View previous topic :: View next topic |
Author |
Message |
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Wed Nov 10, 2010 6:20 pm Post subject: Effect of Rest Days on Efficiencies |
|
|
Rest Day Analysis
NBA Stuffer has done some good work on rest days, but I thought I'd attack it more rigorously.
I compiled all regular season games in the last 3 years, each team's efficiencies and the pace of the game. I also generated their rest situation, using the same set of cases that NBA Stuffer did:
4th in 5 days
3rd in 4 days, B2B
Back-to-back (other)
3rd in 4 days, 1 day rest
1 Day Rest (other)
2 Days Rest
3+ Days Rest
To estimate these values, I ran a regression on each year's data, solving for each team's offensive and defensive ratings, the effect of home court advantage, and the effect of the rest days. I then averaged the 3 year's results, based on the number of observations of each type.
Here are the results:
Code: | Off Rtg Def Rtg Eff Diff Pace Observations
4th in 5 days -0.63 1.34 -1.97 -0.42 230
3rd in 4 days, B2B -0.30 1.37 -1.67 -0.32 942
Back-to-back (other) -0.31 1.13 -1.44 -0.11 608
3rd in 4 days, 1 day rest 0.07 -0.03 0.10 0.10 1126
1 Day Rest (other) 0.14 -0.51 0.64 -0.06 2778
2 Days Rest 0.64 -0.22 0.85 0.21 1164
3+ Days Rest -1.11 -1.14 0.03 0.50 532 |
And here's a pretty chart:
The home court efficiency advantage generated was 4.68. I generated the estimates for each game using the form Team1*Team2/LgAvg for each portion of the estimate (offense, defense, pace). _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
inkt2002
Joined: 10 Oct 2009 Posts: 8
|
Posted: Wed Nov 10, 2010 10:37 pm Post subject: |
|
|
Solid analysis as always. Very interesting. Amazing that the biggest drop-off in efficiency would be offensively when resting 2 days vs. 3+ days.
Thanks for sharing. |
|
Back to top |
|
|
Serhat Ugur (hoopseng)
Joined: 13 Oct 2006 Posts: 209 Location: Basketball Research
|
Posted: Thu Nov 11, 2010 1:53 am Post subject: |
|
|
@DSMok1, thanks for the analysis. Maybe we should increase observations and extend this study by going back to -let's say- 10 years.
From my archive, here's another study on rest days. _________________ http://www.nbastuffer.com |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Thu Nov 11, 2010 9:23 am Post subject: |
|
|
Serhat Ugur (hoopseng) wrote: | @DSMok1, thanks for the analysis. Maybe we should increase observations and extend this study by going back to -let's say- 10 years.
From my archive, here's another study on rest days. |
I had not realized that the home team typically has more rest than the road team! That's interesting...
I have the framework in place to do this for as far back as we want to look, though it takes time to set up the data.
I think that one of the primary reasons for the bad offensive numbers on the 3+ days of rest is that every team has that on the first day of the year, and I think defenses typically dominate the first day.
Perhaps I should add a factor that looks at what point in the season the game occurs? Perhaps offenses progress after, say, the first month, because it takes more practice working together at game speed to get the offense rolling. _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
greyberger
Joined: 27 Sep 2010 Posts: 53
|
Posted: Thu Nov 11, 2010 2:38 pm Post subject: |
|
|
If it's not a huge headache you could try breaking it down further into "3 or more days rest, inside the season" and "Returning from break", or just chopping off the starting day games (and post all-star break games, if needed). |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Thu Nov 11, 2010 6:55 pm Post subject: |
|
|
greyberger wrote: | If it's not a huge headache you could try breaking it down further into "3 or more days rest, inside the season" and "Returning from break", or just chopping off the starting day games (and post all-star break games, if needed). |
I think I'll graph the whole season and see if there seems to be particular trends visually. Your approach may be what is required... _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
Marver
Joined: 21 Jun 2009 Posts: 5
|
Posted: Sat Nov 13, 2010 5:00 pm Post subject: |
|
|
I'm not so sure this is an effect of the actual players performing worse than normal, or if this is largely a function of playing time. Surely some, if not all, coaches temper the playing time they give to their star players when playing multiple games in a short time-span. Granted, this wouldn't explain the differences observed for games played with excess rest, but it certainly could play a large mitigating factor for the other scenarios. Perhaps you could use player efficiency, rather than team efficiency, as the main metric. |
|
Back to top |
|
|
Italian Stallion
Joined: 04 Mar 2009 Posts: 112
|
Posted: Tue Nov 16, 2010 10:52 pm Post subject: |
|
|
Is there any way to convert that into a Points Per Game impact? |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Thu Nov 18, 2010 10:10 am Post subject: |
|
|
Italian Stallion wrote: | Is there any way to convert that into a Points Per Game impact? |
