What Did Hollinger Say?
by Blazer1342 on Aug 13, 2009 3:18 AM PDT
Email
PrintIn light of Ben’s post where Hollinger is quoted,
Nobody talks about the Blazers as serious contenders in the West, but we should. Portland had the best scoring margin in basketball after the All-Star break last season, and as I keep repeating, scoring margin predicts future success better than winning percentage.
http://www.blazersedge.com/2009/8/12/986756/clyde-drexler-blazers-are-1#comments
I’d like to take a look at what that statement signifies.
The only research I’ve ever done which is basketball related, is the direct correlation of scoring margin (Point Differential) to win %. That relationship could not possibly be more of a direct one to one relationship.
e.g.
14 years of NBA data (all teams included) show that a team with a point difference of "0" (i.e. scoring margin) crosses the line at approximately 40.95 wins. Frankly, I personally don’t see how you could get any closer to a direct correlation unless this data resulted in exactly 41 wins (a .500 season),to which it almost did.
Here’s a chart of those 14 years. Number of wins at the top and point differential at the left.
If I have evaluated the data correctly the largest deviation from this straight line at any win% level (listed as number of games won in this chart) is +2.9 points to the high side and –2.6 to the low side. This is an extremely significant number, in 14 years not one team has deviated more than 3 points from this median. If you didn’t know a teams scoring margin, by choosing the median if given the number of wins you could tell within 3 points what their scoring margin was.
Analyzing Hollinger
In my mind that declaration (quote above from Hollinger) can only be interpreted one way.
What Hollinger’s statement suggests (even though scoring margin is a direct correlation to win%) is that scoring margin leads win% for a team on the rise, or stated in other words, win% lags scoring margin. So a team who continually does better in scoring margin than what they actually achieve in win%, should be a team on the rise.
It therefore also means that this must true for a team that is in decline. So a team declining from prominence will have a better win% than what their point diff would suggest.
To restate this, hopefully a little better, If scoring margin is a better indicator of future success, then a team on the way down should have a point differential which is a little worse than their record (win%).
So if Hollinger be correct, point differential leads the way, whether that team is going up or going down.
Scoring margin is a little better than win% if going up, a little worse if going down.
Is this true, I don’t know? But before San Antonio upgraded most would agree they were a team in decline and they had point diff of 3.8 last year which should have resulted in about 49-50 wins and they had 54 wins. So in SA’s case, scoring margin was leading the way down. So with a sample size of 2, Blazers and San Antonio, it is true.
Update: I hate when I have a moment of clarity long after posting.
Bottom line is this: 54 wins means absolutely nothing. going from 41 to 54 means nothing. It does not tell you anything about how many wins will be added next year. You can’t add 13 to 54 and get 67. The more pertinent question is, did I have a point difference of 3.0 or 9.0 when I got those 54 wins. Take two teams with 54 wins, one with each of the afore mentioned scoring margins, and I’ll tell you who’s going to beat who in the near term. (this was my take on Hollinger when putting up the post, I hope it came thru)
What does this mean for the Blazers next year?
Here is a table of the Season, which lists how many games we won for the season, and how many were won for the last 41 & 21 games of the year. It also is a projection of how many games the Blazers would have won if the last 41 & 21 games were extrapolated out to 82 games. There is a projection based on win% so it can be compared to how many games we should have won if based on scoring margin.
The very bottom line is a projection of how many games we should have won if based on scoring margin.
This table shows that scoring margin is leading win%.
I removed 54 games from the table because it was not a projection. Our point diff of 5.34 suggested a 56 win season, 2 games better than the 54 we won. Notice the last 2 lines where this spread widens as the season continues.
In the last 41 games our actual win% would yield 58 games but point diff suggested 4 to 5 games higher at 62-63 wins.
And then the last 21 games win% = 62-63 (62.48), but point diff proposes 68-69 (68.8), spread widens to 6+.
So scoring margin kept leading by a larger spread. See last paragraph where the last 21 games was not a weaker schedule than the last 41 games.
---------------------------------------------------------------------------------------------------------------------------
The same information can be interpreted from the chart below.
The Blazers had a point diff of 9.8 for the last 21 games, which on the chart below shows a 68 win season actually about 68.8. Compared to a win % which would have yielded about 62 wins. All things being equal a 62 win season should have around a 7.4 scoring margin. We were 2.4 points above this level for those last 21 games.
Their last 41 games had a diff of 7.7 which yields about 62 wins (this is eyeballed from chart, table shows 62.48).
Compared to 58 wins if based on a .707 win%. Likewise a 58 win season should yield about a 6.0 margin, but we had 7.7 which is 1.7 better for the last 41 games.
So point differential is leading win% in both of the later segments.
The point diff for the first 41 games was 2.91, which would be in terms of games won, a 49 win season. So the Blazers scoring margin went from lagging win% in the start of the year to leading win% by the end of the year.
(Please note: the way to view this chart is to come across horizontally from any point diff level, when that line hits the median line then draw a vertical line upwards to see projected wins. Since this chart has wins at the top instead of win%, I included in the table above the number of games won based on the win% - see projected games won based on win%)
One way to look at the chart above and probably the best way, is to note that a point differential above the median line is leading win% rather than lagging it. Which is how the Blazers ended the season. Scoring margin leading win%.
