Factors that Impact Team Performance/Home Court Advantage

[v-zero]

That's precisely what I did in my analysis.

One confounding issue I failed to deal with: I accounted for team strength over the course of the season. Individual games coaches *may* strategically concede/give up on early--like Popovich does overtly. In other words--the team may not be as good on a 4th in 5 game, purely because they are resting players somewhat. This would exaggerate the effect that I found. To account for that, I would need to account for exact minutes played and use a player rating system to estimate actual team strength for the given game to account for rest accurately.

Which is exactly what I did, and found that 4 in 5 was not significant AIC/BIC wise as long as the question of it being a back-to-back was taken into account on both sides. However I can't say that's the be all and end all of anything.

[DSMok1]

That's precisely what I did in my analysis.

One confounding issue I failed to deal with: I accounted for team strength over the course of the season. Individual games coaches *may* strategically concede/give up on early--like Popovich does overtly. In other words--the team may not be as good on a 4th in 5 game, purely because they are resting players somewhat. This would exaggerate the effect that I found. To account for that, I would need to account for exact minutes played and use a player rating system to estimate actual team strength for the given game to account for rest accurately.

Which is exactly what I did, and found that 4 in 5 was not significant AIC/BIC wise as long as the question of it being a back-to-back was taken into account on both sides. However I can't say that's the be all and end all of anything.

What was the coefficient, though? I would run 4in5 as a separate case (not including effect of b2b) and look at it that way. It may not have been significant due to the vast difference in number of instances between b2b and 4in5.

[v-zero]

Are you suggesting only considering games in which it is a 4in5 but not back to back? What about 4in5 for one and back to back for the opposition? I don't disagree that there's an issue of sample size with 4in5, but I'm not convinced it can be handled by ignoring other situations. I will probably have a look though, as there may be some hierarchy I can build to sort it.

[DSMok1]

When I did my analysis, I looked at the following cases for each team, concurrently, along with HCA:

Code: Select all

4 in 5        
3 in 4, BTB    
B2B           
3 in 4       
1 Day Rest     
2 Days Rest   
3-4 Days Rest 
5+ Days Rest

I essentially assigned a 0 or 1 to the relevant category when running the regression: only 1 of these categories would have a 1 for each team; the rest would be 0's.

Does that make sense?

[v-zero]

When I did my analysis, I looked at the following cases for each team, concurrently, along with HCA:

Code: Select all

4 in 5        
3 in 4, BTB    
B2B           
3 in 4       
1 Day Rest     
2 Days Rest   
3-4 Days Rest 
5+ Days Rest

I essentially assigned a 0 or 1 to the relevant category when running the regression: only 1 of these categories would have a 1 for each team; the rest would be 0's.

Does that make sense?

Perfect sense, I'll run those at some point this evening and post up what I find.

[v-zero]

Took longer than expected but here are my findings - each of these values should be added to the HCA of 3 (where it ended up when all of these were considered).

Code: Select all

hb2b , -0.6
h4in5 , -0.92
h3in4 , -0.01
h3in4b2b , -1.29
hrest1 , 0.11
hrest2 , 0.67
hrest3or4 , 1.08
hrest5ormore , -0.25
ab2b , 0.91
a4in5 , 1.65
a3in4 , -0.06
a3in4b2b , 0.55
arest1 , -0.97
arest2 , 0.3
arest3or4 , -0.57
arest5ormore , 0.26

Hopefully the names are self explanatory.

[DSMok1]

Wonderful work, v-zero!

Some of the empirical numbers don't quite make sense here-- You'd think that there would be a monotonic progression B2B<3in4B2B<4in5. Of course, empirical numbers don't usually make sense completely!

What was the sample size for each case?

[v-zero]

Wonderful work, v-zero!

Some of the empirical numbers don't quite make sense here-- You'd think that there would be a monotonic progression B2B<3in4B2B<4in5. Of course, empirical numbers don't usually make sense completely!

