In the nearly fifty years the Finals MVP has been awarded, it has been given to a player on the losing team only once. Oddly enough, as many fans know, this happened during the first year of the award; Jerry West won after a scorching series where he scored 38 a game and barely lost in Game 7 to the Bill Russell-led Celtics.
Since then, only winners have gotten the MVP, even if the best player was on the other side. In fact, the Finals MVP is pretty easy to predict most years — the leading scorer on the winning team has gotten it 35 times out of 45. (The exceptions are Billups in 2004, Bird in 1986 where Kevin McHale barely edged him out, Magic in 1982 and 1980, Dennis Johnson in 1979, Unseld in 1978, Walton in 1977, Willis Reed in 1973, Chamberlain in 1972, and, of course, West in 1969.)
How does one improve on this?
Quantitatively, it’s a bit difficult, as we only we one datum on a losing MVP and the awards typically go to who’s perceived as the leader, like Unseld, or who’s perceived as the hero, like Reed, walking back onto the court after an injury in 1970. To adjust for guys like Magic or Walton — who lead their respective teams without leading them in scoring — I used the game score metric popularized by John Hollinger, with a couple tweaks. First, I replaced raw rebounds with rebound percentage to account for the earlier years where rebounds were more prevalent, due to lower field-goal percentages. The second change was a boost to playmakers using the term USG%*AST%^0.5 (the square root was found to be useful after some testing as it seems the voters care more about shooting than passing.)
There are some strong trends in the results, but there are some serious kinks. About 75% of the MVP winners have scored 22 points or more a game, but Unseld won with 9. Some players win with 13 assists a game and some with 2. Some have won because they were the leading rebounders and interior defenders, while Joe Dumars won with a paltry 1.8 rebounds a game.
The Predictive Model
The model was completed using binomial regression in R, and the form was a logit function. Basically, I need something to calculate the odds of a player winning where the results are binary (1 or 0.) I used my modified game score and divided it by the maximum game score found during the series. This accounts for the fact that your odds depend on your competition, and it’s also an indirect adjustment for pace or some era effects. And, of course, I added a dummy variable for whether or not the player’s team won the series, which has a huge impact on the odds. There’s also an adjustment for missed games.
As good as this was given the limited information, it still missed 8 players, but at least the errors were mostly smaller ones like Dennis Johnson with a 39.4% chance at winning (which he ultimately did) compared to leading scorer Gus Williams at 45.0%.
In fact, I like these results more than the real ones.
Larry Bird lost a finals MVP to Cedric Maxwell one year. Maxwell averaged 17.7 points, 9.5 rebounds, and 2.8 assists a game. He outscored Bird, sure, but only by a hair, and Bird had 15.3 rebounds a game — he nearly outrebounded Moses Malone.
The finals MVPs per the model:
Season Team Odds Player
2013 MIA 89.0 LeBron James
2012 MIA 86.0 LeBron James
2011 DAL 59.6 Dirk Nowitzki
2010 LAL 55.5 Kobe Bryant
2009 LAL 88.0 Kobe Bryant
2008 BOS 37.3 Paul Pierce
2007 SAS 50.7 Tony Parker
2006 MIA 99.5 Dwyane Wade
2005 SAS 59.6 Tim Duncan
2004 DET 76.8 Chauncey Billups
2003 SAS 98.6 Tim Duncan
2002 LAL 88.5 Shaquille O’Neal
2001 LAL 91.5 Shaquille O’Neal
2000 LAL 99.6 Shaquille O’Neal
1999 SAS 87.6 Tim Duncan
1998 CHI 90.6 Michael Jordan
1997 CHI 94.3 Michael Jordan
1996 CHI 77.4 Michael Jordan
1995 HOU 63.3 Hakeem Olajuwon
1994 HOU 98.1 Hakeem Olajuwon
1993 CHI 96.2 Michael Jordan
1992 CHI 86.9 Michael Jordan
1991 CHI 93.8 Michael Jordan
1990 DET 83.7 Isiah Thomas
1989 DET 77.3 Joe Dumars
1988 LAL 86.8 Magic Johnson
1987 LAL 94.7 Magic Johnson
1986 BOS 67.5 Larry Bird
1985 LAL 41.7 Kareem Abdul-Jabbar
1984 BOS 91.6 Larry Bird
1983 PHI 66.8 Moses Malone
1982 LAL 44.3 Magic Johnson
1981 BOS 56.2 Larry Bird
1980 LAL 48.2 Magic Johnson
1979 SEA 44.9 Gus Williams
1978 WSB 53.3 Elvin Hayes
1977 POR 67.2 Bill Walton
1976 BOS 49.5 Dave Cowens
1975 GSW 97.1 Rick Barry
1974 BOS 51.7 John Havlicek
1973 NYK 28.1 Walt Frazier
1972 LAL 43.1 Wilt Chamberlain
1971 MIL 47.4 Kareem Abdul-Jabbar
1970 NYK 70.4 Walt Frazier
1969 BOS 63.4 John Havlicek
The model, by the way, is shown below. And note that the standard game score works decently well, but if you want to include the tweaks those are in the notes.
