 |
APBRmetrics
The statistical revolution will not be televised.
|
| View previous topic :: View next topic |
| Author |
Message |
Mike G
Joined: 14 Jan 2005
Posts: 3625
Location: Hendersonville, NC
|
 Posted: Fri Feb 18, 2011 11:46 am Post subject: The 2:1 Game |
 |
|
On Feb. 25, 1998, the 32-22 Blazers went to Chicago and beat the Bulls (yes, Those Bulls), 106-101. This was the Bulls' only home loss between Jan. 25 and April 13; and their only non-overtime loss over a stretch of 27 games.
The great victory pushed Portland to 3-2 with their new point guard Damon Stoudamire; acquired (Feb. 13) from Toronto along with Walt Williams and Carlos Rogers, for Kenny Anderson, Gary Trent, Alvin Williams, and picks.
But then the Blazers went to Indy and had their heads (and their entrails) handed to them. It was 33-14 after one, 59-29 at the half, 84-43 after 3, and with a 40-16 4th, finally ending at 124-59.
This is, as well as I can determine the only game since 1986 at least, in which one team has doubled the score on the other. Normally, when a team has doubled its opponent through 3/4, it lightens up; but not this time.
Both teams used 12 players. The Pacers #8-12 guys totaled 64 Min., 16-21 FG, 38 Pts, 14 Reb, 3 TO. Their counterparts went 62 Min. with 3-16 FG, 2-8 FT, 8 pts, 4 Reb, and 8 TO.
http://www.basketball-reference.com/boxscores/199802270IND.html
I suggest that this historic beatdown from wire to wire sets a benchmark for futility (or for dominance) in the NBA. And given that some 30,000 games have been played since b-r.com has boxscores (1986 to present) with just one example of a 2:1 scoring ratio, that this is as bad as it gets.
If semi-Star players like Rasheed, Sabonis, Stoudamire, Brian Grant, and Rider (and their backups) can lose this badly -- and no one we know has ever done worse -- then perhaps 2:1 (for 48 minutes) is the otherwise unreached zero-point of NBA competitiveness.
This is not at all like baseball or football, in which 2:1 (or worse) is perhaps the norm. Perhaps it's zero contribution above replacement level that produces a 2:1 beatdown.
A team that averages 50-100 scores has a Pythagorean expectancy of 0.003, that is wins per 82 games .
For 59-124, it's .002
_________________
`
36% of all statistics are wrong
Last edited by Mike G on Fri Feb 18, 2011 12:05 pm; edited 1 time in total |
|
| Back to top |
|
 |
mystic
Joined: 18 Dec 2006
Posts: 21
|
| Posted: Fri Feb 18, 2011 12:01 pm Post subject: |
|
|
The Lakers were up 109-53 with 1:10 to go in the 4th against the Cavs this season. At the end it was 112-57, missed the 2:1 by either scoring two points of their own or get one points less against.
In the 2009 playoffs the Nuggets came close to that against the Hornets as they beat them 121-63.
_________________
http://bbmetrics.wordpress.com |
|
| Back to top |
|
|
Mike G
Joined: 14 Jan 2005
Posts: 3625
Location: Hendersonville, NC
|
| Posted: Fri Feb 18, 2011 12:12 pm Post subject: |
|
|
Yes, it's almost like an asymptote that cannot be quite reached, but only approached. It's possible an odd turnover/FG combination at the very end allowed this debacle to cross the threshold.
It's also reasonable to suppose the Pacers all just were on top of their game as much as the Blazers were hung over (or whatever their problem was.) Not one Pacer hit fewer than 1/2 his FGA. In fact just one of 12 was below .545 eFG%.
Switch to 'Advanced Box Score' at that link and notice that the starting lineups are exceedingly lopsided. But the reserves look like NBA AllStars vs high school girls.
_________________
`
36% of all statistics are wrong |
|
| Back to top |
|
|
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
|
| Posted: Fri Feb 18, 2011 12:21 pm Post subject: |
|
|
| Quote: | | Perhaps it's zero contribution above replacement level that produces a 2:1 beatdown. |
No, it's mostly luck. Replacement level in the NBA, per player, is somewhere around -3 to -3.5, as has been discussed elsewhere.
