Hollinger argues Minn worst then NJ. What's the point?

[Fafnir]

Thats a great way to put it Fan from VT.

[BballTim]

Coaching strategy is what effects pace the most in my view, but I'm not going to try and use that to discount what statistics say about teams.

What effects pace the most is the opposing team.

Pace numbers this early in the season are meaningless.
It would be nice if every team would play every opponent the same number of times, since that messes with so many team-oriented statistics.
Injuries are also not part of the pace stat or the efficiency stats.
It doesn´t account for matchups, different refereees or certain changes of team strategies by the coaches over the course of a season.

To be quite honest, all these holes make me wonder how anyone can give a flying rat fart about these stats.

  I don't think it's that early for pace. The top 6 teams in pace are, though the order's different, the top 6 from last year. 4 of the 5 slowest paced teams from last year are in the bottom 5 this year. Injuries can affect pace but by and large teams won't change that much from where they are now.

[MBunge]

As long as the sample size is sufficiently large point differential is a superior indicator of team strength than W-L record.

If you're talking about fantasy leagues or computer games, sure.  In the real world, wins and losses are what count.

Mike

No in real life point differential is a better predictor of future success, both in the regular season and playoffs. I never said point differential is what "counts". I should have said its a better predictor for future results, that is what I meant. In  the end who raises the banner is what counts.

Since 2000, point differential in the regular season has only correctly foreshadowed the winner of the NBA championship 4 times.  Having the most wins has also been right 4 times.  2 times PD was right and wins were wrong, while 2 times wins were right and PD was wrong.

Mike

[Casperian]

All your criticisms are generally true, but there is always a middle ground between "stats are always right and tell you everything you need to know" and "stats are totally worthless and tell you nothing".  Anything that happens on an NBA court is multiply determined; some of the things that cause each event can be quantified and others cannot.  While we'll never be able to quantify things well enough to predict 100% of what events will happen in the future, we can do it well enough to predict well over chance accuracy, and well over the predictive ability of more conventional stats like W-L (don't forget, wins and losses are a statistic just like any other).  

People (not necessarily you) seem to expect that if there's anything at all that can't be captured by a given stat, then the stat is worthless.  But if pt differential or team efficiency is a better predictor, then it's a better predictor - that's all there is to it.  It doesn't mean there still isn't variance that the stat doesn't explain, because there always will be; it just means there's less unexplained variance when using the superior stat.  Using more than one successful predictor reduces unexplained variance further and makes your predictions more accurate as a whole.

Meanwhile, people like Hollinger continue looking for better predictors and ways to tweak current predictors so they explain more variance.  Over time, we get better at anticipating future events. Evidence has shown differential and other higher-level stats to generally be significantly better predictors of future success than the W-L statistic.  That's all there is to it.

I guess you kind of caught me there, since my points of criticism could be indeed applied to almost any stat, if viewed in a vacuum. However, unlike many other stats, the W-L record can make a definitive statement, and that is wether you make the playoffs or not.  

Anyway, it´s absolutely possible that I´m not smart enough to see in which combination a stat like pace would gain any significance.

[Fafnir]

As long as the sample size is sufficiently large point differential is a superior indicator of team strength than W-L record.

If you're talking about fantasy leagues or computer games, sure.  In the real world, wins and losses are what count.

Mike

No in real life point differential is a better predictor of future success, both in the regular season and playoffs. I never said point differential is what "counts". I should have said its a better predictor for future results, that is what I meant. In  the end who raises the banner is what counts.

Since 2000, point differential in the regular season has only correctly foreshadowed the winner of the NBA championship 4 times.  Having the most wins has also been right 4 times.  2 times PD was right and wins were wrong, while 2 times wins were right and PD was wrong.

Mike

Far too small of a sample size to draw any conclusion. Additionally what you really would want to look at is all plaoff series individually. If you did that over the same period I think you'll find that point differential has an advantage.