Only 3 teams in the last 15 years had a scoring margin above 9.8 and all 3 were NBA Champs (Blazers were at 9.8 for last 21 games). (charts above were based on 14 years due to a shortened season)
Conclusion:
If Hollinger is anywhere close to being accurate 60+ wins is close to being in the bag. The only caveat that I did not discuss in this post is how far wins can deviate from point diff. Looking at the chart this appears to have a range of about 20 games wide across the median. i.e. +/- 10 games. But in reality the widest range I could find was about 14 games at the widest point (using increments of .1 for point diff), which yields a +/- of about 7 games. What this means is we could be above or below by 7 games and still be within normal. Taking the worst possible scenario still yields a win of 61 games if the last 21 games are chosen to conjecture upon. If the last 41 games is chosen a worst case senario is somewhere near 56 games.
But then again in this worst-case scenario, win% would be lagging point differential by the widest of margins. So if it were possible to take the end of last year and project it onto this year, 56 wins is our floor. Any way you look at it, things are looking good.
One Final note:
In the table above, if point differential were to lead win% by the same margin next year as in the last 21 to 41 games of this past year. And if in next year all teams were the same. Then the table would be fairly accurate with a range of 58 to 62 wins. The win% of the 1st 41 games of our opponents was .507, the win% of opponents in the last 41 games was .480, the last 21 games .496
How much 4 rookies with a year under their belt, Andre Miller, a healthy Oden & Martell, and a slashing Nic counteracts a weaker 2nd half schedule is anybodies guess.
We’ll have to play the games to find out.
111 comments
|
30 recs |
Do you like this story?
Comments
That's interesting stuff.
I somewhat skimmed it seeing as its really late and I wouldn’t be able to take in so much info at once, but it looks promising. The guys played out of their minds at the end of the year (until the playoffs) and it’s going to be interesting seeing how much of that carries over. I honestly think we can be a 60 win team at least at this seems to support that. There’s much more to winning and losing than point differential though.
I can’t wait for this season to start
Does Nate's coaching style with these players add to point differential more than it does win totals?
Popovich is all about positioning for the playoffs. If his team was down big, like against us last year, he will simply rest his starters. That leads to worse point differential rather than if you fight for every point .
Nate is the opposite. He doesn’t dog pile but he asks his guys to fight back until the last minute at least even when we are down. One reason for this is to get young guys to learn to play the right way. We get a few more wins, like when CP3 went down and we clubbed what were basically a bunch of baby seals in New Orleans. (I include NO’s senseless announcing corps in this.) But basically, we are closing the gap hard in our losses. Like we did coming back in Boston and they had to send their starters back in. We didn’t win, but we made is close. That counts in differential, but not in win totals.
What you’ve said is certainly true, and can account for a point diff that leads win%
But how much does 4 rookies playing 50 games, losing focus for long spells, losing games we should have won, take away from win%. How much did an apathetic showing in most of our 1st quarters take a toll on win%
The exact importance of any one factor is probably beyond statistics to unfold. Hollinger states that scoring margin is a leading indicator, which is the current state of the Blazers position.
But at the same time I’m certain that "scoring margin being a better indicator of future results" is not by any means absolute.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
I am not saying our team is bad -- we're good and were getting better!
I am saying that the way Nate coaches could account for excess point differential that does not convert to wins. Factoring coaching style into this algorithm might help it be even more predictive.
A counterargument to this potential trend is that the Blazers have fewer net possessions. So it would be more difficult to gain a higher differential. Differential/possessions would flatten this (although more possessions per game should slightly reduce expected variance with more instances over which to average out per contest). How close to leading the league would the Blazers be in this new category?
However, estimating win totals using differential/possessions would likely make Portland’s win numbers at even wider variance from the model’s projections. Which leads me back to my first point.
Really, it (your comment) all sounded good to me
in the main post using 2 teams was more a tongue in cheek comment to a question I didn’t know the answer to. SA would definitely be a bad example – not sure many coaches take that approach to that extreme.
I’m trying to put together something that validates Hollingers assumption, but I think having the last 41 games of each season would be ideal, and I don’t have it. The blazers had a 2.9 first half and a 7.7 second half point diff. I don’t think I’m will be successful without that kind of data. One thing for sure that I’ve found, is all teams regress to the mean which goes against Hollinger (which you’d expect otherwise point diff wouldn’t be a tight one to one ratio). So a team that has a nice pop like the Blazers had is vital info.
http://farm4.static.flickr.com/3270/3819453559_0f494c3964.jpg
point diff on top, wins increase/decrease for the next year on left
This is 11 years, had to throw out two years because of the 99 season and 2009 doesn’t have a following year. This is all data so it doesn’t look at a "rising team".
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
this is a chart that compares one year to the next. how many wins does a team gain in the second year in relation to it’s point differential from the first year. you’ll notice teams with positive point diff lost wins, team with a negative P-Diff gained wins.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Did a team with neg 3.5 ish point diff actually win 42 more games the next year?
And did a team with pos 6.5 ish point diff actually lose 39 more games the next year?
Seems very improbable
"Tough times don't last. Tough people do."
