Latrinsorm
01-12-2014, 02:45 PM
I talked before (http://forum.gsplayers.com/showthread.php?87168-NBA-Did-Kobe-Deserve-His-MVP) about how two criteria model MVP award winners very well:
1. Being on a team with the best or second best record in the conference
2. Leading that team in Win Shares
Okay, but MVP voters can vote for up to 5 people, and usually the aggregate field ends up ranging from 10 to 15 people. Obviously the above metric can only capture 4 guys, does it model the top 4 MVP vote getters?
As it turns out, no. If I average the MVP shares over the past 33 years (writer vote era), the values look like this:
Rank Act Pred Ratio
1 0.878 0.755 85.97%
2 0.614 0.403 65.69%
3 0.426 0.220 51.73%
4 0.275 0.153 55.64%
Like I said, identifying the guy ranked 1st (i.e. the MVP award winner) is pretty easy, but the accuracy plummets immediately after. It is certainly possible that voters use one metric (subconsciously) for determining the most valuable player and another one entirely for the second most, third most, etc., because people are not robots. What is one thing people like disproportionate to its value on the court?
Buckets.
If we look at the leaders in points per game, only 7 of the 33 years also won the MVP (Jordan x 5, Shaq, AI), so it's not a reliable indicator of winning MVP, but from 1986 to 2013 the ppg leader has finished outside of the top 5 in MVP voting only twice:
Tracy McGrady's NBA-worst Magic squad in 2004, didn't get a single vote
AI's 6-seed 76ers in 2002 that were only 2 games ahead of the 9 seed, finished 9th
28 years, 26 top 5 finishes, I'd say that's a pretty good correlation. Also important to note is that this isn't the ppg leader of the top 4 teams as described above, which also makes Kevin Durant getting 2nd MVP on the 2010 8 seed Thunder a lot more explicable. Big picture, it suggests that guys with high ppg but low MVP-winning credentials (Carmelo, Kobe, Durant) have inflated career MVP shares, while guys with the opposite (Chauncey, Duncan) are skewed low. Currently Duncan is .055 ahead of Kobe at 10th and 11th career, which to the unenlightened observer might suggest they are about equal. Chauncey is behind Carmelo, which is obviously a travesty but now one that at least makes sense in retrospect.
1. Being on a team with the best or second best record in the conference
2. Leading that team in Win Shares
Okay, but MVP voters can vote for up to 5 people, and usually the aggregate field ends up ranging from 10 to 15 people. Obviously the above metric can only capture 4 guys, does it model the top 4 MVP vote getters?
As it turns out, no. If I average the MVP shares over the past 33 years (writer vote era), the values look like this:
Rank Act Pred Ratio
1 0.878 0.755 85.97%
2 0.614 0.403 65.69%
3 0.426 0.220 51.73%
4 0.275 0.153 55.64%
Like I said, identifying the guy ranked 1st (i.e. the MVP award winner) is pretty easy, but the accuracy plummets immediately after. It is certainly possible that voters use one metric (subconsciously) for determining the most valuable player and another one entirely for the second most, third most, etc., because people are not robots. What is one thing people like disproportionate to its value on the court?
Buckets.
If we look at the leaders in points per game, only 7 of the 33 years also won the MVP (Jordan x 5, Shaq, AI), so it's not a reliable indicator of winning MVP, but from 1986 to 2013 the ppg leader has finished outside of the top 5 in MVP voting only twice:
Tracy McGrady's NBA-worst Magic squad in 2004, didn't get a single vote
AI's 6-seed 76ers in 2002 that were only 2 games ahead of the 9 seed, finished 9th
28 years, 26 top 5 finishes, I'd say that's a pretty good correlation. Also important to note is that this isn't the ppg leader of the top 4 teams as described above, which also makes Kevin Durant getting 2nd MVP on the 2010 8 seed Thunder a lot more explicable. Big picture, it suggests that guys with high ppg but low MVP-winning credentials (Carmelo, Kobe, Durant) have inflated career MVP shares, while guys with the opposite (Chauncey, Duncan) are skewed low. Currently Duncan is .055 ahead of Kobe at 10th and 11th career, which to the unenlightened observer might suggest they are about equal. Chauncey is behind Carmelo, which is obviously a travesty but now one that at least makes sense in retrospect.