[MBunge]

As long as the sample size is sufficiently large point differential is a superior indicator of team strength than W-L record.

If you're talking about fantasy leagues or computer games, sure.  In the real world, wins and losses are what count.

Mike

No in real life point differential is a better predictor of future success, both in the regular season and playoffs. I never said point differential is what "counts". I should have said its a better predictor for future results, that is what I meant. In  the end who raises the banner is what counts.

Since 2000, point differential in the regular season has only correctly foreshadowed the winner of the NBA championship 4 times.  Having the most wins has also been right 4 times.  2 times PD was right and wins were wrong, while 2 times wins were right and PD was wrong.

Mike

Far too small of a sample size to draw any conclusion. Additionally what you really would want to look at is all plaoff series individually. If you did that over the same period I think you'll find that point differential has an advantage.

Uh, how about you supply some actual data instead of just pulling something out of your behind?  It's not like the numbers aren't available.  One would think a believer in stats would rely on, you know, stats to make their point instead of conjecture and speculation.

Mike

[Fafnir]

As long as the sample size is sufficiently large point differential is a superior indicator of team strength than W-L record.

If you're talking about fantasy leagues or computer games, sure.  In the real world, wins and losses are what count.

Mike

No in real life point differential is a better predictor of future success, both in the regular season and playoffs. I never said point differential is what "counts". I should have said its a better predictor for future results, that is what I meant. In  the end who raises the banner is what counts.

Since 2000, point differential in the regular season has only correctly foreshadowed the winner of the NBA championship 4 times.  Having the most wins has also been right 4 times.  2 times PD was right and wins were wrong, while 2 times wins were right and PD was wrong.

Mike

Far too small of a sample size to draw any conclusion. Additionally what you really would want to look at is all plaoff series individually. If you did that over the same period I think you'll find that point differential has an advantage.

Uh, how about you supply some actual data instead of just pulling something out of your behind?  It's not like the numbers aren't available.  One would think a believer in stats would rely on, you know, stats to make their point instead of conjecture and speculation.

Mike

I'm at work at the moment so I don't have time to collate 9 years worth of playoffs. After work during the Celtic game I'll be able to shift through the numbers.

Here's a quick article on the subject though:

http://armchairgm.wikia.com/NBA_Point_Differential_-_The_Most_Power_Stat

[MBunge]

As long as the sample size is sufficiently large point differential is a superior indicator of team strength than W-L record.

If you're talking about fantasy leagues or computer games, sure.  In the real world, wins and losses are what count.

Mike

No in real life point differential is a better predictor of future success, both in the regular season and playoffs. I never said point differential is what "counts". I should have said its a better predictor for future results, that is what I meant. In  the end who raises the banner is what counts.

Since 2000, point differential in the regular season has only correctly foreshadowed the winner of the NBA championship 4 times.  Having the most wins has also been right 4 times.  2 times PD was right and wins were wrong, while 2 times wins were right and PD was wrong.

Mike

Far too small of a sample size to draw any conclusion. Additionally what you really would want to look at is all plaoff series individually. If you did that over the same period I think you'll find that point differential has an advantage.

Uh, how about you supply some actual data instead of just pulling something out of your behind?  It's not like the numbers aren't available.  One would think a believer in stats would rely on, you know, stats to make their point instead of conjecture and speculation.

Mike

I'm at work at the moment so I don't have time to collate 9 years worth of playoffs. After work during the Celtic game I'll be able to shift through the numbers.

Here's a quick article on the subject though:

http://armchairgm.wikia.com/NBA_Point_Differential_-_The_Most_Power_Stat

I looked at that article.  Unless I'm reading it wrong, it's only examining the connection between PD and total wins.  In other words, does point differential more accurately predict how many wins a team will have compared to other stats.  Greater point differential = more wins.  That seems entirely sensible and I'm not at all surprised that's what the numbers show.