-Chauncey Billups
bulls went from .756 to .260 the next year
can’t use wins, the .260 year was the short season, just showing big changes happen. (equivalent to going from 62 wins to 21 if both years 82 games)
Boston went from 24 wins to 66 wins, 07 to 08.
denver 37 to 65 wins, 74 to 75
detroit 59 to 39, 07 to 08
GSW 50 to 26
Miami 44 to 15 back to 43
Spurs from 59 wins to 20 to 56, year 96 to 97 to 98 (they had another year 21 to 56 wins)
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
that Spurs team
was when the Admiral got hurt, they fell into the lottery and got lucky (Duncan) then went on to win 4 trophies
the ’74 Nuggets (ABA) had Steve “Snapper” Jones, while the ’75 squad added Bobby Jones and Mack Calvin. BTW, The ’85 Nuggets improved 14 games (38 to 52) following the imfamous Vandeweghe deal)
the ’79-80 Celtics went from 29 to 62 wins, thanks to a rookie named Larry
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
Here’s to hoping that Oden brings us the same kind of luck.
Not that ours was due to the same reason, injury, but we certainly didn’t deserve that #1 pick.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
sure we deserved it
it balanced out getting #4 pick the previous year. Even though we had the worst record in the league, we got bumped out of all the top 3 slots and ended up the worst we could have been.
Interesting analysis
although I’d be interested to see what the regression looks like with some of the outliers taken out (you might have taken them out of the regression, but I can’t tell). Most of those, like the Bulls and Spurs as you described, are due to the teams the following year having vastly different personnel. Both of those teams had 2 HOF players in their “good” years vs. none in their “bad” years. I’m not sure how much insight those circumstances would give us into predicting the performance of teams whose cores largely stay together.
I’m just guessing here, but it seems that most of the huge loss diffs (20 wins or greater) on good teams are probably due to injuries to all star level players. I’d guess that something similar is the case for the huge win gains as well (all of the 30+ win gains are due to adding an MVP-level player: Nash, Duncan, Garnett, plus the Heat with Wade causing a huge drop and then another huge gain).
I’m not saying that injuries or additions need to be completely controlled for, which is probably impossible, but I think the most extreme effects could be filtered out to at least mitigate the effects of these extenuating circumstances.
No outliers were removed.
Visually it looks like they may have pretty much cancelled each other.
Another thing that looks interesting to me is that it appears as you move left to right the data points get denser. This would suggest the good teams have less change from one year to the next. Which makes sense, you’d expect the bad teams who get the good draft picks would shift faster than a team good team trying to maintain what they have.
I probably should have put a polynomial trend line on that chart which would show a curve steeper on the left which lessens as you move right.
While regression to the mean is a know entity (as polisam points out in another post), I would take a guess that doesn’t happen in the time frame of one year to the next.
Since this is a graph of one year to the next. you’d expect the points getting tighter as you move right across the graph.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
I should also say
I have been able to prove that comparing teams at the same win level, the teams with a lower point diff drop off the following year compared to the teams with the better point diff.
Which in a sense supports Hollinger.
I used teams that had 41 to 55 wins for the first year.
I don’t think you can realistically go above 55 wins and say a team is still on the rise. Even 55 is a little high. The problem I’m running into is the lesser teams get worse, and the better teams flat line. It seems eliminating teams with an extreme high point diff might be an approach. I think one could argue that to have a very high scoring margin meant you were good all year and not improving during the year.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
One aspect that will effect net win +/- is the haves and have nots
Some teams are investing and re-signing and others are divesting in the 2009 free agent market. (Whether this is due to the 2010 free agent sweepstakes does not matter in this analysis.)
In 2009-10, the teams with a strong developing core and who’ve kept their strength together and invested in top free agents should see a lot more easy wins against the bulk of the teams in the league. At the same time, wins against the top teams will be harder to get.
The Blazers tend to beat the teams they should beat, already. So, with little expansion in this segment, Portland’s +/- win total will really be determined by how the Blazers fare against the tougher teams. Now it’s time to take it that next level and be ready for skill and intensity we’ll encounter in the play-offs.
Makes it difficult, it’s a living organism.
Definitely not like trying to measure the bounce of ball dropped from 10 feet. Probably more like trying to measure the bounce, but the experiment is done while a hurricane is blowing.
It’d be nice to know what kind of models KP and others actually look at.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
I think it all evens out with blowouts.
If you remember, when Pop was resting his starters because we were blowing them out, Nate was resting our starters for the same reason. The game was in the bag, so why risk injury?
In fact, the last two weeks of the season it seemed that we had a huge string of 20-30 point wins. The starters never played past the middle of the third quarter. Am I remembering this right?
You're misremembering
at least about the starters playing. That only happened in the last two games of the season, a home blowout over OKC, and us blowing out Denver in a game that was meaningless to them. Before that, Roy and LaMarcus were playing huge minutes in every game. In April, which roughly corresponds to the last two weeks and encompasses our last 8 games, LaMarcus never played fewer than 39 minutes (38:55, technically) except for the two games mentioned, and Roy played under 33 only once (an away win against OKC).
IIRC, it was a major point of discussion here about how unwilling Nate was to rest either of them until the last three minutes or so.
As to the resting starters thing evening out, that’s kind of the point. The difference between a 20 point win and a 30 point win often has nothing to do with any player who matters, but just comes from one team’s scrubs going on a run against another team’s scrubs. Those guys have no effect on actual games, yet raw point differential treats it as if they did. An NBA team could theoretically have an excellent 9 man rotation, and just pull a fan out of the stands and put him in if they were getting blown out, and the win might become twice as bad, but since they’re never going to actually put that fan into a game when it’s on the line, how much should those results really factor in?