In cases like the 2:1 game, it is good players having bad games and probably better players having really good games.
Here are the odds of such a game happening, using the "game performance stdev=12pts/100poss" I estimated in my Bayesian analysis:
Even teams: if the mean is 107 pts/100poss, we're looking for the outlier where with 53 pts/100poss. That's 4.5 standard deviations from the mean. The odds of such a case are 3.4e-6 for even teams.
However, suppose there is a range of matchups in the NBA, where sometimes 1 team is much stronger than the other. Suppose the average matchup is even, but there is a standard deviation on that of about 8 (counting injuries, rest, location, etc.)--I think that's a pretty good approximation. So the updated standard deviation is more like 14.5. Then, doubling a team up is only has the odds of 0.0001123.
That would thus have a 13.8% chance of happening each year. That's probably high, because in general the with a big advantage teams let up (that 12 pt stdev is for all games, and I think the tails on the distribution are thin in reality).
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
| Back to top |
|
|
EvanZ
Joined: 22 Nov 2010
Posts: 307
|
| Posted: Fri Feb 18, 2011 12:28 pm Post subject: |
|
|
| DSMok1 wrote: | So the updated standard deviation is more like 14.5. Then, doubling a team up is only has the odds of 0.0001123.
That would thus have a 13.8% chance of happening each year. That's probably high, because in general the with a big advantage teams let up (that 12 pt stdev is for all games, and I think the tails on the distribution are thin in reality). |
In a decade, the chances of seeing such a game become very high.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir |
|
| Back to top |
|
|
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
|
| Posted: Fri Feb 18, 2011 1:13 pm Post subject: |
|
|
| EvanZ wrote: | | DSMok1 wrote: | So the updated standard deviation is more like 14.5. Then, doubling a team up is only has the odds of 0.0001123.
That would thus have a 13.8% chance of happening each year. That's probably high, because in general the with a big advantage teams let up (that 12 pt stdev is for all games, and I think the tails on the distribution are thin in reality). |
In a decade, the chances of seeing such a game become very high. |
Yeah, except I don't think the distribution is actually normal. I think it's a thin-tailed distribution... We could simply look at the distribution of margins to find out.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
| Back to top |
|
|
Mike G
Joined: 14 Jan 2005
Posts: 3625
Location: Hendersonville, NC
|
| Posted: Fri Feb 18, 2011 2:39 pm Post subject: |
|
|
| DSMok1 wrote: | | Quote: | | Perhaps it's zero contribution above replacement level that produces a 2:1 beatdown. |
No, it's mostly luck. Replacement level in the NBA, per player, is somewhere around -3 to -3.5, as has been discussed elsewhere.
In cases like the 2:1 game, it is good players having bad games and probably better players having really good games.
|
Guessing your -3 to -3.5 refers to per player per 100 possessions?
And so, your replacements would lose to average teams by about 14-17 pts/G ? Let's say they're minus-16.
The Pacers in 1997-98 had an SRS of 6.25; the Blazers were 0.83.
In Indianapolis, add about 3.3 pts to Ind, and suppose they were favored by 9 . They beat that spread by 56 .
Suppose further that this is half due to one team playing its best game of the year, simultaneous with the other playing their worst. Suppose the variance is half due to each team's performance, the Pacers at +28 (from expectation) and the Blazers -28.
Then are you saying that a team of replacements -- somewhat worse than the '93 Mavericks -- could hope to beat this bunch from Portland (on this night) by (-16-(-28)) about 12 points?
This seems plausible. The '93 Mavs, SRS = -14.7, are the worst team ever. They had a couple of guys who were better than replacements.
_________________
`
36% of all statistics are wrong |
|
| Back to top |
|
|
Mike G
Joined: 14 Jan 2005
Posts: 3625
Location: Hendersonville, NC
|
| Posted: Sat Feb 19, 2011 7:38 am Post subject: |
|
|
DSMok1, or EvanZ, or anyone: How likely is it that a team of replacements loses by a 2:1 margin ?
