Improving the MVP System
How well does the MVP voting system actually work? Is it likely that the most deserving MVP candidate wins the award 90% of the time, or 70%, or what?
I developed a set of mathematical models to study that issue. I wrote formulas which create "known values" for each of 100 players, representing a league, and wrote formulas to adjust for how each of those 100 players is seen by each voter. In other words, each voter could "think" that each player was better than he actually was or not as good as he actually was, to a limited but realistic extent. I then studied how that model selects the MVP, starting with 3 voters casting just one ballot each, and working up to a system of 45 voters each casting a 10-person ballot, each time testing the model through hundreds of seasons.
The models suggest that, while the correct MVP is probably selected about 87% of the time with the current system, this could be increased to 90% or more by expanding the panel from 30 voters to 45. The current system is not "broken"; the current system works quite well. However, while it works well, it could work better.
The advantages of studying this with mathematical models are (1) that by working with models, one can create and evaluate thousands of seasons, rather than the limited number supplied by experience, and (2) that in the model, we know for certain who the best player actually was, and thus can determine how often the best player is selected as the MVP, whereas in the real world it is often impossible to be certain who the most-deserving player is, thus impossible to determine whether or not the right player has won the award.
In the model, each vote is determined by two things: the player’s actual value, and the "perceptual error" of the voter, resulting from the voter’s bias and from gaps in his information or gaps in his understanding. Since the perceptual error can be either positive or negative, with an increase in the number of voters it trends toward an average of zero. Thus, as more voters are added, the ratio of actual value to perceptual error increases, and the voting outcomes become more reliable.
The key question is, though, is that a meaningful advantage, given the voting structure, or a trivial advantage?
My conclusion is that it is clearly a meaningful advantage. The addition of a third voter per team would increase the reliability of the selections from about 87% to about 90%. For this reason, I would urge the BBWAA to consider adding a third voter per team.
My studies reached a second conclusion, which is that the additional weight given to the first-place vote by counting the first-place vote as worth 14 points, rather than 10 points (14-9-8-7-6-5-4-3-2-1 rather than 10-9-8-7-6-5-4-3-2-1). . . that actually probably does not help. It probably does more harm than good.
But I don’t want to start a fight about that, because (a) frankly, I could be wrong about that, and (b) it just complicates the issue. Sticking to the point, I urge you to add one voter per team.
The full study, which runs about 25 pages, can be read on this site under the name "The Perfect Voting System". Thank you for your consideration.