As an example, we’re blowing out the Lakers by 26 at the start of the 4th quarter. Kobe stays in for a while, Farmar hits some shots, and the final winning margin ends up as something like 20. Compare that to the Bulls game where we go into the 4th by a similar amount, but our bench stays hot and we end up winning by 40. Team strengths aside, is that 40 point win really twice as impressive as the 20 point win?
I read this earlier tonight and thought I'd share it, here
Let’s start with Brandon Roy, since he’s the leader of this team. Last season he averaged 37.2 mpg, but that number was dragged down in April, when he played just 32.4 mpg as Portland rested him for the postseason. So he’s basically a 38+ minute player.
http://www.rotoevil.com/nba/2009-2010-team-preview-portland-trail-blazers#more-2016
Are you sure you didn’t “misremember” Roy’s PT from last April? Or was the writer of this article looking at the wrong player’s stat line?
The thing that concerns me is that even though the Blazers tried to rest Roy heading into the Rocket’s series, he was admittedly “so tired” during game 6. Some of this had to do with the defense of Artest and Battier, but also Brandon had to carry the load on the offensive end, because Blake, Rudy and Outlaw weren’t getting it done (in game 6, at least)
Wheels said that the Blazers needed to add a reliable “3rd scorer” following that series, and hopefully Miller is the man, or perhaps Andre can help find other “sources” of points that don’t involve Roy working so hard and getting so tired
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
Look at the actual box scores
When there are only 8 games in April, a couple games (the last two, like I mentioned) can have a big effect.
Starting from the end of the season:
DEN: 22 minutes
OKC: 22 minutes
LAC: 39 minutes
LAL: 37 minutes
SAS: 38 minutes
MEM: 38 minutes
HOU: 33 minutes
OKC: 26 minutes
Like I said, in the last two games, they rested Brandon, but before that, he was playing a ton of minutes, above his season average, actually. Certainly, he was playing significantly past the midway point of the third quarter in most of those games, so unless you interpret the ’huge string of 20-30 point wins" to mean only two games, Brandon was still playing a ton of minutes.
Indeed, if you expand to look at the last month, instead he never played fewer than 32 minutes in the preceding couple weeks, and had a stretch of three straight games where he played 46, 40, 46.
That's just no good
Hopefully we can focus on preparing more for the playoffs this year, and spend some time on combinations that don’t involve Brandon and LMA so heavily. These guys don’t need to be averaging more than 35 mpg in the regular season.
Sure, they might play more in some games here and there, and they might play more in the playoffs, but with as much depth and talent as we have, Nate ought to be able to find some combinations that can be effective enough to give these guys some rest.
When I rule the world, everyone will know how to use Excel.
Very nice post
but you know already what I’m going to say.
End of season point differential vs. early season point differential should be adjusted by strength of schedule. If you do that, you’ll still see more strength end of season, because we got better especially as we got Steve and Greg back. But it won’t be quite as pronounced.
It would be interesting to see if win percentage really does lag point differential. I’m not sure that is a good assumption. Our win percentage in 07-08 exceeded our point differential. Last year, it was reversed. Clearly, in 07-08 we were a team on the rise.
When I rule the world, everyone will know how to use Excel.
You’re too fast
I was going to add a final note about the strength of schedule at the end of the year, but you beat me to it.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
I think we had a tougher first-half schedule.
So it would make sense that our point differential increased second half.
"These are dreams that we have." --Rudolfo Fernandez
.506 & .480 respectively. The 1st 41 games compared to the last 41 games.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Summar
There’s evidence to suggest blazers will kick butt next year and over 60 wins is likely but worst case will be 56 and this is all based on the blazers totally demolishing the teams at the end of last season which some believe to be the sign of a team on the rise. (the demolishing/point difference).
fine point being that major injuries don’t derail the team, of course (my fine point).
"Fernandez, to my eyes, is the Blazer who walks that walk most comfortably. A lot of Portland's fans (egged on, dare I say, by their local broadcasters) lament things like how Ron Artest or Yao Ming get to hit Brandon Roy's arms.
But I suspect Fernandez sees all that and thinks: We get to hit arms! Cool!"
http://myespn.go.com/blogs/truehoop/0-39-135/On-Playoff-Experience.html
Nice job supporting my argument that fans who bemoan ...
… the Blazer’s lack of a big trade deadline or off season move, or keep saying the team won’t progress until they “fix” the glaring holes in the lineup are simply advertizing their ignorance.
One can do all the stat sheet comparisons between Blazer players and their counterparts around the league and come up with fantasy league trade ideas and rosters that they want. But the important stats to look at are not those of individuals but of the team as a whole. And I for one like what I see and have every reason to believe that they will get better.
hakkaa päälle !
by timg56 on Aug 13, 2009 7:51 AM PDT reply actions 2 recs
Agreed
although i do like the Miller addition. There are signs that indicate to me that he’ll be beneficial and it’s not like bayless would have been nailing those threes either.
"Fernandez, to my eyes, is the Blazer who walks that walk most comfortably. A lot of Portland's fans (egged on, dare I say, by their local broadcasters) lament things like how Ron Artest or Yao Ming get to hit Brandon Roy's arms.
But I suspect Fernandez sees all that and thinks: We get to hit arms! Cool!"
http://myespn.go.com/blogs/truehoop/0-39-135/On-Playoff-Experience.html
Exactly why I love the point differential stat.
It’s not subject to player rotations and a host of other issues, but a pure stat IMO, and a very good indicator of overall team performance. A stat which actually has meaning and is quantifiable.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Nice post
I think you might be reading a little too much into what Hollinger is saying. Hollinger’s comment has nothing to do with a team being “on the rise” or “on the decline” – at least not to the extent you’re factoring in age and using the Blazers and Spurs as examples.
