BBR News: MVP Award Tracker
Posted by Justin Kubatko on November 6, 2009
Today I would like to unveil a new feature on the site: the NBA MVP Award Tracker. Before I say much more, let me get two disclaimers out of the way:
- The NBA MVP Award Tracker ranks candidates based on a model built using previous voting results. This list does not represent the opinion of this site. Rather, these are the players that the voters are likely to target.
- I am not trying to come up with a formula to pick the MVP Award winner. In my opinion it would be silly to use a single formula as the basis for determining who should win an award.
Got it? Good. Now moving on…
I looked at 54 years of voting results to determine what factors the voters deemed most important when selecting the MVP. In the end, four factors proved to be much more important than anything else: team wins, points per game, rebounds per game, and assists per game. I'm going to omit the fine details, but using those four statistics I came up with a model to rank MVP candidates in a given year. For example, here are the top five candidates from 2008-09, with the player's actual finish in parentheses:
+----+-------------------+-----+------+------+------+------+------+ | Rk | Player | Tm | W | L | PTS | TRB | AST | +----+-------------------+-----+------+------+------+------+------+ | 1 | LeBron James (1) | CLE | 66 | 16 | 28.4 | 7.6 | 7.2 | | 2 | Kobe Bryant (2) | LAL | 65 | 17 | 26.8 | 5.2 | 4.9 | | 3 | Pau Gasol | LAL | 65 | 17 | 18.9 | 9.6 | 3.5 | | 4 | Dwight Howard (4) | ORL | 59 | 23 | 20.6 | 13.8 | 1.4 | | 5 | Chris Paul (5) | NOH | 49 | 33 | 22.8 | 5.5 | 11.0 | +----+-------------------+-----+------+------+------+------+------+
The model correctly predicted LeBron James as the winner (duh), and identified four of the top five candidates, missing on Pau Gasol (Dwyane Wade actually finished third). For some historical perspective, here are the results for every NBA MVP Award:
+------+-----+--------------------------------+--------------------------------+ | Year | Tp5 | Actual Winner PrRk | Predicted Winner AcRk | +------+-----+--------------------------------+--------------------------------+ | 1956 | 3/5 | Bob Pettit ( 3) | Neil Johnston (na) | | 1957 | 3/5 | Bob Cousy ( 1) | Bob Cousy ( 1) | | 1958 | 3/5 | Bill Russell ( 1) | Bill Russell ( 1) | | 1959 | 3/5 | Bob Pettit ( 1) | Bob Pettit ( 1) | | 1960 | 4/5 | Wilt Chamberlain ( 1) | Wilt Chamberlain ( 1) | | 1961 | 4/5 | Bill Russell ( 3) | Wilt Chamberlain ( 4) | | 1962 | 4/5 | Bill Russell ( 2) | Wilt Chamberlain ( 2) | | 1963 | 4/5 | Bill Russell ( 2) | Elgin Baylor ( 2) | | 1964 | 4/5 | Oscar Robertson ( 2) | Wilt Chamberlain ( 2) | | 1965 | 3/5 | Bill Russell ( 1) | Bill Russell ( 1) | | 1966 | 5/5 | Wilt Chamberlain ( 1) | Wilt Chamberlain ( 1) | | 1967 | 2/5 | Wilt Chamberlain ( 1) | Wilt Chamberlain ( 1) | | 1968 | 3/5 | Wilt Chamberlain ( 1) | Wilt Chamberlain ( 1) | | 1969 | 3/5 | Wes Unseld ( 5) | Wilt Chamberlain (na) | | 1970 | 4/5 | Willis Reed ( 2) | Kareem Abdul-Jabbar ( 3) | | 1971 | 3/5 | Kareem Abdul-Jabbar ( 1) | Kareem Abdul-Jabbar ( 1) | | 1972 | 4/5 | Kareem Abdul-Jabbar ( 1) | Kareem Abdul-Jabbar ( 1) | | 1973 | 4/5 | Dave Cowens ( 2) | Kareem Abdul-Jabbar ( 2) | | 1974 | 4/5 | Kareem Abdul-Jabbar ( 1) | Kareem Abdul-Jabbar ( 1) | | 1975 | 4/5 | Bob McAdoo ( 2) | Dave Cowens ( 2) | | 1976 | 4/5 | Kareem Abdul-Jabbar ( 4) | Dave Cowens ( 3) | | 1977 | 3/5 | Kareem Abdul-Jabbar ( 1) | Kareem Abdul-Jabbar ( 1) | | 1978 | 3/5 | Bill Walton ( 1) | Bill Walton ( 1) | | 1979 | 4/5 | Moses Malone ( 1) | Moses Malone ( 1) | | 1980 | 3/5 | Kareem Abdul-Jabbar ( 1) | Kareem Abdul-Jabbar ( 1) | | 1981 | 3/5 | Julius Erving ( 2) | Larry Bird ( 2) | | 1982 | 4/5 | Moses Malone ( 4) | Larry Bird ( 2) | | 1983 | 4/5 | Moses Malone ( 1) | Moses Malone ( 1) | | 1984 | 3/5 | Larry Bird ( 1) | Larry Bird ( 1) | | 1985 | 5/5 | Larry Bird ( 1) | Larry Bird ( 1) | | 1986 | 3/5 | Larry Bird ( 1) | Larry Bird ( 1) | | 1987 | 4/5 | Magic Johnson ( 1) | Magic Johnson ( 1) | | 1988 | 4/5 | Michael Jordan ( 3) | Magic Johnson ( 3) | | 1989 | 3/5 | Magic Johnson ( 1) | Magic Johnson ( 1) | | 1990 | 3/5 | Magic Johnson ( 1) | Magic Johnson ( 1) | | 1991 | 4/5 | Michael Jordan ( 1) | Michael Jordan ( 1) | | 1992 | 3/5 | Michael Jordan ( 1) | Michael Jordan ( 1) | | 1993 | 4/5 | Charles Barkley ( 1) | Charles Barkley ( 1) | | 1994 | 4/5 | Hakeem Olajuwon ( 1) | Hakeem Olajuwon ( 1) | | 1995 | 3/5 | David Robinson ( 1) | David Robinson ( 1) | | 1996 | 3/5 | Michael Jordan ( 1) | Michael Jordan ( 1) | | 1997 | 3/5 | Karl Malone ( 2) | Michael Jordan ( 2) | | 1998 | 5/5 | Michael Jordan ( 3) | Karl Malone ( 2) | | 1999 | 3/5 | Karl Malone ( 1) | Karl Malone ( 1) | | 2000 | 4/5 | Shaquille O'Neal ( 1) | Shaquille O'Neal ( 1) | | 2001 | 4/5 | Allen Iverson ( 5) | Shaquille O'Neal ( 3) | | 2002 | 3/5 | Tim Duncan ( 1) | Tim Duncan ( 1) | | 2003 | 3/5 | Tim Duncan ( 1) | Tim Duncan ( 1) | | 2004 | 4/5 | Kevin Garnett ( 1) | Kevin Garnett ( 1) | | 2005 | 3/5 | Steve Nash ( 4) | Amare Stoudemire ( 9) | | 2006 | 3/5 | Steve Nash ( 7) | Dirk Nowitzki ( 3) | | 2007 | 4/5 | Dirk Nowitzki ( 1) | Dirk Nowitzki ( 1) | | 2008 | 4/5 | Kobe Bryant ( 1) | Kobe Bryant ( 1) | | 2009 | 4/5 | LeBron James ( 1) | LeBron James ( 1) | +------+-----+--------------------------------+--------------------------------+
Let me take one of the lines above and explain what it means…let's look at 2001. In 2001, the model predicted four of the top five vote-getters (not necessarily in the correct order). Allen Iverson was the actual winner, but he was projected to finish fifth by the model. Shaquille O'Neal was the predicted winner, but he finished third in the actual voting.
Overall, the model does well: 36 of the 54 predictions (67%) were correct. Looking at the 18 incorrect predictions, 9 of the predicted winners finished 2nd in the actual voting, 5 finished 3rd, 1 finished 4th, 1 finished 9th, and 2 did not receive any MVP votes. The model also pegged 192 (over 70%) of the top five finishers (again, not necessarily in the correct order).
November 6th, 2009 at 5:39 pm
Of course, by creating this model, you are now changing future history. Henceforth, voters will subtly, subconsciously change their voting habits to avoid seeming so predictable and/or controlled by a formula.
November 6th, 2009 at 5:46 pm
The Johnny Twisto Uncertainty Principle!
This is a pretty cool feature. Looks like Wins is the highest-weighted factor.
November 6th, 2009 at 6:14 pm
Good formula. Should be interesting to see how things turn out this year.
Sorry for the generic response, but it's just the facts.
