Hollinger Forecast: Blazers

[Ed O]

If it makes you feel better to call me "ignorant", so be it, but show me a model that effectively predicts luck, and then show me the model that makes luck an absolute with a universal definition. It's an easy way out for Hollinger to throw "luck" into the mix; he does not dive deep enough into it for me to do anything other than mock him for his ridiculous inclusion of it.

Uh. OK.

Maybe all of statistical analysis of baseball and basketball done to date is wrong, and you're right.

Maybe.

Ed O.

 

[KingSpeed]

Hollinger has no idea what he's talking about. We have great three point shooting. And we had injuries last year. Roy, LMA, and Jones all had injuries. But of course, you all know this.

 

[PapaG]

Uh. OK.

Maybe all of statistical analysis of baseball and basketball done to date is wrong, and you're right.

Maybe.

Ed O.

Please expand on this. Any model that uses "luck" needs to affix the same value to "luck" across the board and for all parties. If not, the model is useless, since luck is an unknown and is subjective in nature. Luck is an emotional reaction.

You say "all of statistical analysis". Let's take baseball. A batting average is a batting average. One can say, well, X amount of balls should not have been hits based on data, but the fact remains that they do end up hits, and the fact remains that no one can predict which soft line drives will fall and which lasers will be caught.

Hence, forward thinking statistical analysis needs to have the same limitations as historical data, and then Hollinger takes it one ridiculous step further by subjectively attributing "luck" to a historical known, i.e. the Blazers 41 wins from last season.

 

[Ed O]

Please expand on this. Any model that uses "luck" needs to affix the same value to "luck" across the board and for all parties. If not, the model is useless, since luck is an unknown and is subjective in nature. Luck is an emotional reaction.

Dude. I gave you a link to read about Expected Wins, and there is some explanation of how luck plays a role there. I don't have the time or inclination to give you more links when there are so many sabermetric and other sources online to explain the role that luck can play in statistical analysis.

Ed O.

 

[PapaG]

Dude. I gave you a link to read about Expected Wins, and there is some explanation of how luck plays a role there. I don't have the time or inclination to give you more links when there are so many sabermetric and other sources online to explain the role that luck can play in statistical analysis.

Ed O.

I am well aware of these analyses, but they all use historical data to attach a value to "luck".

Unless Hollinger gives us which wins or statistics were based on "luck", he looks foolish. Bill James is rolling over in his grave (that's a joke).

You are comparing the work of an undergrad (Hollinger, in this instance) to the work of the professor (James, for one).

 

[Minstrel]

Luck is inherent and an unknown. Hollinger can't quantify luck, so trying to make a supporting argument out of luck is ridiculous. Luck is an unknown; Hollinger may think the Blazers were "lucky" to win 41 games last year; I think they were unlucky to only win 41 games considering the season as a whole and the entire injury problem, Oden included.

I think you're misunderstanding the use of the word "luck" in statistical analysis. Luck refers to variance. All performance has variance up and down. The difference between "expected performance" and actual performance is luck, if you believe in the model that provides the expected performance. It's unknown whether a team will have good or bad luck (or any luck), but for any team that greatly under-performs or over-performs its expected performance, it is rational to expect a regression to the mean.

He's counting Oden as a new player this year, so his injury last year isn't related to his contention that the Blazers were lucky. The Blazers, as they were comprised without Oden, had a lucky season by Hollinger's model. So, if you are looking at last season's record as the baseline for what to expect this season, some of the gains from gaining Oden, Bayless and Fernandez will be lost in regression to the mean.

Unless they again over-perform Hollinger's expected performance. Which could happen. I am not speaking to whether Hollinger's numbers are accurate or the best, but if you believe in his model (as, of course, Hollinger does), that use of "luck" is perfectly reasonable, in my opinion.

 

[Minstrel]

and i'll never understand this. hollinger has heard scouts make a comparison to robinson, so he won't dismiss that? really? hollinger can't think for himself? i'd hope an espn "expert" (even one that focuses on stats) would actually watch some guys play and be able to make their own judgements.

I think this is an odd nit-pick. He's not using other people's judgments instead of his own. He's saying that since many informed people believe that Oden has Robinson-like potential, he can't discount the possibility that Oden will have a Robinson-like rookie season. It's being open-minded and saying that something he doesn't think will happen does exist as something possibly could happen.

 

[oldmangrouch]

According to "pythagorean" analysis, our record last year should have been 37-45. That is a "luck neutral" estimate.

Studies in both basketball and baseball have shown that winning close games is not a "skill" that is retained from season to season. The pythagorean analysis indicates that we won an unusually high number of close games last season. Call it "luck", or an anomaly, or divine intervention - it is not a trend you can rely on.