The Spurs did not have a higher win percentage than their point differential would suggest because they are on the decline – it was luck. Point differential is more indicative of team strength and likelihood of success going forward, and win percentage has more “noise” due to near wins and near losses.
Basically, a team that wins 5 games by 30 points and loses 5 games by 1 point is much better than a team that wins 5 by 1 and loses 5 by 30, right? That’s all Hollinger is saying. Either team could be on the rise or on the decline.
Even though the Blazers are on the rise, their win percentage last year could have just as easily been higher than their point differential would suggest, and the Spurs win percentage could have been (as it was in years past) worse than their point differential. The fact that one team is improving and the other is declining is irrelevant. All Hollinger is saying is that the Blazers ARE better… not that they’re getting better.
Q: Is Greg favoring his knee?
Frye: He favors dunking on your head, that's what he favors.
Or to be precise
The Blazers WERE better at the end of last year.
Which indicates that they were improving over the course of the season, which could well indicate that they are getting better.
But I think you are right about reading too much into Hollinger’s statement.
When I rule the world, everyone will know how to use Excel.
Right...
Now, the question is how much better does margin of victory over the last 41 games predict wins than previous season wins? And what would that predict for the Blazers? That could also be tested, but it would take a bit more work since the data is not formatted to show ppg over the last 41 games.
that was my question too
"Fernandez, to my eyes, is the Blazer who walks that walk most comfortably. A lot of Portland's fans (egged on, dare I say, by their local broadcasters) lament things like how Ron Artest or Yao Ming get to hit Brandon Roy's arms.
But I suspect Fernandez sees all that and thinks: We get to hit arms! Cool!"
http://myespn.go.com/blogs/truehoop/0-39-135/On-Playoff-Experience.html
this is exactly what Hollinger was saying
He wrote a whole article on the subject, I will try and find it really quick.
Ben II Blazersedge.com || New to Blazers' Edge?
here are a few he has wrote
Mr. Berri agree’s with Hollinger
The money quote from Hollinger,
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, the rankings uses points scored and points allowed, which are actually 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.
That explains why, for instance, Phoenix is No. 1 right now even though Dallas has a better record — the Suns have the best scoring margin in basketball.
Conversely, it explains why Miami is No. 24 even though the Heat are close to .500.
Ben II Blazersedge.com || New to Blazers' Edge?
thanks for the links
I hate when I have a moment of clarity long after posting.
Bottom line is this: 54 wins means absolutely nothing. going from 41 to 54 means nothing. It does not tell you anything about how many wins will be added next year. You can’t add 13 to 54 and get 67. What does tell you something, is did I have a point difference of 3.0 or 9.0 when I got those 54 wins. Take two teams with 54 wins, one with each of the afore mentioned scoring margins, and I’ll tell you who’s going to beat who in the near term. (this was my take on Hollinger when putting up the post, I hope it came thru)
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Another thing I think is true about the Spurs,
is that they are a veteran team and likely pull out some wins that a young team might not come away with. Spurs might not have been the best example to choose from.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Some good stuff
Here’s how I would interpret what Hollinger said:
Past points per game differential is a better predictor of future wins than past win percentage.
In its purest form, the best way to test this is to look at data within a season. One would just compare the goodness of fit of a model with win pct and points per game differential as the key predictors (with identical control variables for strength of schedule). That is easy enough to do, but you seem to be primarily predicting wins in one season based on the previous season. Again, the test for Hollinger’s claim is pretty straight forward: what is a better predictor of future wins? Previous wins or previous margin of victory/
I ran the data on NBA season data from 1980 to 2007 and previous margin of victory is indeed the winner. The r-squared for a model with just lagged wins (or previous season wins) predicting wins is .45. The r-squared for a model with just lagged margin of victory is .47. Not a huge difference, but a difference. How much of a difference is that, well the wins only prediction has a standard deviation of 9.3 while the points per game prediction has a standard deviation of 9. So, roughly speaking, in making predictions between season, points per game lagged is only “better” by about .3 games, on average.
For what it's worth
If you use previous margin of victory to predict wins in the following season, the predictions for the Blazers are a bit more modest than the ones you cited above. Here are the three scenarios:
5.4 ppg diff → 51wins +/ 9
7.7 ppg diff → 56 wins +/ 9
9.8 ppg diff → 60 wins +/ 9
Why are the predictions so much more modest? Because they are based on between season predictions and there’s a lot more variation and reversion towards the mean in performance between season.
Toss out the outliers
I wonder what happens if you toss out the outliers, like the ’08 Boston squad? That might eliminate teams that changed due to new rosters, rather than organic growth or decay.
If you drop teams that improved or decined by more than 20 wins...
I re-ran the analysis dropping any team that were 20 games above or below their predicted win total and it would bump up all of the predictions for the Blazers by one game.
Of course, there are a lot of other things you could consider in making predictions… it would take some effort, but it’s actually possible to include trades, draft picks, age (even “win share” weighted age) in those prediction. Might be a fun project.
but then you would have to really cherry pick your data
and given the fact that you are talking about one team over a 14 year period in a 28-30team league means that all you are including is 1 data point in ~450 data points, so you have to ask yourself if they really skew things that much when averaged in with the rest of the data points
Life is exhausting when you are this stupid.
