Hollinger Rankings

[BudweiserCeltic]

http://sports.espn.go.com/nba/hollinger/powerranking

Celtics are currently #4, and considering the stats, they should be #3 in my opinion. The current problem with the formula is that the strength of schedule is being counted twice since we've played few games and it's early in the season. A couple of weeks more and the rankings should start to reflect the current status in the league more accurately.

A win tonight should give the Celtics a bump...

[indeedproceed]

Hollinger's rankings are mindless products of equations...they need a larger sample size.

[Celtic]

He has the Pacers at #6 so I wouldn't take this one too seriously.

[LB3533]

We haven't crushed our opponents like last year either.

[Who]

I'm surprised he put it up so early. I thought he needed a larger sample than this to get some decent results.

[crownsy]

I'm surprised he put it up so early. I thought he needed a larger sample than this to get some decent results.

i would say some of the top 15 bear that out.

[BudweiserCeltic]

I'm surprised he put it up so early. I thought he needed a larger sample than this to get some decent results.

It's fine, since they update each day after games... so at least we can entertain ourselves witnessing the day to day progress.

So as long as we keep than in mind, it's fine. Problem comes when people don't apply the results in that context.

[Hoops]

The key to Hollinger's stuff is that you have to view it with a little perspective. It's just a piece of the puzzle (and one that happens to be more valuable as the sample size increases). It's just data - and I think data is interesting if you look at it with the right perspective. It's not a matter of whether Boston's ranking is "right" or "wrong" - their ranking simply "is" (how's that for some philosophical depth?)

I suppose one can argue about the data models Hollinger is using, but I don't see how anyone can get hot and bothered about the results. Example: x + y = z. Z is simply the result of adding x and y - it's not really arguable. What is arguable is whether x and y are the right variables to use in the first place.

Anyhow, enough of my mathematical/philosophical musings...

[nickagneta]

So it's six days later and the Celtics now currently sit third in the Hollinger Rankings barely a half game ahead of the Atlanta Hawks, a team the Celtics beat and who have lost their last three straight games.

http://sports.espn.go.com/nba/hollinger/powerranking

I understand that Hollinger's rankings are based on a long computation of different variables but perhaps he should de-emphasize the Margin of Victory variable because I don't see how numbers can come up with a position that the Celtics are third in the league, barely ahead of a team they beat that has more losses than they do, when they have had the 2nd hardest schedule, won the most games, and have the second best winning percentage. The only thing I see that they lag behind in versus Cleveland(the team in second place that the Celtics also beat) and LA in is Margin of Victory.

Does this man understand that Margin of Victory/Strength of Schedule should be an integrated number and not seperate variables and that it's graph should be an exponential function and not a linear one. The harder your SOS the lower the MOV but there's no way that is linear. And considering the Celtics SOS is 3% more difficult than the Cavs and 16%+ more difficult than the Lakers, that should be taken into consideration more.

Also, victory over Teams of Quality should have a bonus kicked in. A team could conceivably perform at it's best against quality teams and win more of those games. Those wins should account for more than wins over inferior competition.

I'd also love to see his study that shows MOV is a better determining factor for success than W-L percentage.

[Fafnir]

The Hawks have played one more road game than the Celtics. Because there is an adjustment to margin of victory for home and road games that is likely what is bumping up the Hawks right now.

Hollinger's power ranking isn't a bad tool once you've got a decent sample size, but this early in the season there just isn't enough data. To either use it or criticize it.

[BudweiserCeltic]

So it's six days later and the Celtics now currently sit third in the Hollinger Rankings barely a half game ahead of the Atlanta Hawks, a team the Celtics beat and who have lost their last three straight games.

http://sports.espn.go.com/nba/hollinger/powerranking

I understand that Hollinger's rankings are based on a long computation of different variables but perhaps he should de-emphasize the Margin of Victory variable because I don't see how numbers can come up with a position that the Celtics are third in the league, barely ahead of a team they beat that has more losses than they do, when they have had the 2nd hardest schedule, won the most games, and have the second best winning percentage. The only thing I see that they lag behind in versus Cleveland(the team in second place that the Celtics also beat) and LA in is Margin of Victory.