It depends on the pace of the game (usually around 92 poss/game, check Basketball Reference for each team). The numbers are points/100 possessions, so it's an easy conversion. _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
bstenger
Joined: 10 Nov 2005 Posts: 15 Location: San Francisco, CA
|
Posted: Sun Nov 21, 2010 11:11 pm Post subject: |
|
|
There's recent research from Europe that looks at the how Champions League soccer players' injury rates and athletic performance depending on whether they play one game in a week, or two. The researcher, Gregory Dupont from University of Lille, found that injury rates increased substantially (6X more likely) but athletic performance (based on player-tracking data analysis) did not drop off. ... the paper from the American Journal of Sports Medicine, http://ajs.sagepub.com/content/early/2010/04/15/0363546510361236.abstract
I discuss the research in an article on NBA schedules at http://news.medill.northwestern.edu/chicago/news.aspx?id=171975 |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Tue Nov 23, 2010 10:06 am Post subject: |
|
|
DSMok1 wrote: | greyberger wrote: | If it's not a huge headache you could try breaking it down further into "3 or more days rest, inside the season" and "Returning from break", or just chopping off the starting day games (and post all-star break games, if needed). |
I think I'll graph the whole season and see if there seems to be particular trends visually. Your approach may be what is required... |
Here's that graph:
There are several distinct things to note here:
1) The playoffs are a lot slower paced
2) The pace is higher at the very beginning of the season and right after the all-star break, but is pretty consistent otherwise
3) The overall efficiency seems to increase quite a bit in the first week, then increase at a steady pace throughout the season (apparently, offense takes more practice than defense?)
I will endeavor to tease out these effects. I will change my regressions in the following ways:
1) Add another category: 3-4 days rest and 5+ days rest. The latter only occurs at the start of the season and at the all-star break.
2) Add a modifier for the first week of the season that simply adds a constant to pace and efficiencies
3) Add a linearly increasing term for the efficiencies over the course of the regular season.
4) Add a constant and a second linear term for the duration of the playoffs, for both efficiencies and pace
Once running this more intricate regression, I should be able to separate out all of the effects. _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
mtamada
Joined: 28 Jan 2005 Posts: 377
|
Posted: Tue Nov 23, 2010 4:43 pm Post subject: |
|
|
Is "RTG" the offensive rating, or defensive rating, or both combined? Or the offensive rating for both teams in a given game?
There is a big increase in the range and variance of RTG around Day 200; I'm presuming that's when the playoffs start; if the ratings are for teams rather than for games, then that variance might reflect #1 seeds stomping all over #8 seeds, resulting in sky-high ratings for the winners and low ratings for the losers. But I'm not sure what the RTG figures represent. |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Tue Nov 23, 2010 6:32 pm Post subject: |
|
|
mtamada wrote: | Is "RTG" the offensive rating, or defensive rating, or both combined? Or the offensive rating for both teams in a given game?
There is a big increase in the range and variance of RTG around Day 200; I'm presuming that's when the playoffs start; if the ratings are for teams rather than for games, then that variance might reflect #1 seeds stomping all over #8 seeds, resulting in sky-high ratings for the winners and low ratings for the losers. But I'm not sure what the RTG figures represent. |
Ratings are the average for the game. The playoffs start at day 170; there is a lot of variance because of fewer games/day over the 3 year sample (under 10 at that point).
In my initial regression it doesn't appear that there are any significant differences in the 1st week in the season for pace that are unexpected (other than the fact that after a long rest teams play faster). As I have the opportunity I'll keep everyone posted. _________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1 |
|
Back to top |
|
|
Italian Stallion
Joined: 04 Mar 2009 Posts: 112
|
Posted: Mon Dec 06, 2010 6:20 pm Post subject: |
|
|
What about if a team plays back to back at home both games vs. back to back where traveling is involved? |
|
Back to top |
|
|
DSMok1
Joined: 05 Aug 2009 Posts: 611 Location: Where the wind comes sweeping down the plains
|
Posted: Mon Dec 06, 2010 8:13 pm Post subject: |
|
|
Hmmm. It seems I forgot to update with the results of my proposed changes above.
Here are the results:
There are some changes in the results from the original regressions. I minimized absolute error in this regression, rather than squared error (I'm not sure what effect that has).
EDIT: I forgot to mention a few things. The corrected home court advantage is now 3.24.