What's being contended here, however, is that PD is a better reflection of how good a team is INSTEAD of wins.  That, for example, a team with 45 wins and a PD of 4.8 would be more likely to defeat a team with 52 wins and a PD of 3.7.  Or in my example, is having the highest PD in the regular season more predictive of winning the title than having the most wins.

I wouldn't dispute that PD is predictive OF wins.  You and Hollinger are arguing that PD is more predictive THAN wins.  My extremely premliminary look at the numbers says that's not true.

Mike

[scoop]

It's mind-boggling that there are still people that still don't understand how is the game approached in terms of possessions.

Dean Oliver's book Basketball On Paper discusses all these issues at length and has plenty of data.

Btw, what's the problem with pace? Pace isn't a winning factor, no significant correlation was found between pace and efficiency differential. The main factors are eFG%, rebounding, turnovers and FT/FG, both on offense and defense.

Finals from 74 to 07, in terms of efficiency differential and winning record:
http://www.wagesofwins.com/NBAFinalsEffDif74-07.html

(although I'm not sure if people aren't missing the point that ED is a better predictor of future W/L record than past W/L record).

A propos, former Celtics executive Daryl Morey, now with the Rockets, gets the credit for expanding the Pythagorean Formula to the NBA (although the concept that eff. differential was a better measure of quality was already being used by guys like Dean Smith and Hubbie Brown in the 70s):

Part of Morey's brilliance was his ability to recognize a good idea when he saw one. Years earlier, James had created what he called the Pythagorean Theorem of baseball, a method which predicts wins based on runs scored and runs allowed. While at STATS, Morey simply took that idea and tweaked it to make it accurately apply to other sports as well.

"That was probably the first thing I got any notice for," says Morey. "It really was stealing from Bill James — which I was proud to do — and taking his approach in baseball, which shows that run differential is much more predictive of future winning — and also of the quality of your team — than your actual winning. Winning close games is a skill that no team could perpetuate and has never shown a consistent ability to do. So basically, I stole from Bill and showed that the same rules apply to basketball as well."

[nickagneta]

What's mind boggling is the amount of arrogance some people have sometimes when they have a little bit of knowledge.

The overall point I was trying to make way back is that in Hollinger's article he makes a big deal about pointing out that Minnesota is worse than New jersey because of Point Differential.

This is a man who I think invented his own Player efficiency Rating based on 10 or more different separate statistics that are then adjusted based on pace, league averages, and other variables.

This is a man who invented his own Power Rankings formula to measure how well a team is playing based on 8 or more measured variables that are then adjusted based on more variables.

But when he is going to determine statistically which team betwwen two is better he uses one stat, point differential.

All I am saying is that he has a vast amount of other stats available to him, some that he even proudly displays on ESPN.com as "Hollinger Team Stats" and uses none of them to make an argument. He doesn't even use his offensive and defensive efficiency stats except very briefly to say NJ's is 14th, which makes them a decent defensive team, not even mentioning where Minnesota is ranked or how the two compare with offensive efficiency.

He doesn't mention which team turns the ball over more or forces more turnovers. he doesn't mention which team passes better or shoots better or rebounds better. He just runs straight to point differential and discusses little else.

That's what led to my mention of the other stats and thats what led to, I guess, my mistaken defense mention offensive rebounds as different possessions(OMG, how could I have not known that!! Doesn't everyone!!! That's like not knowing 1+1=2 ).

[Fafnir]

That's what led to my mention of the other stats and thats what led to, I guess, my mistaken defense mention offensive rebounds as different possessions(OMG, how could I have not known that!! Doesn't everyone!!! That's like not knowing 1+1=2 ).

Your use of those stats was still flawed nick.

In those stats Minnesota has a higher assist%, a lower turnover%, a higher eFG% and higher TS%, a lower defensive rebound rate, a higher offensive rebound rate and an almost identical rebound rate(48.0 to 48.1) and differential in the offensive and defensive efficiencies(13.8 to 13).

Funny how when Hollinger wants to make stat based arguments he always excludes anything that doesn't prove his point....even his own derived statistics.

