|
dsparksWeb Site: http://arbitrarian.wordpress.com/ BallHype Member since June 20, 2007 |
Comments Posted
Choosing the MVP, geometrically
-
dsparks +3Thanks Jason, for the clarification. I think it would be possible, then, to use Game Scores to produce a similar graph. Maybe someday, if I get the time... I think the results would likely be pretty similar, although it looks as though slightly more calculating went into creating Game Scores than my metric...
Choosing the MVP, geometrically
-
dsparks The thing about Game Scores and PER and the like are that they are time-rated, which Winshares are not. So, per unit of time, I am certain that Garnett is more valuable, but in terms of actual wins produced, Pierce played more minutes, and so had more opportunity to contribute to team success. I think that when measured in this specific way, it's reasonable to view Pierce as more valuable.
Giant Network Diagram of NBA Players
-
dsparks I'm glad you find it fascinating... I do, too. I could just stare and stare at it. I can't get over the colors, and how everything makes so much sense. The diagrams attached to this particular post are for players' entire careers between 1979 and 2007. However, I can fairly easily make a similar diagram for any set of players or even teams, and I could do this for different sports, as well, and plan to eventually. I thought I'd debut it with Jordan, Bird, et al. front and center though. It's pretty compelling. The way these diagrams were created, I removed team statistics, including wins and losses. If you read the comments on this post: http://arbitrarian.wordpress.com/2008/02/19/chris-mullin-is-the-next-michael-jordan/
I sort of explain why -- basically, I'm just looking at playing style, and team success skews the results in odd ways. Anyway, I'm glad you're interested.
Giant Network Diagram of NBA Players
-
dsparks Just so everybody knows, if you're looking at the PDFs, just hit Ctrl-f and that will search the document for you. Much faster that way, if you want to find someone specific.
NBA Game Theory, Part 1
-
dsparks I just ran a real quick regression in response to Patrick above regarding the importance of team vs. opponent assist ratio, and found that when you use team's ppg, opponent ppg, team's ast ratio, and opponent ast ratio to predict team winning percentage, you get an R^2 of about .94 on the strength of the ppg numbers alone--neither of the assist ratio statistics are significant.
Even though it's not theoretically sound, I reran the regression using team ast ratio, opp. ast ratio, team fg%, and opp fg%, and I got an R^2 of about 0.66, and team assist ratio is not statistically significant, which gives some credence to Mr. Schuhmann's story. Then, when I try to predict winning percentage using only team and opponent assist ratios, the R^2 is about .45, and both indepentent variables are significant, with opponent ratio being about 25% more detrimental than team ratio is beneficial, in terms of winning percentages, which again is evidence in favor of the theory. I would like to disclaim any sort of statistical validity, and I hope no real statisticians are reading any of this. Nevertheless, it's interesting stuff.
