Comments on: BBR Rankings: Schedule-Adjusted Offensive and Defensive Ratings (January 21, 2011)

By: joe

joe — Mon, 24 Jan 2011 09:05:36 +0000

wow, the cavs are really bad. does anyone know what their ranking was last year? i would guess that this is the biggest one year turnaround in nba history.

By: DSMok1

DSMok1 — Sun, 23 Jan 2011 14:34:43 +0000

Yep, Yariv, L2 is tied to arithmetic mean and L1 to median. What I was investigating was whether, because team performance is not normally distributed, it is beneficial to adjust for the skew somewhat by using a value on the median side of the arithmetic mean or on the opposite side of the arithmetic mean from the median. Further research is hinting that, while exponents above 2 tend to be better than 2, it's not a big difference.

By: yariv

yariv — Sat, 22 Jan 2011 19:22:57 +0000

I personally would stick with L2 norms, they are widely used and I believe there is a theoretical statistical reason for this (although I'm not sure, probably because they are tied to averages. L1 norms are tied to medians). If there is a strong (statistically) evidence that some other value is significantly better predictor, I might consider it, but I would require it to be quite strong.

By: DSMok1

DSMok1 — Sat, 22 Jan 2011 14:40:14 +0000

Neil, remember our earlier discussion about |residual|^2 vs. other exponents? I am running some out-of-sample analyses on last year's data and am getting some interesting results. It's appearing so far that while my choice of ^1.5 was quite poor, ^2 is not the best either--something in the range of ^2.35 was best in my first test. I'll conduct more and eventually post on my blog and attempt to interpret.

By: Greyberger

Greyberger — Fri, 21 Jan 2011 21:43:50 +0000

Good gravy the Cavs are bad. There goes the Kings' pick.

By: DSMok1

DSMok1 — Fri, 21 Jan 2011 15:36:21 +0000

That's points per 100 possessions. Most teams run around 92 possessions per game.

If the Heat (7.88) were playing the Lakers (6.34) on a neutral court, the Heat would be expected to win by about 1.54 points per 100 possessions played, or probably around 1.4 points per game.

The home court advantage is added on to the 100 possession number either direction--so Miami would be favored by (1.54+3.68=5.22pts) per 100 poss. at home.

By: Garron

Garron — Fri, 21 Jan 2011 14:50:44 +0000

Was just wondering, if team X has 7 points, and team Y has 6 points, whats the points per possession advantage that team X has over team Y in a neutral field? Just trying to put the numbers into context