Those are all factors in the fact that Minnesota is a better offensive team than NJ. That doesn't really undercut Hollinger's point, which is that Minnesota has a worse efficiency differential than NJ. (Because of their defense is bad)

You are right, Hollinger could have gotten much in more depth on all the factors going into predicting wins. But his Per Diem articles are always pretty short and efficiency differential is the best single stat to look at.

[nickagneta]

That's what led to my mention of the other stats and thats what led to, I guess, my mistaken defense mention offensive rebounds as different possessions(OMG, how could I have not known that!! Doesn't everyone!!! That's like not knowing 1+1=2 ).

Your use of those stats was still flawed nick.

In those stats Minnesota has a higher assist%, a lower turnover%, a higher eFG% and higher TS%, a lower defensive rebound rate, a higher offensive rebound rate and an almost identical rebound rate(48.0 to 48.1) and differential in the offensive and defensive efficiencies(13.8 to 13).

Funny how when Hollinger wants to make stat based arguments he always excludes anything that doesn't prove his point....even his own derived statistics.

Those are all factors in the fact that Minnesota is a better offensive team than NJ. That doesn't really undercut Hollinger's point, which is that Minnesota has a worse efficiency differential than NJ. (Because of their defense is bad)

You are right, Hollinger could have gotten much in more depth on all the factors going into predicting wins. But his Per Diem articles are always pretty short and efficiency differential is the best single stat to look at.

He never mentioned efficiency differential. He mentioned point differential...scoring margin. Two completely different things.

And I am not going to debate you anymore on my interpretation of the stats. I think I'm right. You think you're right. Once again, let's leave it at that.

[Fafnir]

That's what led to my mention of the other stats and thats what led to, I guess, my mistaken defense mention offensive rebounds as different possessions(OMG, how could I have not known that!! Doesn't everyone!!! That's like not knowing 1+1=2 ).

Your use of those stats was still flawed nick.

In those stats Minnesota has a higher assist%, a lower turnover%, a higher eFG% and higher TS%, a lower defensive rebound rate, a higher offensive rebound rate and an almost identical rebound rate(48.0 to 48.1) and differential in the offensive and defensive efficiencies(13.8 to 13).

Funny how when Hollinger wants to make stat based arguments he always excludes anything that doesn't prove his point....even his own derived statistics.

Those are all factors in the fact that Minnesota is a better offensive team than NJ. That doesn't really undercut Hollinger's point, which is that Minnesota has a worse efficiency differential than NJ. (Because of their defense is bad)

You are right, Hollinger could have gotten much in more depth on all the factors going into predicting wins. But his Per Diem articles are always pretty short and efficiency differential is the best single stat to look at.

He never mentioned efficiency differential. He mentioned point differential...scoring margin. Two completely different things.

Er they are and they aren't. They're slightly different, but both tell pretty much the same story. How much you score and how much you're scored on. One is just regularized a bit for pace.

Fundamentally you won't get different conclusions from using one instead of the other.

[scoop]

]He never mentioned efficiency differential. He mentioned point differential...scoring margin. Two completely different things.

Completely different? Why? The only difference is the adjustment for pace. As their difference in pace is about 3 possessions per game and the differential remains basically the same I don't see what makes such a big difference.

And Hollinger uses other arguments besides point differential in the article, including some non-quantitative ones.

In any case, different metrics measure different things. He used the most proper metric to make his case. I guess he could have developed it and explain what factors contribute more to offensive and defensive efficiency, but in my view that would go way beyond the scope of the article.  

[fairweatherfan]

Er they are and they aren't. They're slightly different, but both tell pretty much the same story. How much you score and how much you're scored on. One is just regularized a bit for pace.

Fundamentally you won't get different conclusions from using one instead of the other.

True, the ratio of pts scored/pts allowed and off/def efficiency should always be practically identical; efficiency's just the same data adjusted for pace.