Hollinger Humor

[Birdbrain]

Pt diff will be the death of the knowledgeable sports fan. 

[Mike-Dub]

Miami beating Cleveland by 6 points is enough for them to pass Boston on his rankings because they only beat the Knicks by 2. 

Holliner does not deserve a job covering the NBA or basketball in general.  Along with many of the other analyst out there.

He needs to get away from his formula sometimes and use common sense.

[BballTim]

How can somebody as tied to stats as much as Hollinger look at those numbers, and say "yeah, Miami is better"?  Ultimately it doesn't matter, but you'd think that he'd want to tweak his formula so that it could pass the smell test.

The rankings are 110.441 for the Heat and 110.371, I don't think that .06 of a point in his rating system says that there is much separation between the two teams.

If you look at his projections they are essentially identical for the two teams, both are projected to win the title 30% of the time.

Regardless, Miami is ranked higher, despite Boston losing four fewer games, beating Miami twice head to head, having played a tougher schedule (and a tougher schedule in the last ten), having played fewer home games and more road games, and having an almost identical margin of victory.

You really think, based on all of the above, that Miami should be ahead, or even tied with, Boston?  That seems like a silly conclusion, leading me to believe that it's a silly formula that needs to be tweaked.

  I don't know that the problem is his formula, per se, but more of a smallish sample size. A lot of things generally smooth out better over, say, 60+ games. Teams can be on hot (or cold) streaks now that comprise half of their season to date, and schedules should somewhat even out.

[celticinorlando]

biggest hack around

[guava_wrench]

Miami beating Cleveland by 6 points is enough for them to pass Boston on his rankings because they only beat the Knicks by 2. 

No it isn't.

The last 10 games are weighted more heavily. Obviously, when a game is played, another game gets knocked out of the last 10 games. For Miami, that game was a loss. For us, it was probably better than a 2 point win.

They currently have a really large margin of victory. It is what it is.

Think about it. If Amare gets the final 3 off in time and the Knicks win by 1, how much of a ranking penalty is that 3 points worth? It is easy to give too much weight to record when the difference between a win and a loss is only 2 points, or in last night's case, .3 seconds.

The heavier weighting of the last 10 games is very defensible since teams change over the course of the season as they gel. Miami is likely a good example of this.

[Chris]

Yup, still hate this guy.

I hate basketball stats.  I actually think Hollinger is an intelligent guy, and a good writer...this is just his role at ESPN.

[Taklamar]

Straight from Hollinger's "Intro and Method" link in the upper right of the power rankings, explaining the rankings;

"I created these rankings to give a quick assessment of all 30 teams so far in the season, since sometimes the standings can be misleading in this department."

"One of my goals was to create a system that told us more about a team's quality than the standings do.

So instead of winning percentage, these rankings use points scored and points allowed, which are better indicators of a team's quality than wins and losses.

So instead of winning percentage, these rankings use points scored and points allowed, which are better indicators of a team's quality than wins and losses.

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."

"Another key variable in the formula is recent performance, which I included for two reasons.

First, it stands to reason that more recent games are more valid indicators of how strong a team is currently.

Second, I wanted these rankings to follow the model of Marc Stein's "human" power rankings, on the site each Monday, in which a team's recent play is a huge factor."

"Caveats
Since this is an entirely automated ranking, you'll notice certain "human" factors missing.

It doesn't know which players are about to come back from injury or which teams have been playing without their best players for the past 10 games.

Along the same lines, it doesn't take into account injuries, trades, controversial calls or any other variables -- just the scores, please.

Nonetheless, it can be very useful because it allows us to see what the landscape looks like when we remove our usual filters.

We hope you enjoy our daily power rankings."

=============================================================

He created a *tool*, to try to give a better and deeper look at current team strength over W/L record.  An *objective* look.  To get people to look at stats and have a better understanding of them and to promote discussion.  Nothing more, nothing less.

You can argue about how he does the weights of various factors, sure.  Overall though, by the end of the season, his rankings are pretty solid.

As far as being an idiot or not knowing anything about the game...no clue how anyone could say that.  If you've read his articles and chats, he obviously is very knowledgeable about basketball and an intelligent guy.  He takes a more stats-based angle than a subjective angle on basketball.  You can dislike that, but calling the man an idiot and unknowledgeable is petty and ignorant.

/rant off

Tak

[fairweatherfan]

As far as being an idiot or not knowing anything about the game...no clue how anyone could say that.  If you've read his articles and chats, he obviously is very knowledgeable about basketball and an intelligent guy.  He takes a more stats-based angle than a subjective angle on basketball.  You can dislike that, but calling the man an idiot and unknowledgeable is petty and ignorant.

/rant off

People tend to be threatened by alternative perspectives, especially ones that are somewhat difficult to understand and contradict what they currently believe.  The easiest mental out is to conclude that the alternative perspective is worthless and stupid.  This justifies both the person's preferred viewpoint and their lack of understanding of the rationale behind the alternative view (why bother to try to understand something that's stupid and worthless?) You see the same thing all the time in politics, for example.

There's plenty of room for legitimate criticism of methods like Hollinger's, and there's lots of that in this thread too, but this is generally why you see so much anger and hyperbole like the stuff you're talking about.

[Brendan]

1. As a fan who gives a hoot if Hollinger has a team ranked low or high?

2. Second if you cannot understand what the tool he has produced is for, don't complain.

3. The tools is a first pass at team quality - you have to then add all subjective factors AFTER you look at this ranking. So you look at this and say "OK but the C's are missing all their big men and beat the Heat head to head, but are behind by a tony fraction, we'll give them the nod."