What was the sample size for each case?

I completely agree that some of these are very unintuitive theoretically, and probably flat out wrong - sample sizes were as such:

Code: Select all

hb2b , 286.0
h4in5 , 209.0
h3in4 , 1253.0
h3in4b2b , 463.0
hrest1 , 2420.0
hrest2 , 1279.0
hrest3or4 , 690.0
hrest5ormore , 478.0
ab2b , 568.0
a4in5 , 330.0
a3in4 , 1117.0
a3in4b2b , 1043.0
arest1 , 2154.0
arest2 , 1059.0
arest3or4 , 452.0
arest5ormore , 355.0

[mystic]

Hopefully the names are self explanatory.

They are. Maybe you want to include a variable for the difference in days of rest between the home and away team. Meaning, the home team had 1 day rest and the away team 0, the variable would be 1; if the home team has a b2b and the away team 3 days rest, it would be -3. Maybe that makes the rest of the variables obsolet and will dimish the influence of the differences in sample size.

[v-zero]

Hopefully the names are self explanatory.

They are. Maybe you want to include a variable for the difference in days of rest between the home and away team. Meaning, the home team had 1 day rest and the away team 0, the variable would be 1; if the home team has a b2b and the away team 3 days rest, it would be -3. Maybe that makes the rest of the variables obsolet and will dimish the influence of the differences in sample size.

I have tried that, but it's no better because it completely ignores whether a team has had one day of rest but three straight prior to that etc etc. I also think there is good reason to believe that home fatigue and away fatigue will be different (do you feel more tired after a day of travelling a lot but doing little or a day of relaxing at home/in comfort?).

[mystic]

I have tried that, but it's no better because it completely ignores whether a team has had one day of rest but three straight prior to that etc etc.

Well, in that case, you can include the amount of games within a certain timespan as well for each team.

I also think there is good reason to believe that home fatigue and away fatigue will be different (do you feel more tired after a day of travelling a lot but doing little or a day of relaxing at home/in comfort?).

Do you account for the previous game in that fashion? Only because a team has a b2b at home doesn't mean that they didn't travel at all, because the previous game could have been on the road. Imagine the Spurs play at home against the Mavericks and both are on b2b, the Mavericks played the previous game in Houston while the Spurs just had a game in Orlando. Which team had the tougher travelling schedule here? If you want to make it more accurate, taking into account the travelling distance as well as the game time would be an idea. Question would be, how much does that increase the predictive power? Is that worth it?

[v-zero]

I have tried that, but it's no better because it completely ignores whether a team has had one day of rest but three straight prior to that etc etc.

Well, in that case, you can include the amount of games within a certain timespan as well for each team.

I also think there is good reason to believe that home fatigue and away fatigue will be different (do you feel more tired after a day of travelling a lot but doing little or a day of relaxing at home/in comfort?).

Do you account for the previous game in that fashion? Only because a team has a b2b at home doesn't mean that they didn't travel at all, because the previous game could have been on the road. Imagine the Spurs play at home against the Mavericks and both are on b2b, the Mavericks played the previous game in Houston while the Spurs just had a game in Orlando. Which team had the tougher travelling schedule here? If you want to make it more accurate, taking into account the travelling distance as well as the game time would be an idea. Question would be, how much does that increase the predictive power? Is that worth it?

I agree that these are useful to include, with the other big deal thing to include being atmospheric pressure. The thing with handling these low samples is all about finding a reasonable number of bins such that you get adequate granularity (travelling 100 miles != travelling 1000 miles) but you also get a decent sample size.

[DSMok1]

When I did my research, I lumped away and home rest situations and just let the HCA deal with that. Splitting them apart--isn't that redundant with HCA?

[v-zero]

When I did my research, I lumped away and home rest situations and just let the HCA deal with that. Splitting them apart--isn't that redundant with HCA?

I would think not with HCA as a constant.