Latrinsorm
09-19-2013, 08:40 PM
I have mentioned before how it is possible to model (http://en.wikipedia.org/wiki/Pythagorean_expectation) wins and losses with the rate of points scored, paying no regard to clutchness or anything else. But another way of looking at this is what could those remainders or differences between the model and the actual be? Better home court advantage? Coaching 'em up? Superstar calls? Clutchness? Or just random noise?
Well, I tabulated all the differences (please note: not the magnitudes or squares of those differences, as we might do if we wanted to maximize our regression's accuracy) for each team over the last 10 years, giving us 299 team-seasons (as the Bobcats only came into the league in 2005). The total sum of these differences was -19.24, or the average team in an average season underperformed its projection by 0.06 wins. That is very close to zero, which is a good start for the random noise theory, which should increasingly cancel out when divided over increased time. If there are equal amounts of clutchness and chokeness, though, we would see the same result. We can also look at the distribution against a Gaussian model:
http://img.photobucket.com/albums/v456/johnnyoldschool/NBAPythagorasClutch2_zps76e6a2de.jpg
...again, it looks pretty close.
Next I summed all the differences for each team over this span and got:
ATL: 5.63
BOS: -11.82
BRO: 16.63
CHA: -0.10
CHI: 3.95
CLE: 10.07
DAL: 22.67
DEN: 4.29
DET: -0.32
GS: -8.34
HOU: -15.75
IND: -8.81
LAC: 6.29
LAL: 12.67
MEM: -7.43
MIA: 2.50
MIL: -10.10
MIN: -19.72
NO: 9.02
NY: -7.86
OKC: 1.54
ORL: -14.68
PHI: -5.63
PHO: 6.88
POR: 10.47
SAC: -5.82
SA: -8.48
TOR: -15.05
UTA: 11.39
WAS: -3.35
DAL: 22.67
BRO: 16.63
LAL: 12.67
UTA: 11.39
POR: 10.47
CLE: 10.07
NO: 9.02
PHO: 6.88
LAC: 6.29
ATL: 5.63
DEN: 4.29
CHI: 3.95
MIA: 2.50
OKC: 1.54
CHA: -0.10
DET: -0.32
WAS: -3.35
PHI: -5.63
SAC: -5.82
MEM: -7.43
NY: -7.86
GS: -8.34
SA: -8.48
IND: -8.81
MIL: -10.10
BOS: -11.82
ORL: -14.68
TOR: -15.05
HOU: -15.75
MIN: -19.72
Or in graphical form:
http://img.photobucket.com/albums/v456/johnnyoldschool/NBAPythagorasClutch_zps1d01fce7.jpg
These are (almost) all sums over 10 years, so Dallas' 22 means they have outperformed their projection by a little over 2 wins per season, on average. I am pretty sure I didn't correct for Charlotte only being in the league 9 years because 1. who cares about Charlotte, (b) their bar is so puny you wouldn't be able to see a difference anyway, and IV - blame it on LeBron. Also keep in mind that OKC was Seattle for part of this era, and New Orleans was homeless. I can see two plausible non-noise hypotheses being raised from this data:
1. Coaching matters.
Carlisle, Phil Jackson, Sloan in the top 5, the only name coaches in the bottom 5 are Stan Van and Doc, and Doc has famously had some trouble with end of game situations. It's a little iffy because... the Nets? But the Nets did have Lawrence Frank for 5 and a half years.
2. Home courts are not all created equal.
The bottom 5 are just lousy cities. What are you doing in Toronto? Nothing. Well-rested and ready to go the next day. Whereas at the top you've got Dallas bang bang shoot guns, New York City, Los Angeles, Portland (get that weed). Salt Lake stands out but instead of having night life they have famously berserk fans. It's a little iffy too because how come the New York Knicks are so low? Because the Knicks suck.
But it's probably just noise.
.
Two final notes: if we take Cleveland 2004-2010 and Miami 2011-2013, we get a value of 9.732. If we take the last 10 NBA Champions, we find that in the regular season they underperformed their projections by an average of 0.04 wins per year and had an average rank of exactly 15th, so in each case they were a very very very tiny bit above average. The two least clutch teams: 2007 Spurs and 2004 Pistons. The two most clutch: 2009 Lakers and 2013 Heat.