4. There's a limit to how complicated he's going to make it - for example I could see figuring out an adjustment based on head to heads, but its probably very complicated to do so, both analytically and data processing-wise.

The first writer worth complaining about at ESPN is Thorpe, once he's gone I'll consider others.

[dlpin]

Miami beating Cleveland by 6 points is enough for them to pass Boston on his rankings because they only beat the Knicks by 2. 

That is not the reason the heat passed the celtics. the reason for that is that his rankings weigh the last 10 games more heavily. That is, performance over the past 10 games count for a great deal. So the 2 point win over the knicks replaced the 23 point win over the hawks, so the celtics' score went down, while the heat's 6 point win replace an 11 point loss to the mavs in the "last 10" score.

[nickagneta]

Hollinger's problem is his weighting system that puts so much emphasis on margin or victory in the last ten games over everything else.

Tweaks I would make would be a deduction for losses weighted according to the winning percentage of the team lost to, a deduction for bad losses(losses to by more than a certain margin), and a deduction for every loss against a team within 1 point of your final score.


These tweaks could be the difference in where to properly place two teams that under Hollinger's current system have similar numbers and yet one team has beaten the other a couple of times.

Also take into effect here that his system used L10 as a current indicator but once the season is more than 40 games old he goes to L25% of games. It makes for less fluctuations later in the year.

Still though, I've never been a fan of his formulations and will continue to not be a fan of them. As Roy said, many times, they just don't pass the smell test. Sort of like that small company that barely ever makes a profit and if it does it's barely anything but has been in business for 50 years and the owners are multimillionaires. Math shows that the business struggles. Smell test says the accountants are having fun with the numbers. It's all good math, just doesn't pass the smell test.

[Roy H.]

A few things:

1. Just because somebody rejects Hollinger's formula doesn't mean that they reject all statistical models. Hollinger's seems flawed, in part because it totally rejects H2H and winning percentage. Similarly, there are better measures of SOS than simply looking at opponent's winning percentage.

2. Hollinger's model may be based upon data rather than emotion, but "objective" doesn't mean" accurate". It's a formula created by a fallible man. It's like arguing that the BCS computers should determine the national champion, because they're" objective".

[LooseCannon]

2. Hollinger's model may be based upon data rather than emotion, but "objective" doesn't mean" accurate". It's a formula created by a fallible man. It's like arguing that the BCS computers should determine the national champion, because they're" objective".

The computer rankings used in the BCS are examples of formulas that are tweaked because people don't like the results.  Unlike Hollinger's formula, they don't take margin of victory into consideration (an adjustment because the "wrong" teams were being ranked too highly).

As mentioned, one reason Hollinger does his ranking is to promote discussion.  The continued existence of threads like this are what validates his model.

[Fan from VT]

A few things:

1. Just because somebody rejects Hollinger's formula doesn't mean that they reject all statistical models. Hollinger's seems flawed, in part because it totally rejects H2H and winning percentage. Similarly, there are better measures of SOS than simply looking at opponent's winning percentage.

2. Hollinger's model may be based upon data rather than emotion, but "objective" doesn't mean" accurate". It's a formula created by a fallible man. It's like arguing that the BCS computers should determine the national champion, because they're" objective".

For point #1, though, in order to criticize this particular method for rejecting head to head and winning percentage, don't you have to provide some evidence that winning percentage and head to head matter significantly when trying to predict future results? I think those two criteria are rejected for a reason: they have historically had little to do with what happens in the future relative to scoring margin. In order to add variables to the equation, there would have to be a consistent correlation of those criteria to the outcomes trying to be predicted.

Edit-I agree with the SOS criteria though. I've actually written him a couple times to see if he'd address this in a PERdiem, but he hasn't. I think it is absolutely flawed that he used this system that explicitly ignores winning percentage on the basis that it is a poor predictor of true ability, yet uses win percentage to formulate SOS! I basically asked him why he doesn't use "opponent margin of victory" instead of SOS if margin of victory is a better surrogate for team strength than win percentage. Or, if power rankings is really a good measure of how well a team is playing at a given time, why couldn't you replace SOS with "Opponent Power Ranking at time of game?" Seems like both would be better.

[Brendan]

A few things:

1. Just because somebody rejects Hollinger's formula doesn't mean that they reject all statistical models. Hollinger's seems flawed, in part because it totally rejects H2H and winning percentage. Similarly, there are better measures of SOS than simply looking at opponent's winning percentage.

2. Hollinger's model may be based upon data rather than emotion, but "objective" doesn't mean" accurate". It's a formula created by a fallible man. It's like arguing that the BCS computers should determine the national champion, because they're" objective".

For #1 - the OP does explicitly reject "basketball statistics". And on this website, mostly the complaint is "his stat model ranks team X (usually LAL and wherever Lebron is playing) too high versus team Y (usually the C's.)"

It would be interesting to know why he uses that for SoS. I'm guessing that he was limited in how much time he could put into it, and what stats he could use (I'm guessing ESPN didn't want him to use something from a bball content site for example.) Additionally he's probably trying to keep things on the simple side: most casual fans repudiate the use of second order stats already, making it too complicated confuses things. But what's the actual target here? Are we expecting him to refactor the formula continually or is it good enough that even in cases like this (where one team is marginally better than the other despite some compelling evidence to the contrary) he's getting the numbers pretty close to target. Should he spend another 120 hours improving the formula so BOS' rank moves a few points and MIA's falls a few points? What's the ROI in that for him or ESPN?

Anyways with anything like this you really need a confidence interval which I'm sure he would say is there is no difference between Heat and BOS.