Page 2 of 3
Re: Home Court Advantage
Posted: Tue Dec 11, 2012 10:54 pm
by Eternal
No, by confounded I mean the two would be intertwined. It would be like having an alloy of gold and silver. What percentage of each can't be determined simply by weighing it. You'd need to know something else, like the volume, in order to determine the percentages of the two (assuming you know the density of gold and silver).
-Chris
Re: Home Court Advantage
Posted: Wed Dec 12, 2012 6:43 pm
by v-zero
Rest issues should be otherwise capturable, and these should mostly be capturing the other effects (largely altitude I would imagine).
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 3:26 am
by Eternal
The NCAA D1 team with the greatest home court advantage? Denver (elevation 5000+ feet). And all four Utah teams are well above average.
#2 D1 is Wisconsin (fans, noise?)
Top D2 - Billings-Montana (elevation 3000 feet, long travel time to reach)
#2 D2 - Western NM (elevation 6000 feet)
Top D3 - Sul Ross State (elevation 4500 feet)
#2 D3 - Colorado College (elevation 6000+ feet)
-Chris
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 3:49 am
by Eternal
University of Denver win percentage by year:
Code: Select all
year | home | away | n
-----+-------+-------+----
2002 | 0.462 | 0.143 | 28
2003 | 0.769 | 0.182 | 32
2004 | 0.692 | 0.357 | 27
2005 | 0.923 | 0.400 | 31
2006 | 0.909 | 0.200 | 31
2007 | 0.286 | 0.000 | 29
2008 | 0.750 | 0.000 | 30
2009 | 0.800 | 0.143 | 31
2010 | 0.938 | 0.154 | 32
2011 | 0.733 | 0.167 | 30
2012 | 0.875 | 0.538 | 31
2013 | 0.125 | 0.000 | 7
Overall:
Code: Select all
home | away | neutral | n
------+-------+---------+-----
0.683 | 0.189 | 0.552 | 339
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 10:13 am
by Eternal
Wisconsin is equally insane:
Code: Select all
year | home | away | n
-----+-------+-------+----
2002 | 0.923 | 0.429 | 32
2003 | 0.941 | 0.545 | 32
2004 | 1.000 | 0.455 | 32
2005 | 0.938 | 0.455 | 34
2006 | 0.875 | 0.200 | 31
2007 | 1.000 | 0.700 | 36
2008 | 0.889 | 0.833 | 36
2009 | 0.813 | 0.364 | 33
2010 | 0.941 | 0.500 | 33
2011 | 1.000 | 0.455 | 34
2012 | 0.778 | 0.636 | 36
2013 | 0.857 | 0.000 | 11
(12 rows)
home | away | neutral | n
------+-------+---------+-----
0.915 | 0.500 | 0.603 | 380
-Chris
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 12:40 pm
by DSMok1
Eternal wrote:Wisconsin is equally insane:
Code: Select all
year | home | away | n
-----+-------+-------+----
2002 | 0.923 | 0.429 | 32
2003 | 0.941 | 0.545 | 32
2004 | 1.000 | 0.455 | 32
2005 | 0.938 | 0.455 | 34
2006 | 0.875 | 0.200 | 31
2007 | 1.000 | 0.700 | 36
2008 | 0.889 | 0.833 | 36
2009 | 0.813 | 0.364 | 33
2010 | 0.941 | 0.500 | 33
2011 | 1.000 | 0.455 | 34
2012 | 0.778 | 0.636 | 36
2013 | 0.857 | 0.000 | 11
(12 rows)
home | away | neutral | n
------+-------+---------+-----
0.915 | 0.500 | 0.603 | 380
-Chris
If you have 320 universities and the HCA's are distributed on a bell curve randomly, you would expect some outliers simply due to statistical variation. (Not necessarily reflecting a truly significant difference).
