I am a new (if voracious) reader and infrequent commenter on Canis Hoopus. I am also a college senior here in the Twin Cities and a statistics minor. For my Probability class, I proposed a final project that calculates the probability of the Timberwolves making the playoffs in the 2012-2013 season. My professor has little to no experience in sports statistics, and since CH is populated by more than a few statisticians, I was hoping I could get some help with where to start/good resources to use.
My original plan was to either simulate the seasons of every team in the Western Conference and see where the Wolves end up or to just try simulating a Wolves season a few dozen/hundred times and see how many times they have a higher winning percentage than the 8-seed last year. I'm hoping that as we get further into the semester, I will learn methods to run these simulations. But what I am really looking for right now is a good way to determine the probability of the Wolves winning any given game. The simplest method that my professor suggested was to just use the winning percentages from last year. But I really want to see how much the team's off-season moves affect their performance. Is there a preferred statistic that could be used to compute this? Just from reading comments, I have seen that most people favor Win Shares or PA100 over PER, but can these be used to calculate a game-by-game probability for the whole team? Would I be better off summing individual player metrics, or team things like points scored/allowed per 100 possessions? Is what I am trying to do even possible with the knowledge/expertise I have? Any help would be greatly appreciated.
The Indefatigable Charlie Holmes