Does this man understand that Margin of Victory/Strength of Schedule should be an integrated number and not seperate variables and that it's graph should be an exponential function and not a linear one. The harder your SOS the lower the MOV but there's no way that is linear. And considering the Celtics SOS is 3% more difficult than the Cavs and 16%+ more difficult than the Lakers, that should be taken into consideration more.

Also, victory over Teams of Quality should have a bonus kicked in. A team could conceivably perform at it's best against quality teams and win more of those games. Those wins should account for more than wins over inferior competition.

I'd also love to see his study that shows MOV is a better determining factor for success than W-L percentage.

The current problem with the powerrankings at the moment is that it has various variables that get counted twice because of the "last 10 games" factor. So some variables are having more weight than others, like SOS and the Margin. That's because we've only played about 10 games.

Once the we play more games, and "the last 10 games" are a separate entity rather than a duplicate, then the stats should be more "real".

[Hoops]

So it's six days later and the Celtics now currently sit third in the Hollinger Rankings barely a half game ahead of the Atlanta Hawks, a team the Celtics beat and who have lost their last three straight games.

http://sports.espn.go.com/nba/hollinger/powerranking

I understand that Hollinger's rankings are based on a long computation of different variables but perhaps he should de-emphasize the Margin of Victory variable because I don't see how numbers can come up with a position that the Celtics are third in the league, barely ahead of a team they beat that has more losses than they do, when they have had the 2nd hardest schedule, won the most games, and have the second best winning percentage. The only thing I see that they lag behind in versus Cleveland(the team in second place that the Celtics also beat) and LA in is Margin of Victory.

Does this man understand that Margin of Victory/Strength of Schedule should be an integrated number and not seperate variables and that it's graph should be an exponential function and not a linear one. The harder your SOS the lower the MOV but there's no way that is linear. And considering the Celtics SOS is 3% more difficult than the Cavs and 16%+ more difficult than the Lakers, that should be taken into consideration more.

Also, victory over Teams of Quality should have a bonus kicked in. A team could conceivably perform at it's best against quality teams and win more of those games. Those wins should account for more than wins over inferior competition.

I'd also love to see his study that shows MOV is a better determining factor for success than W-L percentage.

Interesting points - I don't necessarily disagree. It's one silly example, but his power rankings did have the Celtics on top almost all year last year and the C's did, in fact, prove to be the best team...

[cordobes]

I'd also love to see his study that shows MOV is a better determining factor for success than W-L percentage.

Well, I have no idea if Hollinger has a study on this, but that's basically the Pythagorean method used to estimate win %. Yes, it's better, there's not even need for empirical evidence, there's a theoretical explanation. Anyway:

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

http://www.rawbw.com/~deano/articles/JordvsOlaj/linwtscorr.html

[Boris Badenov]

One of the things that is a little frustrating about Hollinger's numbers is that he doesn't open the "black box." I understand why, because he views the underlying model as intellectual property that he has to protect, but it does lead to confusion about what's driving his results.

And I do agree that if in the first ten games he's double-counting some numbers, that is a problem. The simplest fix would be to wait a little longer before posting the rankings. A more correct fix would be to continuously re-weight the numbers as the number of games changes. That wouldn't be too hard.

[Fafnir]

One of the things that is a little frustrating about Hollinger's numbers is that he doesn't open the "black box." I understand why, because he views the underlying model as intellectual property that he has to protect, but it does lead to confusion about what's driving his results.

And I do agree that if in the first ten games he's double-counting some numbers, that is a problem. The simplest fix would be to wait a little longer before posting the rankings. A more correct fix would be to continuously re-weight the numbers as the number of games changes. That wouldn't be too hard.

Actually doesn't he break down his power rankings formula? link
He has other statistics that he doesn't reveal the formulations, but power rankings seems straight forward.