Well, I tabulated all the differences (please note: not the magnitudes or squares of those differences, as we might do if we wanted to maximize our regression's accuracy) for each team over the last 10 years, giving us 299 team-seasons (as the Bobcats only came into the league in 2005). The total sum of these differences was -19.24, or the average team in an average season underperformed its projection by 0.06 wins. That is very close to zero, which is a good start for the random noise theory, which should increasingly cancel out when divided over increased time. If there are equal amounts of clutchness and chokeness, though, we would see the same result. We can also look at the distribution against a Gaussian model:
http://img.photobucket.com/albums/v456/johnnyoldschool/NBAPythagorasClutch2_zps76e6a2de.jpg
...again, it looks pretty close.
Next I summed all the differences for each team over this span and got:
ATL: 5.63
BOS: -11.82
BRO: 16.63
CHA: -0.10
CHI: 3.95
CLE: 10.07
DAL: 22.67
DEN: 4.29
DET: -0.32
GS: -8.34
HOU: -15.75
IND: -8.81
LAC: 6.29
LAL: 12.67
MEM: -7.43
MIA: 2.50
MIL: -10.10
MIN: -19.72
NO: 9.02
NY: -7.86
OKC: 1.54
ORL: -14.68
PHI: -5.63
PHO: 6.88
POR: 10.47
SAC: -5.82
SA: -8.48
TOR: -15.05
UTA: 11.39
WAS: -3.35
DAL: 22.67
BRO: 16.63
LAL: 12.67
UTA: 11.39
POR: 10.47
CLE: 10.07
NO: 9.02
PHO: 6.88
LAC: 6.29
ATL: 5.63
DEN: 4.29
CHI: 3.95
MIA: 2.50
OKC: 1.54
CHA: -0.10
DET: -0.32
WAS: -3.35
PHI: -5.63
SAC: -5.82
MEM: -7.43
NY: -7.86
GS: -8.34
SA: -8.48
IND: -8.81
MIL: -10.10
BOS: -11.82
ORL: -14.68
TOR: -15.05
HOU: -15.75
MIN: -19.72
Or in graphical form:
http://img.photobucket.com/albums/v456/johnnyoldschool/NBAPythagorasClutch_zps1d01fce7.jpg
These are (almost) all sums over 10 years, so Dallas' 22 means they have outperformed their projection by a little over 2 wins per season, on average. I am pretty sure I didn't correct for Charlotte only being in the league 9 years because 1. who cares about Charlotte, (b) their bar is so puny you wouldn't be able to see a difference anyway, and IV - blame it on LeBron. Also keep in mind that OKC was Seattle for part of this era, and New Orleans was homeless. I can see two plausible non-noise hypotheses being raised from this data:
1. Coaching matters.
Carlisle, Phil Jackson, Sloan in the top 5, the only name coaches in the bottom 5 are Stan Van and Doc, and Doc has famously had some trouble with end of game situations. It's a little iffy because... the Nets? But the Nets did have Lawrence Frank for 5 and a half years.
2. Home courts are not all created equal.
The bottom 5 are just lousy cities. What are you doing in Toronto? Nothing. Well-rested and ready to go the next day. Whereas at the top you've got Dallas bang bang shoot guns, New York City, Los Angeles, Portland (get that weed). Salt Lake stands out but instead of having night life they have famously berserk fans. It's a little iffy too because how come the New York Knicks are so low? Because the Knicks suck.
But it's probably just noise.
.
Two final notes: if we take Cleveland 2004-2010 and Miami 2011-2013, we get a value of 9.732. If we take the last 10 NBA Champions, we find that in the regular season they underperformed their projections by an average of 0.04 wins per year and had an average rank of exactly 15th, so in each case they were a very very very tiny bit above average. The two least clutch teams: 2007 Spurs and 2004 Pistons. The two most clutch: 2009 Lakers and 2013 Heat.