APBRmetrics

Performed a "test" running two ridge regressions of the season to date Regression one: Each row is one game Y = (home margin - away margin) X are the whole teams rather than their offensive and defensive halves. 1 for home, -1 away, and HCA always set to 1. Regression two: Each row is one ...

What are the alternatives to not including it as a variable? Calculate HCA at the team level with SRS home/away difference compared to league average but don't forget to include the effect of B2B. This way you'll have all the data you need for all arenas instead of being limited to PBP data 2000+ (...

Also, why not just use win probability as a proxy for game state? That's how I would start doing this. http://i.imgur.com/dhMeAzK.gif I love that. Another question: Last night, the Spurs played the second game of a back to back at Memphis. The night before, they were home against Milwaukee, making ...

I'm more a fan of having a single column for HCA, and toggle it for home (1) and away(0) possessions If you were to do two columns for HCA, how would you do it? The 1/-1 way (It looks good to me?)? For matters of prediction/interaction with other units, knowing a team plays the same on offense but ...

Why are you treating the "game state" (point margin) as a dummy variable and not just a continuous variable? Do you have reason to believe the effect (if any) will be non-linear? I mean, maybe it will be, but are you really going to have a column for each possible value of point margin? S...

I have never used HCA as a variable on the regression stage to calculate ARPM/RAPM since I haven't cared about it. I doubt J.E. does it this way but if I implemented HCA as a variable to the regression, I would do it differently than what you describe. What are the alternatives to not including it ...

This is extremely confusing to me. What's m, jt, pm, ts, sc, dg, ab? m <- margin. m2 if team is up 2, m1 if up 1, etc Remaining initials are those of the players for this example -- Jeff Teague, Paul Millsap, Thabo Sefolosha. Just for the purposes of this mock-up did I do 2 vs 2, actual data set is...

http://i.imgur.com/ZDQcoKJ.png Sample design matrix for 2v2 game Golden State Warriors AT Atlanta Hawks Using 1/-1 dummy variables for offense/defense not as a matter of preference but so the regression "sees" who the home team is: when Atlanta is on offense, their players are asssigned 1...

What were your methodologies for picking games? Sum up player ratings times % of minutes for both teams, include HCA and profit?

thanks. i have heard about Sport Vu. I need the raw video files to train our computer vision models. How good of quality do you need? nba.com/stats regular season box scores have videos of every event, (probably) encompassing a continuous 48 minutes in overlapping segments. Beyond that... ballstrea...

Having problems remembering where this was posted, but you don't need video: player tracking from SportVU on NBA.com possession chalkboards can be parsed into X,Y something like 20 times a second

Priors. The problem is priors. Apparently there's no way to use a prior if you're using glmnet? Well, that's not entirely true -- apparently you can use a prior by including more rows -- from the previous season or seasons, or of a pre-baked SPM -- and weigh them differently? How does that look? Say...

Let's take a step back for a moment. When you talk about 'adding HCA' are you talking about estimating how individual players' performance changes with home/away games, or trying to produce RAPM numbers that are 'controlled for' home court effects? The latter. Home court effects, score effects, per...

I had then used a number of different columns representing different score differentials (hometeamup4, hometeamup9, etc) This should work. One column for each possible score differential, although you might want put all situations with e.g. score_diff>30 into one bucket Makes sense. With a big enou...

I would be inclined to handle home court by having the same kind of split as with offense and defense - P1oh, P1dh, P1or, P1dr. To add a linear time left or scoring differential factor per player, you could have P1oh-diff, P1dh-diff, P1or-diff, P1d-diff *in addition to* the other independent variab...