LOL @ Hollinger

[TheNorthCydeRises]

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

He's going to act like the biggest Hawk fan ever, if we keep winning, and his "formula" keeps us near the top. That's what he gets for overvaluing Chill, like he was the most vital piece to our team.

[dlpin]

Normally I wouldn't defend Hollinger because I think basing everything on numbers alone leads to some inconsistencies.

But here I find myself defending him. He is not "acting" like a hawks fan. His ranking are based purely on his formula, and he does not have any reason to skew them in any way. After all, his career success depends on him being accurate.

There are two things that are the key to his formula, more than anything else: margin of victory (really THE key stat in his formula) and strength of schedule. So hawks are number 2 not because he skewed the numbers to show so, but because the hawks have the second largest margin of victory average and the 10th largest "strength of schedule" score.

[Packfill]

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

He's going to act like the biggest Hawk fan ever, if we keep winning, and his "formula" keeps us near the top. That's what he gets for overvaluing Chill, like he was the most vital piece to our team.

I am LOL at all the people who think Hollinger has an agenda other then pushing his statisical rankings.

[mrhonline]

I would think that scoring margin would tend to be highly volatile in a smaller sample, which means we probably can make even less of Hollinger's ranking the Hawks at #2 than the other rankings.

[crimedog]

I am just always skeptical of Hollinger's work, someone has to question his methods though.

No doubt. I like hearing your gripes about his stuff a lot more than the people who just blindly say, "Lol at him, I bet he doesn't even watch our games."

[Admin]

Hollinger is certainly one of the more Hawks knowledgeable (and perhaps the most) of the national media and he's quick to admit that he was wrong so I give him major props for that. I think a lot of the reason why some don't like him is that over the past decade we've become a bit like one of those mangy mutts that get beaten over and over again and are quick to bite whenever someone comes near us. We've had nothing but disrespect from everyone for a decade now and we're just too sensitive as fans because of it.

[AHF]

I posted this somewhere else, but I wonder what Hollinger's intuition behind using margin of victory and not win-loss at all is. He simply stated in his explanation that:

"This might not sound right at first, but studies have shown scoring margin to be a better predictor of future success than a team's win-loss record. Thus, scoring margin is a more accurate sign of a team's quality."

Any idea on what studies actually led him to this? While I agree that margin of victory is a good predictor, I get the feeling that Hollinger has some twisted reason behind using it. I would guess he fooled around with his formula long enough to get his desired results and then to describe the formula he went through and found "studies" to describe it. I am just always skeptical of Hollinger's work, someone has to question his methods though.

I don't know where he got the numbers but I suspect someone did a basic statistical analysis correlating actual end season results and what the numbers indicated at a given point (or did a combined study of the predictions every game for a period of the season) based on margin versus record.

I'd be interested to hear if anyone gets an answer from him on that.

Either way, I agree with those who believe there is no agenda for Hollinger re the Hawks other than promoting his own numbers. I am sure he would like them to do well so he can see some better games/more high profile games in Atlanta.

[AHF]

I would assume it has to do with Pythagorean Wins. It states your number of wins should be roughly equal to:

Games*(team points^14 / (team points^14+opponent points^14))

There have been analysis done and it shows to be a fairly reliable predictor. Its root means square error is 3.14 wins when using 14 as the arbitrary exponent (essentially saying the predicted value is within 3.14 wins (plus or minus) of the actual value). It was applied from Bill James who did this for baseball with great results.

I am just skeptical whether or not this is Hollinger's reasoning. He never explicitly states this in his description of the rankings. He just says "studies", which is vague language.

No, I agree that it is vague. I would assume that numerous formulas like the above have been tested against past seasons to determine which has the best predictive value, but you know the expression about assuming...

[crimedog]

No, I agree that it is vague. I would assume that numerous formulas like the above have been tested against past seasons to determine which has the best predictive value, but you know the expression about assuming...

From what I've read (just off espn.com) he mentions a lot of stuff about how it generally predicts playoff performance better than record (San Anto for their last couple of titles, that Jersey team that got beat by Tim Duncan and some other guys, stuff like that). I'm sure there are more in depth analyses done somewhere but I haven't read his book and I'd imagine he lays a lot of his stuff out in there but doesn't want to spill too much otherwise people wouldn't buy it (sposing anyone buys it).

[dlpin]

From what I remember him saying, the basic insight is that close games are true toss ups decided by random factors, as opposed to decisive wins. That is, the rationale is that blow outs tell you more about a team's skill than tight games. So a team that decisively won 50 games and lost the other 32 by 1 or 2 points is better than a team that decisively won 30 games, and just squeaked by winning another 25 at the last second. Case in point last night's hawks loss: that PP shot was at best a 50% shot, so the loss (or win) was more due to luck and randomness than true decisive superiority.

My problem with Hollinger is that he creates these "black boxes" and then places blind faith in them. Methodologically, my problem is that his statistics are purely descriptive. He makes no effort to model trends and swings, or even specific match ups, other than arbitrarily discounting certain numbers over others.

Case in point, last year's playoffs when he underestimated the cavs (the hawks caught everyone by surprise, even though a more statistically savvy person would have modeled how the Celtics did against athletic teams as opposed to just looking at the overall number) despite their improvements late, and later underestimated the celtics in the final. Explicitly including trends and other momentum and experience indicators in a true model as opposed to a descriptive statistic would have allowed him to be much more precise in forecasting the cavs success against the wizards and the celtics success against the lakers.