I will talk about DeJuan Blair no more forever
thanks
I was just thinking about a team like the Spurs when Robinson went down and how that would even out an anomaly like the Celtics without having to go through the wishy washiness of cherry picking data.
Life is exhausting when you are this stupid.
I will talk about DeJuan Blair no more forever
Nice post
Scoring margin over an entire season does seem like it would be a great predictor. But somehow I cant quite wrap my head around:
If I have evaluated the data correctly the largest deviation from this straight line at any win% level (listed as number of games won in this chart) is 2.9 points to the high side and –2.6 to the low side. This is an extremely significant number, in 14 years not one team has deviated more than 3 points from this median.Surely the scoring margin when Boston won only around ~20 games did nothing to predict the 60 win season that they had the next year after the big trades. Same thing with San Antonio before and after drafting Tim Duncan.
Even some of the data in your post would indicate differently
But before San Antonio upgraded most would agree they were a team in decline and they had point diff of 3.8 last year which should have resulted in about 49-50 wins and they had 54 wins.San Antonio had 4 wins more than predicted which is outside the +2.9 to -2.6 range.
Like the effort. Would be eager to hear more on the subject.
Referring to the top quote
You read the numbers 2.9 and 2.6 in terms of the error in the number of wins predicted, given margin of victory. The numbers he gave-whether knowingly or not—is the largest error in the predicted margin of victory, given win total. That explains the San Antonio case that you gave.
Also, note that those numbers are prediction errors within a given season, which is why there are no huge outliers, like Boston improving by 20 games.
Huh?
Isnt the San Antonio error of +4 games outside the range of +2.9? Am I misunderstanding something?
Thought that the whole point about this article is trying to use previous years margin of victory to predict the next years results. If the prediction errors are only within a season, do they have any bearing on next years performance. Can you really even use margin of victory as a predictor from one year to the next?
Isnt the San Antonio error of +4 games outside the range of +2.9? Am I misunderstanding something?
The way you interpreted what he said makes sense. I think he misread his graph, but I was getting him the benefit of the doubt. Basically, I am saying you were right in pointing out a mistake that he said… And, I was also explaining why he made the mistake.
If the prediction errors are only within a season, do they have any bearing on next years performance
Agree. Probably not.
Can you really even use margin of victory as a predictor from one year to the next?
Well, you can use anything you like as a predictor; it’s just a question of how good it is. As I said above, it turns out that margin of victory in the previous season predicts wins in the next season better than wins in the previous season. I would characterize both as relatively rough predictors. Are they better than “expert” subjective projections? I don’t know. Probably not, but possibly. Experts are notoriously poor at making predictions.
just to be clear
The errors in predicting wins from margin of victory should be read from the left and right of the regression line in the graph above. I think he read the errors in predicting wins from margin of victory from the vertical distance from the regression line… but that is the error in predicting margin of victory, based on win total.
Adding units to the numbers clarifies the statement
San Antonio was +4 games won based on their +3.8pts/gm differential. The 2.9pts/game differential is the range of deviation from the expected norm. So no San Antonio was not outside the range. My concern is that the wrong error bars are being discussed (the plus mins 3pts/ game differential). Given a pts/game differential, you should care about the error in the number of games won error. This was stated to be plus or minus seven games. Therefore San Antonios +4 game win margin was well wihtin typical norms. This error also shows how impresice the pts differential is. The alternative conclusion is that given a teams win loss record you can predict a teams differential margin with plus or minus 3 pts/game. This is also true given the stastical data but a detail which nobody really cares about.
right... but to be clear, 2.9 pts/game is the largest outlier... not the normal deviation
Same goes for the +/- 7 games for wins. That’s the largest error in 14 season. The average error is about a game or so. So, I would not call the ppg predictor imprecise (at least within a season).
AAAh so. Thanks.
your (and polisam’s) patience is much appreciated.
Poli has addressed your question, but if I can state it a little differently
point diff is across the y-axis, -2.6 to +2.9, number of wins is across the x-axis and is about 14 games wide at the widest point at any given level. If you were to draw a band (similar to bollinger bands in stocks) or two straight lines around the range ( and you’d want these line to be parallel) it looks to be about a 20 game range, which might be a better number. I did not crunch any numbers, just eyeballing. Bollinger bands is probably a bad example, just thought if you were familiar with them you’d know what direction I was going with it.
And yes, a complete roster overhaul blows any of this kind of data out of the water.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
I'm very optimistic right now.
Please, Basketball Gods, PLEASE let this team stay healthy!!!!!
But I like to be here. Oh, I like it a lot! Said the Cat in the Hat. To the fish in the pot.
by you'vegottomakeyourfreethrows on Aug 13, 2009 9:42 AM PDT reply actions 1 recs
Great post...I like this stuff.
I also analyzed win / loss data for the past 32 years. I trended it through several algorithms, smoothed the data with an inverse square hyperpolynomial function and projected it using geometric base mean Vorscheckian gobble rule. I then ran it through the dryer on the “fluff” setting before mixing on medium speed for two minutes. The results were shocking, to say the least. Every NBA champion for the last 32 years won the last game they played that year. This trend held without fail.
I suggest tanking until game 7 of the finals.
The cowards never started
The weak died along the way
Only the strong survived
They were the Trailblazers
by lukeyhere on Aug 13, 2009 10:32 AM PDT reply actions 10 recs
I shouldn't have rec'd you
without seeing the graphs to prove your point.
