LMAO @ Hollinger

[Seventyniner]

I think most people who are not familiar with his system cannot wrap their mind around the concept, thus they continue to insist that Hollinger is biased. Point differential is a legit factor, but it's not infallible. Over the long haul it might be the most infallible next to win/loss however. Point differential indicates both the strength of the offense as well as the defense in one neatly packaged number. It's drawback is that some teams intentionally run scores up while others do not. That doesn't change anything as far as the game goes, but it does affect Hollinger's system.

I'd love to see a point differential statistic that somehow factors out garbage time, and perhaps includes injuries.

Also, should a 5-point OT win for a team count as +5, 0 (because it was tied at the end of regulation), or something in between?

[jestersmash]

He weighs his rankings IMO on some spurious assumptions: that the last 10 games are more important than a larger body of work; that one bad performance can severely change a ranking; and that the ability to close out games is not truly that important.

Moreover, Pop is more likely to pull his starters either when winning a game big or losing by a large margin more than some coaches and it would be hard to build that into the picture. There is also importantly no adjustment for the fact that the West wins more than 60% vs. Eastern teams so that the SOS of all Western teams is probably understimated. Also, there is no adjustment for the fact some teams play considerably more back to backs - Chicago and Magic are playing a ton while the Lakers play the least.

That being said SOS and point differential are historically pretty good predictors of success but a better model could be designed and Hollinger should publish every year how well his model worked or did not work.

See, I can respect these sorts of criticisms about his ranking system, and I wish more people would do this. You're exactly correct.

Conversely, if the Spurs are blowing a team out by 30 points through 3 quarters and insert their scrubs early in the 4th while the opposing team keeps their starters in and brings the differential down to 18 or 15 by the end of the game, obviously the end of game differential becomes somewhat muted to what it should be in terms of how the game had been played when both teams were trying their best to win. I can imagine a system that somehow adjusts for minutes played by the top 5 players on each team in order to rectify this issue.

As for some of Hollinger's assumptions being spurious - not so much, actually. His decision to place greater emphasis on the last 10 games isn't so much spurious as it is arbitrary. It's really his prerogative to place greater emphasis on the last X games because he wanted his ranking system - by design - to reflect recent performance. His decision to use 10 instead of 12 or 15 is arbitrary. His decision to use point differential as the foundation of his ranking system is evidenced-based, however.

At the end of the day many people seem to fail to realize the strengths and limitations of a mathematical model. There's almost this sense of "well if it isn't absolutely perfect and infallible then it's worthless." That's totally unrealistic. There are perhaps a handful of models in science that are - for all intents and purposes - virtually infallible. Current models for quantum electrodynamics predict many quan ies that agree with experiment to the 10th decimal place. That's pretty damned impressive but it's by no means a realistic standard for the level of accuracy that mathematical models should offer.

Too many people are just so egregiously misinformed and uneducated about simple concepts like mathematical models or the fact that point differential has factually been a better predictor of championship success than NBA.com standing, I can't help but pity them. Mostly - I just don't find their arguments to be compelling. Refuting Hollinger's rankings by calling them "horse statistics" - while creatively insulting - just isn't compelling as an argument to me. Sorry "703 Spurzzzzzz"

You have to wonder - how many "HURR DURR TROLLINGER HORSE STATISTICS LOL" Hollinger critics realize that his rankings have better predicted playoff success than using season-long point differential alone (and certainly have better predicted playoff success than using NBA.com standings)? Not many I'd presume. Most cite his rankings off without actually spending sufficient time to assess how useful they've been historically.