November 6th, 2009 at 6:58 pm
Interesting observation: Steve Nash won the award twice despite the model not showing him as even a top 3 candidate. He won it once with ranking of 7th!!!!
I will forever argue that Nash did not deserve either of those awards (I would have given it to him in 2007 though as he deserved it that season). I thought Shaq should have won in 2005 and Kobe in 2006.
Neil, could you list the modeled results for the top 5 from 2005, 2006, and 2007?
I think it is obsurd that Nash will finish with two MVP's while never getting to the finals, let alone winning a ring, while Kobe and Shaq, who were more dominant at their peaks and over a longer time span, won at least 4 titles each, and made six finals appearances, will each only be a one-time winner.
November 6th, 2009 at 11:51 pm
hey that s a cool tool to follow the MVP race
one suggestion might be that you limit it no more than two people per team to more
otherwise i think the top ten gets a little crowded with people from the same high win squad
having carmello on there makes sense
chauncey, not as much, but maybe
kenyon? no way
same with boston
i think this would more accurately reflect the mvp race cause 99 times out of 100 it is going to go to a team's top dog (scorer) maybe not, if they are balanced, other guy is very strong in other areas, but i think it would be really very rare for it to go to the third guy down the line, even on teams with a "big three'
anyway nice work!
November 8th, 2009 at 6:26 am
I have been recently reading Bill Simmons' The Book of Basketball and he often states a player's Points, Rebounds, Assists and Wins when comparing players' candicy for MVP status. Although we all know there is so much more than just these four statistics, it does seem to have a great impact on discussions. Being a fan of the advance statistics found on this website, I am somewhat appalled by this formula's popularity.
November 9th, 2009 at 12:00 pm
This seems like a formula for who actually should have won MVP.
November 9th, 2009 at 1:43 pm
I was curious about the two "na" entires under the predicted winners (Neil Johnston in 1956, and Wilt Chamberlain in 1969):
--1956 was the first year that the MVP award existed. Johnston did not receive any MVP votes at all, but the voting system used in 1956 and 1957 appears to have been very different from that used in almost every year since then, which it makes it hard to draw a good comparison to later years. In 1956 and 1957, voters simply chose one player, rather than voting for first/second/third place etc. In 1956, only eight players received votes, and Johnston was not one of them. But in modern terms, Johnston essentially didn't get any first-place votes. It's interesting that no one voted for him, but it's impossible to say where he would have finished under a modern-style voting system. I suspect he wouldn't have gone completely voteless.
--Between 1965-66 and 1971-72, the MVP voting records on this web site only list the top five, even though the voting system in use during that era almost certainly resulted in more than five players getting votes. (I'm not sure if the top five is all that the NBA publicly released back in that era, or if somewhere along the line someone compiled the top five for some reference source and that's all that is readily available today.) As a result, all that we can really say about Wilt in '69 is that he wasn't in the Top 5. He probably did receive some votes.
Based on the factors that voters seem to place the most weight on, 1969 has both one of the most unlikely winners and one of the most overlooked predicted winners. It must have been a strange MVP race; I'm guessing that there were just no obvious candidates.
Another year that jumps out at me is 1975. You often hear that Rick Barry should have won that year, but was denied due to his unpopularity. According to this methodology, it looks like McAdoo was a more likely choice than Barry, and Dave Cowens even more likely than either of them. Golden State's relatively unimpressive regular-season record (48 wins) is probably pulling Barry down. The same principle also likely explains why Kareem Abdul-Jabbar in 1976 and Moses Malone in 1982 are among the most unlikely MVPs. Both put up monster seasons playing for teams that were on the playoff bubble down to the wire (the '82 Rockets just barely squeaked in; the '76 Lakers missed the postseason, though that was due as much to a quirk in the seeding rules in use in '70s as to anything else, as they had a better record than two of the five Western Conference teams that qualified).
IINM, the voting was done by players up until around 1980, then changed to members of the media. From the table above, it doesn't look like this change had much of an effect on the predicted results. Rick Barry and Steve Nash aside, the pre-1980 and post-1980 portions of the chart don't really look much different.
November 9th, 2009 at 6:07 pm
Looking at the standings today (Nov. 9), it seems that the amount of games the player played doesn't factor into the tracker (or at least not enough). Bynum's hot start has him in 4th place, but he's missed 2 out of the 7 Laker games. It seems fairly improbable that voters would put Bynum so high if the data so far projects out to a 82 game season.