But I did.
When I rule the world, everyone will know how to use Excel.
It's just that people keep giving me so many of them
and I’m not into hoarding them.
When I rule the world, everyone will know how to use Excel.
by jscot on Aug 13, 2009 12:49 PM PDT up reply actions 1 recs
Great discussion post
And lots of nice data and charts to look at.
Well done.
I know less than half of you half as well as I should like, and I like less than half of you half as well as you deserve.
Another thing we learned, last April
is that “point differential in the last 21 games” means very little when you’re facing a team with a superior playoff roster, even if you have the homecourt advantage
KP still needs to strengthen his frontcourt depth with a veteran PF/C, before next spring
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
by two4larue on Aug 13, 2009 10:39 AM PDT reply actions 1 recs
And I hope we learned that lesson well.
Without that first game blowout, we might have had a chance. Roy said he was extremely tired for that game. It would bother me a bit if you could rest a little the last few games of the year, then do it, but doesn’t seem likely in the West.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
by Blazer1342 on Aug 13, 2009 6:04 PM PDT up reply actions 1 recs
That should say "would not bother me..."
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
rec! i love you guys...
you take one sentence and create a full blown statistical discussion.
Go Blazers!!!!!!!!!! Wooooooooooot Wooooooooooooot!!!!!!!!!
It's August
When I rule the world, everyone will know how to use Excel.
wasssup!?!!?!
I started to make a reply in Ben’s post but it just kept growing, seemed like a better idea to make a seperate post.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
To paraphrase Mark Twain,
“[Isn’t it interesting to] get such wholesale returns of conjecture out of such a trifling investment of fact.”
As much as I respect his work, the one thing missing from Hollinger’s hypothesis are actual facts. They may exist, but I haven’t seen the study, and it appears no one else commenting here has, either. PoliSam offers some new data above, but so much is missing from his quickly composed regressions (no disrespect inferred).
Most notably, even assuming the hypothesis is generally true (and I’m not saying it is), over what periods of time is it so? Projecting from one whole or partial season to an immediately succeeding whole or partial season is probably the worst possible relational case, given roster changes, etc.
And more specifically, after controlling for strength of schedule, what past number of games is the best predictor of what immediately succeeding future number of games? (20 past games to predict the next 10? Or 20? Or 80? Or 300? Clearly, the further out you try to predict, the less significant any relationship is likely to be. Perhaps point differential is best at predicting won-loss just a few games in advance based on PD in a much larger number of games in the past. Or perhaps not.
And if we are to trust the outcomes, there had better be a rather consistent relationship of the data across many past time periods to future ones, and not just for a few past-future period combinations where there are positive correlations intermixed with other combinations in which there are negative ones.
But ultimately, what does the data show? It is fruitless even to conjecture on projected Blazer wins based on the Hollinger hypothesis until the data is seen. As Twain said, “Truth is the most valuable thing we have. Let us economize it.”
Add....
Frankly, I wouldn’t be surprised that the best relationship of past point differential to future win probability, after controlling for strength of schedule, home-away, etc., is simply predicting only the very next single game.
by blazerwizard on Aug 13, 2009 12:40 PM PDT up reply actions
I have respect for Hollinger's magic formulas.
He shows how they are constructed and gives simple explanations for his reasoning and admits that they aren’t perfect.
I think he gets bashed for them because his predictions are always there in black and white. There are a thousand other predictions out there on the radio, TV, internet, etc. They are all based on some guys off the cuff gut feel and all of them are forgotten as soon as the page is turned or the segment is over. Hollinger’s stuff stays there to be picked at.
Yeah, it’s not perfect, but it is surprisingly accurate for how simple it is.
The cowards never started
The weak died along the way
Only the strong survived
They were the Trailblazers
Like Hollingers work as well
Its interesting to watch how teams move in his power ranking which are purely formula driven. Dont know if he plans to tweak the formulas further to continuously improve the model.
Efforts like these eventually could produce the Basketball equivalent of SaberMetrics. Suspect that we are still a ways away, especially until NBA tracks defensive stats at sufficient detail, and metrics like assists leading to Free Throws are still not counted. But still Billy Beane’s use of Sabemetrics is said to have kept Oakland competitive with the big market teams ilke Boston and New York; and even shifted focus from BA to OPS, and ERA to WHIP. Look forward to have similar shifts shape the game that we enjoy so much.
RT: Look forward to have similar shifts shape the game that we enjoy so much
My observation: this “stat-geek” over-analysis has shredded otherwise-friendly baseball message boards for the past 10 years
just be careful what you wish for
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
Not familiar with RT
does it stand for Reply To?
or "responding to"
I’ve seen this abbrev used with twitter, so I thought I’d stick in in front of the sentence that I’m “responding to”
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
on twitter it means - re tweet
Senior Asian ambassador of Blazers Edge
by Philthyanimal on Aug 15, 2009 11:41 AM PDT up reply actions
works for me
I could also use “IRT” I suppose
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
Yes, I like Hollinger, too
He’s actually my favorite NBA analyst, and his statistical approaches are most impressive because of their strong theoretical underpinnings.