 

[rocketeer]

I think this is an odd nit-pick. He's not using other people's judgments instead of his own. He's saying that since many informed people believe that Oden has Robinson-like potential, he can't discount the possibility that Oden will have a Robinson-like rookie season. It's being open-minded and saying that something he doesn't think will happen does exist as something possibly could happen.

he says "since I've heard scouts make that comparison I'm not going to dismiss the possibility" which to me is an incredibly stupid comment to make.

 

[Ed O]

he says "since I've heard scouts make that comparison I'm not going to dismiss the possibility" which to me is an incredibly stupid comment to make.

He's not a scout. He's not a professional talent evaluator. He works with statistics and models and projections and analysis. He relies on people who get paid to scout players to provide the qualitative input.

It's like a scout who's writing a column saying that Hollinger points out that a player's PER was higher than the average for a guy at his position.

Why is that an incredibly stupid approach?

Ed O.

 

[number 10]

Hehe.

That's 5 more wins than Hollinger is being crucified for predicting... does anyone really think that he expects his predictions/projections to have THAT much accuracy?

Ed O.

It's Hollinger's own fault for being so specific instead of using a range of wins.

 

[gatorpops]

I think you're misunderstanding the use of the word "luck" in statistical analysis. Luck refers to variance. All performance has variance up and down. The difference between "expected performance" and actual performance is luck, if you believe in the model that provides the expected performance. It's unknown whether a team will have good or bad luck (or any luck), but for any team that greatly under-performs or over-performs its expected performance, it is rational to expect a regression to the mean.

He's counting Oden as a new player this year, so his injury last year isn't related to his contention that the Blazers were lucky. The Blazers, as they were comprised without Oden, had a lucky season by Hollinger's model. So, if you are looking at last season's record as the baseline for what to expect this season, some of the gains from gaining Oden, Bayless and Fernandez will be lost in regression to the mean.

Unless they again over-perform Hollinger's expected performance. Which could happen. I am not speaking to whether Hollinger's numbers are accurate or the best, but if you believe in his model (as, of course, Hollinger does), that use of "luck" is perfectly reasonable, in my opinion.

So, what is the value of "expected preformance" if luck has some "value"? So in his "expected preformance" is lower than mine ----?????????? Luck affects both up and down seems to me?

g

 

[Nikolokolus]

I wouldn't take too much offense about Hollinger's prediction, it's not like he's just randomly throwing numbers out there, he does have a formula (or formulae) and I tend to think that this team overachieved last year, and our .500 mark also came with a negative points differential.

Hollinger's a bit conservative, but I've thought for awhile that this team is probably going to win about 46 games next season and if that happens and they get an 8th or 7th seed out of the deal I think we should all consider ourselves fortunate, because that will also mean almost everyone on the team stayed healthy and nobody had a big fall off from the year before.

Viva la Blazers!

 

[gatorpops]

I stick by my prediction that the Blazers win 50 and their luck is that they win 50.


g

 

[PapaG]

I think you're misunderstanding the use of the word "luck" in statistical analysis. Luck refers to variance. All performance has variance up and down. The difference between "expected performance" and actual performance is luck, if you believe in the model that provides the expected performance. It's unknown whether a team will have good or bad luck (or any luck), but for any team that greatly under-performs or over-performs its expected performance, it is rational to expect a regression to the mean.

He's counting Oden as a new player this year, so his injury last year isn't related to his contention that the Blazers were lucky. The Blazers, as they were comprised without Oden, had a lucky season by Hollinger's model. So, if you are looking at last season's record as the baseline for what to expect this season, some of the gains from gaining Oden, Bayless and Fernandez will be lost in regression to the mean.

Unless they again over-perform Hollinger's expected performance. Which could happen. I am not speaking to whether Hollinger's numbers are accurate or the best, but if you believe in his model (as, of course, Hollinger does), that use of "luck" is perfectly reasonable, in my opinion.

True, but objective variance has a plus and a minus value. Hollinger attributing a minus solely to the "luck" value is subjective data. Even with that, he pegs Portland at 38 wins, which is a -3 game difference out of 82, and which therefore has a P-value that isn't statistically significant if the standard of .05 is followed. So basically, we are arguing about nothing.

 

[Minstrel]

he says "since I've heard scouts make that comparison I'm not going to dismiss the possibility" which to me is an incredibly stupid comment to make.

Why? I explained why I didn't think so...that he was allowing that the possibility exists, due to other informed opinions. Why is it stupid to admit the possibility, due to what scouts think?

 

[Minstrel]

True, but objective variance has a plus and a minus value. Hollinger attributing a minus solely to the "luck" value is subjective data. Even with that, he pegs Portland at 38 wins, which is a -3 game difference out of 82, and which therefore has a P-value that isn't statistically significant if the standard of .05 is followed. So basically, we are arguing about nothing, and Hollinger again looks a bit foolish.

Over single season samples, we are always talking about too little data to make concrete conclusions for whole teams...but since teams change too much over multiple seasons and we can't run the same season over and over, we're forced to go by that if we want any statistical analysis at all.