Also, how likely that it's 2.2 to 1 after 3/4 ?
| Quote: | | The Pacers in 1997-98 had an SRS of 6.25; the Blazers were 0.83. |
Other than this 64-point loss, they'd have been 1.62
This one game changed both teams' Pythagorean by ~2.3 ?
_________________
`
36% of all statistics are wrong |
|
| Back to top |
|
|
greyberger
Joined: 27 Sep 2010
Posts: 53
|
| Posted: Sat Feb 19, 2011 12:22 pm Post subject: |
|
|
| Quote: |
The Blazers were 0.83.
Other than this 64-point loss, they'd have been 1.62
This one game changed both teams' Pythagorean by ~2.3 ? |
You're making a strong case for outlier control, I think. Don't we get a clearer picture of this Blazers team if we throw out the biggest blowouts both ways, or at least adjust them down to more typical margins? I'd be tempted to ignore or 'cap' 64-point losses and 45-point wins when working with a season's worth of games. |
|
| Back to top |
|
|
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
|
| Posted: Sat Feb 19, 2011 12:24 pm Post subject: |
|
|
| greyberger wrote: | | Quote: |
The Blazers were 0.83.
Other than this 64-point loss, they'd have been 1.62
This one game changed both teams' Pythagorean by ~2.3 ? |
You're making a strong case for outlier control, I think. Don't we get a clearer picture of this Blazers team if we throw out the biggest blowouts both ways, or at least adjust them down to more typical margins? I'd be tempted to ignore or 'cap' 64-point losses and 45-point wins when working with a season's worth of games. |
I did research on that recently, and I didn't get any better results over a large sample than with simply a normal average.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
| Back to top |
|
|
Mike G
Joined: 14 Jan 2005
Posts: 3625
Location: Hendersonville, NC
|
| Posted: Wed Feb 23, 2011 9:15 am Post subject: |
|
|
| DSMok1 wrote: | ... Replacement level in the NBA, per player, is somewhere around -3 to -3.5, as has been discussed elsewhere.
.. |
I had the impression that -3 to -3.5 (from average) was a 'reasonable guess'.
Has there been a confirmation or a strict definition?
_________________
`
36% of all statistics are wrong |
|
| Back to top |
|
|
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
|
| Posted: Wed Feb 23, 2011 9:32 am Post subject: |
|
|
| Mike G wrote: | | DSMok1 wrote: | ... Replacement level in the NBA, per player, is somewhere around -3 to -3.5, as has been discussed elsewhere.
.. |
I had the impression that -3 to -3.5 (from average) was a 'reasonable guess'.
Has there been a confirmation or a strict definition? |
It's "somewhere around" that number. How to calculate replacement level is a bit convoluted. It's hard to tell where the level is above which the player is actually helping the team, despite being below average.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
| Back to top |
|
|
EvanZ
Joined: 22 Nov 2010
Posts: 307
|
|
| Back to top |
|
|
DSMok1
Joined: 05 Aug 2009
Posts: 611
Location: Where the wind comes sweeping down the plains
|
| Posted: Wed Feb 23, 2011 11:40 am Post subject: |
|
|
| EvanZ wrote: | | Why not make replacement level the average of all D-League callups during a season? Seems reasonable to me. |
I don't know if we have enough data from D-League callups yet.
Also... bad teams will call up players below replacement level, while good teams probably won't call anybody up. What we need is the average replacement level.
What makes it more confusing is that the rosters are big enough in the NBA that players sit on the bench, while still on salary.
_________________
GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
| Back to top |
|
|
EvanZ
Joined: 22 Nov 2010
Posts: 307
|
| Posted: Wed Feb 23, 2011 2:00 pm Post subject: |
|
|
| DSMok1 wrote: |
Also... bad teams will call up players below replacement level, while good teams probably won't call anybody up. What we need is the average replacement level.
|
Bad teams will call up the best D-League players. To me, by definition, those are replacement level players. The best possible replacement that you can easily get.
_________________
http://www.thecity2.com
http://www.ibb.gatech.edu/evan-zamir |
|
| Back to top |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
