I know exactly the problem with the MVP formulas. I nearly published a system for ranking and assigning values for fantasy sports. My system could have been easily adapted for any sport. The only reason I chose not to publish it was because I felt it would spoil the game for most people who used it. It certainly did that with me.
Originally posted by PhillyFossil
The math part I could actually fix. I created my own algorithms for ranking players in fantasy sports which was customizable for individual sports, teams, and leagues. It's the same type of math.
In fact, I'll point out the fatal flaw in the MVP calculations right now - they are simply generating a number based on totals rather than converting each stat to a point value based on a ceiling (league leader) and a floor (league minimum). Point values allow each statistical category to be ranked. The cumulative ranks for each player can then be sorted as a total or an average (makes no difference which) to generate a list of rankings. That's the simplified explanation. The actual math requires a lot of thought and adjustment as it is refined.
That's part of my post from a private thread. I'm not sure if I have the time to actually fix the math myself. It would likely take me months, but I know I could do it. I spent 3 years developing and refining my own system. Point is, it can be done, and it can be done accurately.
If I can find my copy of MS Office, I can dig up my old files, and adapt them. However, most of my newer files (last 18 months or so) are on a broken hard drive in need of data recovery. Still, I have two other hard drives with the original files. I just lack Excel to open them at the moment. Otherwise, I have to recreate the formulas from memory and check for errors.
Stats against playoff teams don't mean much for one reason: league disparity. Some leagues have teams 8 - 10 levels ahead of everyone else in the league. That doesn't mean they have the best stats though, even against CPU teams....and CPU teams do reach the playoffs. In fact, CPU players will reach the leader boards. More often than not though, the best players - usually human players - will rise to the top in a normal ranking based system. It will also prevent a one-dimensional player from dominating the rankings. The league leader categories are meant for giving such players recognition, but it's the overall performance compared to other players that is meant to determine MVP awards.
Edit:
I should add that after this thread began, I ventured to the Epic Suggestions forum. There I found the formulas being discussed for both the Pro Bowl and MVP calculations. MVP calculations were implemented, but were based off the Pro Bowl calculations - which are both listed and which were incorporated into a script written by Pabst. I installed the script, took a look at the values, noticed some problems, went back to the formulas, then back to the values, and noted problems and inconsistencies akin to my first few attempts at ranking players for fantasy sports. I fixed those problems, and to do so, I needed a floor and a ceiling. On top of that, I had to generated a value per stat based on the pool of stats I was dealing with.
To cut things short, I already have gone through the process several times while refining my own system. That's why I suddenly recognized what is wrong with MVP calculations. I finally saw the math and the results, nodded my head, and said, "Ayuh, been there. Fixed that."
