball, only a potential one. Still, the example of the Cabrera/Trout MVP battle is instructive for why a Borda count (and other systems that give voters incentives to dishonestly hurt a candidate, such as Range Voting) might be a bad idea for political elections. In an election with secret ballots in which voters actually have a vested interest in the outcome, there would be a lot more strategic voting. It’s hard to imagine too many Obama voters giving Romney nine points when they could give him zero, and vice versa.
So, should baseball change its MVP voting system? It’s well worth thinking about.
RSS Feed