But … speaking as a statistician, I must tell you that there is no substitute for a rigorous examination of the methodology and the detailed results. I want to see the actual study that leads to his conclusion. When I email him, he does not (probably for obvious reasons) respond.
by blazerwizard on Aug 13, 2009 5:15 PM PDT up reply actions 1 recs
And Mark Twain also said
“There are three kinds of lies: Lies, Damn Lies and Statistics…”
Mark Twain also said
We were astonished at the audacity and sheer magnitude of his lies
by southern oregon on Aug 13, 2009 6:59 PM PDT up reply actions
The secret of life is honesty and fair dealing.
If you can fake that, you’ve got it made, Groucho Marx
I think it’s like predicting the weather, far more accurate in the near term. How near, not sure. It seems the last 41 games would be a good indicator for the next season, if big changes weren’t made with the team.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
To paraphrase W.C. Fields,
“Everything being equal, I’d rather we got Andre Miller out of Philadelphia.”
Did Hollinger also write something about the Pythagorean Formula that he uses in his season previews?
http://armchairgm.wikia.com/Pythagorean_Formula
Well, all in all I take it we are as good or better than our win totals predicted in the last 2 years. Good job.
"I'm addicted to polo y'all...respect my fresh" - Travis25Outlaw
Thanks for the vote of confidence.
Should have probably held back one day and tried to make some of the thoughts a little clearer before posting. I’ve already made countless small changes (one big) which the early posters don’t get the benefit of.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Actually, as I recall, the year before last the Blazers Pythagorean wins were around 37
I think the Blazers point differential suggested they should’ve been a few games under .500, which was part of the reason Hollinger only predicted the Blazers would finish something like 42-40 this year. Hollinger is not the best at predictions from season to season.
But yeah, last year, the Blazers were a little better than their 54 wins suggested. Part of that may have been skewed by extreme depth, though – Sergio and Rudy and Nic and Travis playing some garbage time minutes swelled some of those 20 point blowouts to 35. Not many teams can go 11 or 12 deep like Portland, so my guess is their point differential will continue to be better than their win percentage.
Q: Is Greg favoring his knee?
Frye: He favors dunking on your head, that's what he favors.
What hypotenuse are you trying to make?
We are long at the 5. Our 3 is strong. But, from the right angle, you can see we still need 4 back up.
I think you're trying to say
that the Blazer’s roster isn’t “equilateral” yet
When reached 39 years of following Portland basketball you have, be as passionate of the Trail Blazers you will not!
John Hollinger is going to be on Wheels at Work in the next hour if anyone is interested.
"Aneurysm".
When Outlaw wins a game on a last-second shot, it’s called an "annthefaneurysm". QualityPie
Could you..
pull out past 54-win teams and chart out their next season’s record vs their pt differential?
That would visually give an idea as to whether your idea’s on or not, right.
I would do it but I have no idea how to and I don’t have whatever data in front of me that you do.
Regardless, nice post and good thinking!
Peep the sig.
Good thought, if I have time to get to that I’ll do it.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Should also note that 54 games will be too limiting.
But choosing a range say 45 to 55 games would bring in more data. The question is what range of wins is a good set. Also, another thing I can not do is crunch it with the last 41 games of the season, which I think would be a far better gauge.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Man o man I can not .. W A I T … for the new NBA Season.
by Brad.Peltier on Aug 13, 2009 6:53 PM PDT reply actions 1 recs
Even with the lofty 54 wins of last year
I keep getting this feeling we could still totally blow some people away next year. The potential seems to be there. Is a good strong year like we had enough? I also get the feeling that the Blazers off season work ethic is a disciplined and beneficial as it gets, but that’s purely a feeling.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Remember how we just wrecked the Spurs
We are going to blow up a lot of teams that way this year.
War playing ball in the first quarter.
he was on Wheels after Work today on The Game. here is the link to the podcast (I think he joins about halfway through):
http://vault.desjalogic.com/rcrweb/podcasts/sports/waw8-13-09hour3.mp3
Tony Luftman Is an Animatronic Robot created by Cyberdyne Industries for the purpose of the destruction of the entire human race, no one shall be spared.
Question
I thought everyone was pretty clear on the fact that efficiency differential was vastly superior to W-L records when it came to predicting wins.
Is the news then that Hollinger is touting point differential for some reason instead of efficiency differential?
I’m a little confused.
Don't feel bad, Channing. We can't rebound either.
You mean to tell me
…that a team that scores more points than their opponent frequently also ends up winning frequently?! I think I need to look into this phenomenon. Honestly, it seems a little too good to be true.
:D
"HA HA HA HA HA
I'm not laughing, I'm just listing the five ugliest Blazers ever."
- rockingharder
Strange, isn’t it. but its twue, its twue
Really tho, I think Hollinger’s point was more along the lines of how the Blazers scoring margin ballooned at the end of the year. Hence, it didn’t impact the win column for the current year, but should yield dividends in the next year.
Romance me with that Roy rainbow shot which took flight from way beyond the arc and sailed so high that before it came back down to earth sealing the victory, it kissed the rafters and said "You're mine baby."
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
FanShots
Quick hits of video, photos, quotes, chats, links and lists that you find around the web.
Recommended FanShots
Recent FanShots
In Association With
Ball Don't Lie - NBA - Yahoo! Sport
- Days of NBA Lives: Wherein Roy Hibbert tap-dances to Maine
- Grizzlies Valentine’s Day cards are the key to love
- Kevin Love would like the NBA to take the MLB’s approach on All-Star games
- Kobe Bryant fires a few hilarious salvos at Jeremy Lin
- Brandon Jennings states the obvious regarding his 2014 free agency, everyone goes nuts
Certain photos copyright © 2012 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
