Better home/road adjustments for NBA 2006-07
I wanted to develop a better "home court advantage" factor to use in making a line. Last year I would just look at the home team and credit them for something extra based on ratings I'd find published elsewhere.
Main changes this year:
1. Dealing with both teams involved in the game. The adjustment now accounts for two things -- the degree to which the road team is worse on the road and the degree to which the home team is better at home. (Something like "combined home/road bias" is actually a better term for what's measured than is "home court advantage.")
2. Talking the language of both pointspreads and moneylines. After last week (with big thanks to Ganchrow) there is very solid conversion data for winning percentage to pointspread points and vice versa. So we can create one table for accurate ML adjustments (expressed as an adjustment in probable winning percentage) and a second table for accurate pointspread adjustments.
3. Dealing with a seven-year track record of home and road winning percentage for each team. The years are weighted for recency (2005-06 counts for 24% of the rating down to 5% for 1999-00 -- except for the Bobcats, whose two years are weighted 62-38%.) I think this is worth doing: while players and coaches change rapidly, the cities and arenas (for the most part) don't change or only change once. (That Denver has generated one of the widest spreads helps me think that 7 years are relevant.)
So here it is.
Biggest home/road swings: CLE, DEN, MIL
Smallest home/road swings: HOU, PHI, TOR
How to use (if you're so inclined):
CALCULATING WINNING PERCENTAGE FOR A GIVEN GAME
Start with your base, neutral-court calculation of each team's chances to win the game. Add the home team's number below to their estimated winning percentage and subtract the road team's number below from their estimated winning percentage:
ATL===12.4%
BOS===8.5%
CHA===10.5%
CHI===9.1%
CLE===15.0%
DAL===8.5%
DEN===14.7%
DET===10.2%
GS===10.8%
HOU===5.1%
IND===12.6%
LAC===12.0%
LAX===12.0%
MEM===11.3%
MIA===11.9%
MIL===13.5%
MIN===11.3%
NJ===12.6%
NO===9.1%
NY===11.5%
ORL===12.6%
PHI===7.2%
PHX===9.6%
POR===9.6%
SAC===12.4%
SAN===12.0%
SEA===7.3%
TOR===7.3%
UTH===9.7%
WAS===12.0%
Example: you handicap CLE at 52% to beat WAS on a neutral court. For a game at CLE, their 52% goes to 67.0% and WAS's 48% drops to 36.0%. The ML is 67/36, or CLE -186.
The average team's percentage adjustment is 10.7 -- which would put two dead-average teams at 60.7%/39.3% in favor of the home team, which is basically the true league percentage for all games actually played.
CALCULATING POINTSPREAD FOR A GIVEN GAME
Begin with your estimated pointspread on a neutral court. To account for home/road bias you would combine the road team's number with the home team's number and adjust the spread, in the home team's favor of course, by all of that combined number.
ATL===1.9
BOS===1.4
CHA===1.6
CHI===1.5
CLE===2.5
DAL==1.4
DEN===2.5
DET===1.6
GS===1.7
HOU===1.0
IND===1.9
LAC===1.8
LAX===1.8
MEM===1.7
MIA===1.8
MIL===2.1
MIN===1.7
NJ===1.9
NO===1.5
NY===1.8
ORL===1.9
PHI===1.3
PHX===1.5
POR===1.5
SAC===1.9
SAN===1.8
SEA===1.3
TOR===1.3
UTH===1.5
WAS===1.8
Although it may not look so radical, this creates a meaningful range. A CLE-DEN game would have a 5.0 point adjustment in the home team's favor, and a HOU-PHI game less than half that (2.3).
The average home team is a 3.4 point favorite over the average road team, which again converts to 61% SU winners, right on the historical average for the actual games.
Main changes this year:
1. Dealing with both teams involved in the game. The adjustment now accounts for two things -- the degree to which the road team is worse on the road and the degree to which the home team is better at home. (Something like "combined home/road bias" is actually a better term for what's measured than is "home court advantage.")
2. Talking the language of both pointspreads and moneylines. After last week (with big thanks to Ganchrow) there is very solid conversion data for winning percentage to pointspread points and vice versa. So we can create one table for accurate ML adjustments (expressed as an adjustment in probable winning percentage) and a second table for accurate pointspread adjustments.
3. Dealing with a seven-year track record of home and road winning percentage for each team. The years are weighted for recency (2005-06 counts for 24% of the rating down to 5% for 1999-00 -- except for the Bobcats, whose two years are weighted 62-38%.) I think this is worth doing: while players and coaches change rapidly, the cities and arenas (for the most part) don't change or only change once. (That Denver has generated one of the widest spreads helps me think that 7 years are relevant.)
So here it is.
Biggest home/road swings: CLE, DEN, MIL
Smallest home/road swings: HOU, PHI, TOR
How to use (if you're so inclined):
CALCULATING WINNING PERCENTAGE FOR A GIVEN GAME
Start with your base, neutral-court calculation of each team's chances to win the game. Add the home team's number below to their estimated winning percentage and subtract the road team's number below from their estimated winning percentage:
ATL===12.4%
BOS===8.5%
CHA===10.5%
CHI===9.1%
CLE===15.0%
DAL===8.5%
DEN===14.7%
DET===10.2%
GS===10.8%
HOU===5.1%
IND===12.6%
LAC===12.0%
LAX===12.0%
MEM===11.3%
MIA===11.9%
MIL===13.5%
MIN===11.3%
NJ===12.6%
NO===9.1%
NY===11.5%
ORL===12.6%
PHI===7.2%
PHX===9.6%
POR===9.6%
SAC===12.4%
SAN===12.0%
SEA===7.3%
TOR===7.3%
UTH===9.7%
WAS===12.0%
Example: you handicap CLE at 52% to beat WAS on a neutral court. For a game at CLE, their 52% goes to 67.0% and WAS's 48% drops to 36.0%. The ML is 67/36, or CLE -186.
The average team's percentage adjustment is 10.7 -- which would put two dead-average teams at 60.7%/39.3% in favor of the home team, which is basically the true league percentage for all games actually played.
CALCULATING POINTSPREAD FOR A GIVEN GAME
Begin with your estimated pointspread on a neutral court. To account for home/road bias you would combine the road team's number with the home team's number and adjust the spread, in the home team's favor of course, by all of that combined number.
ATL===1.9
BOS===1.4
CHA===1.6
CHI===1.5
CLE===2.5
DAL==1.4
DEN===2.5
DET===1.6
GS===1.7
HOU===1.0
IND===1.9
LAC===1.8
LAX===1.8
MEM===1.7
MIA===1.8
MIL===2.1
MIN===1.7
NJ===1.9
NO===1.5
NY===1.8
ORL===1.9
PHI===1.3
PHX===1.5
POR===1.5
SAC===1.9
SAN===1.8
SEA===1.3
TOR===1.3
UTH===1.5
WAS===1.8
Although it may not look so radical, this creates a meaningful range. A CLE-DEN game would have a 5.0 point adjustment in the home team's favor, and a HOU-PHI game less than half that (2.3).
The average home team is a 3.4 point favorite over the average road team, which again converts to 61% SU winners, right on the historical average for the actual games.