Here is the table of actual numbers:
Code: | Case ORtg DRtg Differntial Pace Observations
4 in 5 -0.44 3.24 -3.68 -0.27 230
3 in 4, BTB -0.54 1.64 -2.19 -0.37 942
B2B -0.11 1.37 -1.48 0.20 608
3 in 4 -0.30 0.71 -1.01 0.10 1126
1 Day Rest 0.31 -1.63 1.94 -0.17 2778
2 Days Rest 0.71 0.31 0.40 0.31 1164
3-4 Days Rest -0.23 -0.04 -0.19 0.30 353
5+ Days Rest -3.25 1.48 -4.73 0.96 179 |
Also, I found a best-fit "scaler" that ends up reducing the projected efficiency differential if it's above average and increasing the projected efficiency differential if it is below average. Basically, this scales to adjust for the fact that good teams will take it easy against the worst teams. That "scaler" is 93.6%, and the league-average game differential is 6.394. So, if a game has a projected differential of 15 (rare) then the adjusted projected differential would be 6.394+(.936*(15-6.394))=14.44. It's just a slight tweak, but I thought it interesting. /EDIT
As a byproduct of these, here are the ratings of all of the teams in the last 3 years. These include the playoffs:
Code: | 2009-2010 Off Def Margin
ORL 3.29 -3.73 7.03
PHO 8.14 1.22 6.91
CLE 4.54 -2.05 6.59
UTA 4.00 -2.43 6.42
LAL 2.44 -3.81 6.24
ATL 5.43 0.48 4.96
BOS 0.37 -4.04 4.41
SAS 1.69 -2.60 4.29
DEN 3.35 0.14 3.21
POR 2.61 0.37 2.24
DAL 0.80 -1.36 2.16
OKC -1.16 -3.19 2.02
MIL -4.32 -5.97 1.66
MIA -0.40 -1.63 1.23
CHA -2.25 -3.23 0.98
HOU -0.07 -0.90 0.84
CHI -4.08 -3.77 -0.32
TOR 4.04 5.22 -1.18
SAC -1.66 0.67 -2.33
MEM -0.18 2.71 -2.89
NOH -0.36 2.61 -2.97
NYK 0.57 3.78 -3.21
GSW 0.69 3.91 -3.21
WAS -1.90 1.32 -3.22
IND -3.81 0.28 -4.09
PHI -2.33 1.81 -4.14
LAC -4.52 1.62 -6.14
DET -1.42 4.98 -6.40
MIN -7.04 2.92 -9.97
NJN -6.47 4.67 -11.14
2008-2009 Off Def Margin
CLE 5.52 -5.17 10.69
LAL 3.83 -5.14 8.96
BOS 2.40 -5.38 7.78
ORL -0.01 -6.98 6.97
HOU 0.92 -5.74 6.65
DEN 3.64 -0.97 4.62
POR 4.76 0.89 3.87
DAL 2.78 -0.44 3.21
UTA 0.83 -1.78 2.61
ATL 0.57 -0.67 1.24
CHI 1.17 -0.05 1.22
SAS -3.18 -4.37 1.19
NYK 1.44 0.42 1.03
NOH 0.14 -0.78 0.92
PHI -0.87 -0.72 -0.15
PHO 3.89 4.17 -0.28
IND -0.66 0.16 -0.82
CHA -2.60 -1.41 -1.19
MIA -1.41 0.30 -1.71
DET -1.89 -0.02 -1.87
MIL -1.07 0.94 -2.00
NJN 0.19 3.60 -3.41
GSW 1.50 5.15 -3.66
TOR 0.37 4.34 -3.96
MIN -2.16 3.14 -5.29
MEM -4.90 1.14 -6.04
OKC -5.35 1.56 -6.91
SAC -1.49 5.69 -7.17
LAC -5.65 2.27 -7.92
WAS -2.71 5.84 -8.55
2007-2008 Off Def Margin
BOS 3.45 -7.47 10.92
LAL 7.21 -3.18 10.38
NOH 6.88 -0.27 7.15
UTA 7.16 0.27 6.89
HOU -1.45 -7.58 6.13
DET 2.27 -3.51 5.78
SAS -0.20 -5.70 5.50
ORL 4.02 -0.78 4.80
PHO 6.86 2.26 4.59
DAL 4.50 0.20 4.30
DEN 1.37 -2.01 3.37
GSW 4.51 1.55 2.96
TOR 3.80 0.92 2.88
CLE -1.51 -4.15 2.64
WAS 2.10 0.62 1.48
POR 1.63 1.21 0.42
IND -0.88 -0.77 -0.11
PHI -1.83 0.46 -2.29
SAC -1.21 1.99 -3.21
CHI -4.01 -0.60 -3.41
CHA -2.81 1.49 -4.30
ATL -1.62 3.32 -4.94
NJN -4.08 1.25 -5.33
MIL -2.21 4.35 -6.55
MEM -2.50 4.13 -6.62
MIN -4.38 3.71 -8.09
LAC -6.38 1.85 -8.23
NYK -3.44 5.11 -8.55
SEA -8.88 0.13 -9.01
MIA -8.37 1.18 -9.55
|
_________________ GodismyJudgeOK.com/DStats
Twitter.com/DSMok1
Last edited by DSMok1 on Fri Dec 10, 2010 11:03 am; edited 1 time in total |
|
Back to top |
|
|
|