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 6:51 pm
by ed küpfer
It would be nice if these numbers could be tied into a more general theory of home court advantage. For example what is the relation of HCA to elevation? I remember reading a study once about acclimation to high elevation, which takes (IIRC) up to ten days and is not persistent. So some other questions that could be addressed for games at high elevation
1) does the home team perform better if they are playing during a homestand than if they are returning from a road trip?
2) does the road team perform better if they are traveling from another high elevation compared to a low elevation?
3) do players moving from a high elevation team to a low elevation team perform better than expected when they return to high elevation locations? ie is high elevation acclimation persistent through a player's subsequent career (eg does Nene perform better when he goes back to Denver or SLC than expected)?
and so on.
Re: Home Court Advantage
Posted: Sat Dec 15, 2012 7:31 pm
by mtamada
ed küpfer wrote:It would be nice if these numbers could be tied into a more general theory of home court advantage. For example what is the relation of HCA to elevation? I remember reading a study once about acclimation to high elevation, which takes (IIRC) up to ten days and is not persistent. So some other questions that could be addressed for games at high elevation
1) does the home team perform better if they are playing during a homestand than if they are returning from a road trip?
2) does the road team perform better if they are traveling from another high elevation compared to a low elevation?
3) do players moving from a high elevation team to a low elevation team perform better than expected when they return to high elevation locations? ie is high elevation acclimation persistent through a player's subsequent career (eg does Nene perform better when he goes back to Denver or SLC than expected)?
and so on.
Those tests will only work if altitude acclimation can get lost (or gained) in a small number of days, which seems unlikely given the 10-day figure you cite. I'd be more inclined to assume that the high altitude teams have a permanent advantage (when performing at high altitude) compared to low-altitude teams coming to visit them. Easy to check, and the figures for Denver and Utah suggest that this may be happening.
Of course, it's possible that Denver and Utah just plain perform better at home than they do on the road; there are always some teams (my impression is typically younger, not so good teams) that have a wider-than-usual gap between their home vs away performances. This is a classic example of an identification problem in econometrics. Do Denver and Utah play well at home due to altittude, or for other homecout-advantage reasons?
Easy way to resolve it: look at Denver and Utah's records when they play each other. If their homecourt advantage is caused by altitude, it will become smaller or disappear when two high altitude teams play each other. If it's not the altitude, then the two teams will keep on winning on their respective homecourts and losing on their opponent's court.
Small sample size will be a problem, one will have to look at a lot of seasons to get enough Den-Utah game results. Resorting to looking at college games might be an alternative, now there'll be a lot more teams and games and a bigger sample size. But we will need a good model of team strength and win probabilities, because of the large number of college teams with widely varying quality levels. A team might have a good homecourt record because they play a lot of weak teams, so we need good estimates of team strength. E.g. I don't know if Colorado College's data can be used; it's a Division 3 school, where estimates of team strength will be less accurate. In contrast NBA teams are smaller in number and play each other more frequently, making accurate team strength estimates easier to do.
And we could measure altitude as a continuous variable rather than as a binary high-vs-low variable. It's almost certainly a non-linear relationship however; I doubt there's much difference between being at sea level vs 3,000 feet; but expect a substantial difference between say 4,000 feet and 7,000 feet.
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 3:35 am
by Eternal
Interesting idea - I can pull elevations for all of these schools. I'll just need to add it to my code and re-run the scraper.
-Chris
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 6:12 am
by Eternal
Yep, sure enough, a team suffers when going up in elevation, but shows no penalty or a small gain when going down in elevation.
I ran it on data from 2011-2013 to get a snapshot, now I'm running it on the full data set.
-Chris
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 7:32 am
by Eternal
NCAA data from 2002-2013. Model includes distance traveled by team and opponent, individual home court factors, change in elevation by team and opponent.
Code: Select all
Team offensve impact. Level is altitude change in meters/250, factor is percentage change in scoring (1.0 = no change, higher means increase in scoring).
Negative level means team went up in altitude from home court.
