Espn rankings

[fairweatherfan]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

I disagree that they're random - that last one in particular seems to show a relationship between winning close games and success overall.  .45 is a pretty solid correlation, though it's probably not as strong a predictor as pt differential.  Let's not throw out the baby with the bathwater.

[nba is the worst]

I didn't see that anyone has posted the link to Stein' rankings:

http://espn.go.com/nba/powerrankings/_/season/2011/week/3

[Fafnir]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

I disagree that they're random - that last one in particular seems to show a relationship between winning close games and success overall.  .45 is a pretty solid correlation, though it's probably not as strong a predictor as pt differential.  Let's not throw out the baby with the bathwater.

Eh the correlation is pretty weak, good teams win more close games than bad teams, but random noise plays a bigger role than most people believe.

[Greenbean]

I did't see that anyone has posted the link to Stein' rankings:

http://espn.go.com/nba/powerrankings/_/season/2011/week/3

I have to say it is kind of cool to compare the two lists back to back one purely opinon based (I mean it is ONE GUYS opinion!) vs. purely number crunching.

I think the truth at this point in the season for the C's lies somewhere in between Hollinger and Stein's rankings.

[nba is the worst]

I did't see that anyone has posted the link to Stein' rankings:

http://espn.go.com/nba/powerrankings/_/season/2011/week/3

I have to say it is kind of cool to compare the two lists back to back one purely opinon based (I mean it is ONE GUYS opinion!) vs. purely number crunching.

I think the truth at this point in the season for the C's lies somewhere in between Hollinger and Stein's rankings.

I was actually surprised that Stein put the Celtics in first above the undefeated Hornets - but it certainly seems that they deserve no lower than 3rd, given that the Spurs have just the loss to the Hornets so far.

I guess Stein enjoyed the C's dominating the Heat as much as I did!

[LooseCannon]

I know stats cant account for those scenarios perfectly but just because the Celtics were not blowing out bad teams every night didnt mean that the Magic were a better team for example.

You could interpret the stats to say that the Celtics were penalized last season for blowing big leads that other top teams would have held.

People seem to have an attitude that stats are worthless because they didn't predict the Celtics playing in the NBA Finals.  Which is some weird re-writing of history.  I think the consensus mood here in the playoffs went from "I think the Celtics will beat Miami, but I wouldn't be shocked if they lose and, based on how they played in the regular season, I doubt they can beat both Orlando AND Cleveland" to guarded optimism that it would be close against Cleveland to a belief that this team ought to be a favorite against Orlando, based on the past two series.

[fairweatherfan]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

I disagree that they're random - that last one in particular seems to show a relationship between winning close games and success overall.  .45 is a pretty solid correlation, though it's probably not as strong a predictor as pt differential.  Let's not throw out the baby with the bathwater.

Eh the correlation is pretty weak, good teams win more close games than bad teams, but random noise plays a bigger role than most people believe.

Yeah, it's been my experience that people underestimate how much of a role random noise plays in just about everything.  Human beings love us some dichotomies, that's for sure.

[Fafnir]

BUMP!

I just read an article at one of the most useful basketball sites out there, Basektball-Reference. Neil Paine addresses one of the very topics we've debated in this thread:

http://www.basketball-reference.com/blog/?p=8159

Basically as a predictor of post-season success (in the CFs/Finals) the number of 10+ point wins is more important than your record in close games. (whether or not you're playing a good or bad team)

[MBunge]

Basically as a predictor of post-season success (in the CFs/Finals) the number of 10+ point wins is more important than your record in close games. (whether or not you're playing a good or bad team)

Uh...is anyone arguing any different?  I think everyone would probably agree that a team which wins X number of games by 10+ points is likely better than a team which wins X number of games by 5 or less points.

The argument is, to take Boston and Miami as examples, whether or not point differential justifies ranking a 6-4 team over an 8-2 team, particularly when the 8-2 team has already beaten the 6-4 team twice.  To me, when you rank the 6-4 team over the 8-2 team in that situation, it demonstrates there's something profoundly wrong with your evaluation.

Mike

[Fafnir]

The argument is, to take Boston and Miami as examples, whether or not point differential justifies ranking a 6-4 team over an 8-2 team, particularly when the 8-2 team has already beaten the 6-4 team twice.  To me, when you rank the 6-4 team over the 8-2 team in that situation, it demonstrates there's something profoundly wrong with your evaluation.

In a subjective ranking made by people, I think you're right. There is no reason that two Boston wins head to head over Miami shouldn't have them higher. (When combined with Boston's otherwise stellar record)

But adding head to head conditions to an automated power ranking wouldn't make it more accurate. Fundamentally head to head match ups are a small sample size and the raw scores don't tell the story. Not to mention you're ranking a ton of teams so head to head factors would create a nightmare for a regression. It wouldn't increase the predictive power of the model. (Cavs swept the season series with the Spurs in 2006-2007 for example)

The system doesn't consider match up factors, injuries, and a whole host of other subjective factors. (for example back to backs or the last game in a 4 in 5 days sequence)

[MBunge]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

A more powerful predictor of what?  No one is arguing that winning games by big margins would tend to indicate a team in pretty good.  Again to use Boston and Miami as an example, the argument is whether point differential justifies ranking a 6-4 team over an 8-2 team, even if the 8-2 team has already beaten the 6-4 team twice.

Mike

[Fafnir]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

A more powerful predictor of what?  No one is arguing that winning games by big margins would tend to indicate a team in pretty good.
Mike

You were actually arguing just that earlier in the thread. You argued that points differential is a weaker predictor than raw record. (along with general dislike of Hollinger's ratings)

[MBunge]

But adding head to head conditions to an automated power ranking wouldn't make it more accurate.

That's true, but it highlights the point that any sort of power ranking based on "objective" stats is going to be guilty of stupid mistakes like any subjective observer.

Mike

[Fafnir]

But adding head to head conditions to an automated power ranking wouldn't make it more accurate.

That's true, but it highlights the point that any sort of power ranking based on "objective" stats is going to be guilty of stupid mistakes like any subjective observer.

Mike

Only for a small sample size. Overall once the teams have played a large sample of games and SoS has evened out such head to head factors fade to being relatively insignificant. Like a lot of people have said, this system really doesn't work well at all until you have a decent sample.

Now that doesn't mean that you should use Hollinger's system to predict a playoff series. Playoff series are much more match up based than that. It's just an indicator of how well the teams have played overall in the regular season. Personally I look at match ups more than point differential.

Understanding what a model is built to do and what it doesn't do is very important.

[MBunge]

http://www.nba.com/thunder/news/bulls070201.html

http://www.basketballprospectus.com/article.php?articleid=1067

http://basketballprospectus.com/article.php?articleid=573

These links breakdown how point differential is a more powerful predictor than record and how close games winning percentages are random.

A more powerful predictor of what?  No one is arguing that winning games by big margins would tend to indicate a team in pretty good.
Mike

You were actually arguing just that earlier in the thread. You argued that points differential is a weaker predictor than raw record. (along with general dislike of Hollinger's ratings)

No.  What I said is you are what your record says you are.  If a 6-4 team is actually better than 6-4, its record will eventually reflect that going forward and I fully expect Miami will be better than 12-8 after 20 games or 18-12 after 30 games.  My objection is people waving around point differential as though its the only thing that matters or the only thing that demonstrates how good a team is.

Mike