Wednesday, June 28, 2017

MDS Model for the BIG3 Basketball League

I haven't done much with the MDS Model recently, so this summer's BIG3 basketball league seemed like a good test case to break it out again. I applied my standard methodology on the 8 teams in the league, and I'll be using the typical techniques to project win probabilities and margin of victories as well. The nature of scoring in BIG3 does necessitate a deviation on the normal distribution I use to project spreads for other leagues, since in BIG3 teams play to a set score (60, win by 2). Based on my observations with Ultimate Frisbee (which has a similar end of game structure), I tweaked the formula to the log-normal distribution, which gives more weight to the extremes.

With that in mind, here's how the standings currently look (only based on margin of victory right now, since the teams aren't "connected" yet in the graph network):


RankTeamsPFPAPythMatrixMDS
1Trilogy60450.6500.50.650
23s Company61510.5950.50.595
3Power62590.5270.50.527
43-Headed Monsters62600.5180.50.518
5Ghost Ballers60620.4820.50.482
6Tri-State59620.4730.50.473
7Ball Hogs51610.4050.50.405
8Killer 3s45600.3500.50.350

I then used this to project the next round of games in Charlotte, NC on July 2:

Team ATeam BPickWin ProbPick By
Ball HogsTri-StateTri-State56.91%Tri-State by 2
Trilogy3-Headed MonstersTrilogy63.37%Trilogy by 4
PowerKiller 3sPower67.37%Power by 5
3s CompanyGhost Ballers3s Company61.19%3s Company by 3

1 comment:

  1. Wow, absolutely fantastic blog. I am very glad to have such useful information. Thanks. For more information visit legal herbal empire for sale

    ReplyDelete