Overall correlation between altitude and offensive impact is -0.55.
level | factor
-------+--------
11 | 0.920
10 | 0.942
9 | 0.976
8 | 0.972
7 | 0.950
6 | 0.989
5 | 0.988
4 | 0.994
3 | 0.994
2 | 0.997
1 | 0.997
0 | 1.000
-1 | 0.995
-2 | 1.004
-3 | 0.996
-4 | 0.994
-5 | 1.000
-6 | 1.004
-7 | 0.999
-8 | 1.007
-9 | 1.016
-10 | 0.990
-11 | 0.949
Opponent defensive impact. Level is altitude change in meters/250, factor is percentage change in scoring (1.0 = no change, higher means allows more scoring).
Negative level means opponent went up in altitude from home court.
Overall correlation between altitude and defensive impact is -0.70.
level | factor
-------+--------
11 | 0.956
10 | 1.008
9 | 0.984
8 | 0.992
7 | 1.001
6 | 1.006
5 | 1.007
4 | 1.008
3 | 1.013
2 | 1.000
1 | 0.999
0 | 1.000
-1 | 1.006
-2 | 1.010
-3 | 1.005
-4 | 1.007
-5 | 1.006
-6 | 1.031
-7 | 1.006
-8 | 1.025
-9 | 0.998
-10 | 1.042
-11 | 1.047
-Chris
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 7:34 am
by Eternal
There are some sample size issues regarding extreme changes in altitude, but I'm comfortable in saying that a team going from a low to an extreme altitude gets punished tremendously, whereas going from an extreme to a much lower altitude yields either no impact or a small positive.
-Chris
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 7:46 am
by Eternal
Accounting for elevation, here are the top 5 home courts for each division. Team is the ID from ncaa.com.
Code: Select all
school_name | team | home
--------------------+-------+-------
Wisconsin | 796 | 1.080
St. Louis | 609 | 1.065
Oklahoma | 522 | 1.064
Denver | 183 | 1.064
Ohio | 519 | 1.059
school_name | team | home
---------------------+-------+-------
Findlay | 1079 | 1.059
Tarleton St. | 1395 | 1.057
Western N.M. | 1445 | 1.049
Delta St. | 181 | 1.048
Grand Valley St. | 262 | 1.048
school_name | team | home
---------------------+-------+-------
Millsaps | 426 | 1.047
Juniata | 325 | 1.047
SCAD | 16006 | 1.044
Wis.-Stevens Point | 802 | 1.044
Concordia-Mhead | 161 | 1.044
-Chris
Re: Home Court Advantage
Posted: Sun Dec 16, 2012 7:50 am
by Eternal
Here's D2 Findlay's home/away record. They were undefeated in 2009 with a perfect 36-0 record (and D2 championship).
Code: Select all
year | home | away | n
-----+-------+-------+----
2002 | 0.857 | 0.667 | 30
2003 | 0.867 | 0.583 | 31
2004 | 0.933 | 0.583 | 32
2005 | 1.000 | 0.750 | 34
2006 | 1.000 | 0.750 | 30
2007 | 0.952 | 0.889 | 31
2008 | 1.000 | 0.643 | 33
2009 | 1.000 | 1.000 | 36
2010 | 1.000 | 0.500 | 31
2011 | 1.000 | 0.727 | 28
2012 | 0.938 | 0.538 | 31
2013 | 0.833 | 0.500 | 8
(12 rows)
home | away | neutral | n
------+-------+---------+-----
0.958 | 0.684 | 0.742 | 355
-Chris
Re: Home Court Advantage
Posted: Tue Nov 26, 2013 5:42 am
by theeditor
ed küpfer wrote:1) does the home team perform better if they are playing during a homestand than if they are returning from a road trip?
Yes, a home team off a home game has a higher HCA than if they come in off a road game. The difference is not all that large though. The magnitude change in HCA when the *road* team is off a home vs road game is much larger.
The general principle is: whatever perturbation you can think of, the magnitude of the change in HCA is greater when the perturbation affects the road team. Doesn't matter; could be a site switch, could be rest, whatever. The change in HCA will be larger for the road team.