APBRmetrics

I'm still finalizing all the details, so it remains to be seen how much time I end up allocating to what.

-Chris

Looking my latest power rankings, the top 6 are still the same as they've been all season, but now the Clippers have vaulted to the top behind a 15-game win streak. rk | team | str | ofs | dfs | sos ---+------------------------+-------+-------+-------+------- 1 | Los Angeles Clippers | 1.094 | 1.042...

I'm temporarily closed my GitHub, but I'm sure I'll be reopening it at least partially in the near future. I'm consulting for an NBA team, now, so I'll need to check with them out of courtesy. I don't believe it'll be an issue.

-Chris

You can include subjective input in statistical models, too. It's all just information.

-Chris

How do you deal with the DRose situation in this type of model? Models like these assume constant team strength subject to random fluctuations. If this is no longer true (Rose recovers from his ACL injury), you have to revise the model. It'd be easy enough to add an indicator for Rose playing/not p...

If there's interest I could write code to compute playoff probabilities. No need to run simulations, as it's easy enough to compute exact probabilities under the model assumptions.

-Chris

My current power ratings, sorted by defensive strength. rk | team | str | ofs | dfs | sos ---+------------------------+-------+-------+-------+------- 1 | Memphis Grizzlies | 1.064 | 0.992 | 0.933 | 1.003 2 | Indiana Pacers | 1.002 | 0.948 | 0.946 | 0.995 3 | Chicago Bulls | 1.012 | 0.968 | 0.956 | ...

Power ratings typically consist of offensive and defensive ratings. These, in turn, could be estimated from such parameters as rebounds, steals, turnovers, 2-point shots, 2-point efficiency, etc.

-Chris

Yes, one interesting approach would be to fit a team level power ranking model, then fit another model to predict offensive and defensive strength as a function of other statistical parameters.

-Chris

Helpful video and links. I'd suggest watching it at 720p and full screen due to the smallish text.

http://randyzwitch.com/sqldf-package-r/

-Chris

Are you smoothing using a local regression?

-Chris

What's the source of the shot distance? The ESPN data?

-Chris

Here's D2 Findlay's home/away record. They were undefeated in 2009 with a perfect 36-0 record (and D2 championship). year | home | away | n -----+-------+-------+---- 2002 | 0.857 | 0.667 | 30 2003 | 0.867 | 0.583 | 31 2004 | 0.933 | 0.583 | 32 2005 | 1.000 | 0.750 | 34 2006 | 1.000 | 0.750 | 30 200...

Accounting for elevation, here are the top 5 home courts for each division. Team is the ID from ncaa.com. school_name | team | home --------------------+-------+------- Wisconsin | 796 | 1.080 St. Louis | 609 | 1.065 Oklahoma | 522 | 1.064 Denver | 183 | 1.064 Ohio | 519 | 1.059 school_name | team |...

There are some sample size issues regarding extreme changes in altitude, but I'm comfortable in saying that a team going from a low to an extreme altitude gets punished tremendously, whereas going from an extreme to a much lower altitude yields either no impact or a small positive.

-Chris