Likelihood = 1/[1+exp(14.452 – 11.37*Game score – 4.418*Series winner(1 or 0) )]
Odds = Likelihood / sum(Likelihood of each player in series)
The 2014 Finals MVP?
Even the model was unsure of Jerry West’s award in 1969 because of the history against a player on the losing side. For whatever reason, we’ve changed the definition of Finals MVP from the MVP of the Finals series to the MVP of the title-winning team. This is probably based on the faulty reasoning of post hoc ergo propter hoc: after a team loses, we assign the blame to the leading player on the losing side and the credit to whoever sticks out on the winning side. Willis Reed, I would argue, was not the best player in the 1970 Finals. He missed a game, but the media saw him return from injury and assigned the credit to him, ignoring Walt Frazier’s command of the offense and his ballhawking defense.
If the Heat lose, LeBron James will not win the Finals MVP unless he does something truly special. And even then, people would argue against him because, apparently, “he could have done more.” You could say that about every player, even the winners.
As I touched on earlier, your odds of a Finals MVP are based on your competition, and the opening for LeBron is contingent on no other player separating himself from the field. Based on the box score stats and the model — and this does include the advanced stats rebound percentage, usage, and assist rate — LeBron has decent odds at winning the Finals MVP, only behind Kawhi Leonard:
Odds Player
34.4 Kawhi Leonard
20.2 LeBron James
17.5 Tim Duncan
17.3 Tony Parker
4.2 Manu Ginobili
2.5 Boris Diaw
2.2 Danny Green
If the series had already ended, LeBron’s odds at winning would be the same as Jerry West’s when he won, who had a 32.5% at winning compared to Havlicek’s 63%. LeBron, in fact, has the closest chance at having the award than any loser has ever had. Of course, going by box score stats, you miss out on a lot of defense, which is the forté of both Leonard and Duncan. And LeBron has been inconsistent on defense; those are the caveats. But we can’t let the media decide this one for us, just because it’s been a certain way in the past. We also hadn’t seen the leader in points per game in the regular season win a title for 20 years, until Michael Jordan.
If the Heat win the series, and LeBron plays as well as he has so far, he’d win in a landslide: 96.6% odds.
(For a funny side note, the model sees Jeff Ayres as a more likely MVP than Wade assuming San Antonio wins. There’s a harsh penalty for losing, but that’s an awful sign.)
Kawhi Leonard could win a Finals MVP because his team is balanced and no one sticks out, and he has a leading story right now guarding LeBron James. Many are saying he’s been better than LeBron because of his defense on James, never mind that LeBron has a 70 TS% and a 33 usage%, which has never happened in the Finals before. Leonard’s a good player, yes, but he’s not much of a playmaker and has a better supporting cast around him.
If you were asked to choose one player for the Finals for your team — based on how everyone’s played so far — how would you not choose LeBron? He’s been the most valuable player, despite his teammates.
LeBron could score 50 points in a gGame 5 loss and still lose on a Finals MVP. An award like that is worth something when it’s earned.
This is 2014. We have more tools than ever. We can make a better case than “can’t win because his team didn’t win.” I’m not even arguing LeBron deserves it, whatever that means.
Just at least give him a chance.
*Notes: in the game score formula, replace rebounds with 0.59*ORB%*MPG+0.25*DRB%*MPG. This approximates the standard weights in the original formula very well. For the years where offensive and defensive rebounds are not tracked separately, use 0.4*TRB%*MPG. Then add the “playmaker” boost: 0.45*AST%^0.5*USG%*MPG. All the advanced stats, like DRB% and USG%, are not on a percent scale (1 to 100) but go from 0 to 1. For example, LeBron’s usage is 0.334 right now. The NBA didn’t always track steals, blocks, and turnovers, so the formula is less accurate in earlier eras. However, for a quantitative approach, this is the best I can do.