If you feel any team-level projections are useless due to lack of sample size, that's a reasonable position. But, in that case, I don't see a reason to blast Hollinger. For those who are interested in it, he's doing what he feels is the best possible with the data available. I don't think that approach is flawed or silly.

 

[rocketeer]

Why? I explained why I didn't think so...that he was allowing that the possibility exists, due to other informed opinions. Why is it stupid to admit the possibility, due to what scouts think?

because it doesn't belong anywhere in hollinger's analysis. to say "oh i think oden will be this but i've heard scouts say he could be this so i'm going to throw that possibility out there even though i'm actually going to disregard it in my projections" is stupid.

 

[Minstrel]

So, what is the value of "expected preformance" if luck has some "value"? So in his "expected preformance" is lower than mine ----??????????

"Expected performance" can be generally done a couple of ways:

-the individual performances the players of the last X seasons (different models use different amount of the past). Based on how different types of production (scoring, rebounding, scoring efficiency, etc) correlate with wins, the types of production are given weights. That creates the statistical model for placing value on players. Such models either put the values in "wins" or have translations for turning the values into "wins" (the expected amount of minutes each player plays are a factor here). Sum up the wins of all the players, and you have an expectation of wins for the team. This is predictive, you do this before the season happens.

-How the team level numbers turned into wins. There is a strong historical correlation between the amount of points a team scores/allows and the number of wins they record. That correlation has been turned into an equation. So if the amount of wins and losses predicted by this equation (using the points the team actually scored and allowed over the course of the season) is greater or less than the number of wins the team actually got, you have evidence (not proof) of luck. Good luck if the team won more games than expected, bad luck if the team won less games than expected. These differences are usually ascribed to good or bad luck in close games. This is reactive, you do it after the season is over, looking back.

Luck affects both up and down seems to me?

Over a large enough sample. Over just one season, luck doesn't always even out.

 

[Minstrel]

because it doesn't belong anywhere in hollinger's analysis. to say "oh i think oden will be this but i've heard scouts say he could be this so i'm going to throw that possibility out there even though i'm actually going to disregard it in my projections" is stupid.

That's not part of his analysis. That's an aside. His analysis is that it won't happen. He's simply saying it's not impossible...call it a piece of "FYI" trivia.

 

[gatorpops]

"Expected performance" can be generally done a couple of ways:

-the individual performances the players of the last X seasons (different models use different amount of the past). Based on how different types of production (scoring, rebounding, scoring efficiency, etc) correlate with wins, the types of production are given weights. That creates the statistical model for placing value on players. Such models either put the values in "wins" or have translations for turning the values into "wins" (the expected amount of minutes each player plays are a factor here). Sum up the wins of all the players, and you have an expectation of wins for the team. This is predictive, you do this before the season happens.

-How the team level numbers turned into wins. There is a strong historical correlation between the amount of points a team scores/allows and the number of wins they record. That correlation has been turned into an equation. So if the amount of wins and losses predicted by this equation (using the points the team actually scored and allowed over the course of the season) is greater or less than the number of wins the team actually got, you have evidence (not proof) of luck. Good luck if the team won more games than expected, bad luck if the team won less games than expected. These differences are usually ascribed to good or bad luck in close games. This is reactive, you do it after the season is over, looking back.



Over a large enough sample. Over just one season, luck doesn't always even out.

I am not even sure there is "luck" in a game. Was it luck that Outlaw put in that last shot at Menphis (?) I don't think it was. It was the result of coaching, practice, decision, and many other factors.

I can see that his formula for "expected performance" is valid but just ain't so that luck can have some kind of value. Not in my mind. It is the result of the above factors and many others. Expectations were too low, because of poor data.

g

 

[Ed O]

I am not even sure there is "luck" in a game. Was it luck that Outlaw put in that last shot at Menphis (?) I don't think it was. It was the result of coaching, practice, decision, and many other factors.

Sure there's luck. The things that you list can influence percentages but can't influence luck.

He can increase his rate of success through hard work and other factors, but he can NEVER get it to 100%. Luck influences whether a shot rattles in or out, and in close games (such as Outlaw's gamewinner in Memphis) luck is particularly important.

Ed O.

 

[gatorpops]

Sure there's luck. The things that you list can influence percentages but can't influence luck.

He can increase his rate of success through hard work and other factors, but he can NEVER get it to 100%. Luck influences whether a shot rattles in or out[/B], and in close games (such as Outlaw's gamewinner in Memphis) luck is particularly important.

Ed O.

Don't see it ED. That shot goes in because of the physics. Right spin on the ball, hitting just the right spot, whatever.

In my mind, Outlaw put himself in the position and the defense did not stop him from putting it in the hole. We call it luck, but it is just doing and being in the right place at the right time.

 

[PapaG]

Bump