Those of you who were loyal readers of John Hollinger's annual "Pro Basketball Forecast" may recall that he often included a rant about the Most Improved Player (MIP) award voting. John's complaint was that the voters too often cast their ballot for players who got big increases in playing time -- which usually led to proportional increases in their per-game statistics -- rather than players who actually showed significant improvement from one season to the next.
In order to create a reasonable starting list of candidates for the MIP award, I came up with my own method to assess improvement. The first step was to determine which players would be eligible, and I decided to include all players who (A) had played at least 1,250 career minutes coming into this season, and (B) are on pace to play at least 1,250 minutes this season.
The next step was to determine how to measure a player's improvement from one season to the next. I decided to look at the difference between a player's win shares per 48 minutes (an estimate of the number of wins a player contributed to his team through his offense and his defense) in the current season and his baseline expectation (weighted average of win shares from previous three seasons) coming into the season.
In order to give some context to this number, I computed the probability that a randomly selected player would beat his baseline expectation by a difference this large or larger. Here's an example using James Harden's 2011-12 season:
• Harden's baseline expectation entering last season was .147.
• Harden's average last season was .230.
• Harden's improvement was .230 minus .147, or .083.
• The probability that a randomly selected player will beat his baseline expectation by at least .083 is 0.012. In other words, only about 12 players out of 1,000 will beat their baseline expectation by .083 or more.
Let me make it perfectly clear that I am not suggesting that the NBA use a formula to determine its MIP winner. However, I do think this is a good way to whittle down the list of potential candidates.
Here are the top five MIP award candidates so far this season:
Baseline: .037 | Current: .177 | Probability: 0.0001