2017-18 team win projection contest

[Crow]

There are some projections out there but not here, at least not yet. If they were published by deadline they "could" be included in broader overview as in past but I'm probably not going to chase and copy them.

[shadow]

But this team has under-won, relative to their season point-differential, for 10 straight seasons.
Last year, Minny 'should have won' 38 but took just 31 -- minus 7. Previous years: -3, -4, -8...

Dean Oliver touched on this in his book and I found an excerpt here: http://www.rawbw.com/~deano/articles/BellCurve.html

"The Phoenix Suns won 59 games in 1994-95, but if you look at their offensive and defensive point distributions, they were estimated to win only 52. Such a large difference means Phoenix got a bit lucky to win those other 7 games. It's normal for teams to have a difference of up to about four or five games, but seven is pretty large. Part of the Suns difference is the fact that they were 4-1 in games decided by two points or less and 8-2 in games decided by three points or less, usually games that involve a bit of luck."

There was also a study done on college basketball regarding close games in college basketball here: http://blog.minitab.com/blog/the-statis ... ll-part-ii

They looked at teams who over-performed, under-performed and were roughly 0.500 in close games in the first half of the season and how they performed in close games the second half of the season. The teams in the over and underperforming groups regressed very close to 0.500 in the second half of the season in close games.

Their conclusion was:

"The p-value of .213 is greater than 0.05, so we cannot conclude that the difference in winning percentage in close games between the two groups is significant. So overall we can conclude there just isn't any skill in winning or losing close games. "

Since winning close games appears to be dictated more by random variance than skill, we'd expect the differentials between pythag and actual wins for teams to be roughly normally distributed. If you look at the graphs in the article linked below that appears to be the case.

https://fansided.com/2017/09/18/nylon-c ... tribution/

They found roughly 60% of teams were within two wins and 94% of teams were within five wins of their pythag total, with a bell curve like distribution.

Different sport, but last year the Texas Rangers exceeded their pythag win total by 13 games primarily due to their record in 1 run games which was an absurd 36-11. I believe that was the record for the modern era. I'm not exactly sure how much roster turnover they had (which may explain some of the dropoff, but probably not a significant amount), but this season they had the worst record in the league in 1 run games at 13-24. If winning close games was skill based, I'm pretty sure the chances of a team experiencing that kind of decline in their record in 1 run games to be incredibly small. Which makes sense considering all the evidence I've seen indicates random chance has more to do with how teams perform in close games than skill.

[RyanRiot]

So far, we have 12 entries plus 538, and we collectively predict 48 wins for the T-Wolves. Vegas concurs, at 47.
But this team has under-won, relative to their season point-differential, for 10 straight seasons.
Most or all of our 'systems' predict scoring margin, and translate that into wins, yes?

Last year, Minny 'should have won' 38 but took just 31 -- minus 7. Previous years: -3, -4, -8...
Looking up the last 3 seasons' over-unders, weighting them by recency (4-3-2), we get these weighted averages:

Code: Select all

Wins-PythW     Wins-PythW
GSW   3.8      Min   -5.1
Mem   3.7      Uta   -3.2
Was   2.5      Cha   -2.4
Atl   2.0      Phx   -2.1
Bos   1.5      Phl   -2.0
Mil   1.3      Brk   -1.8
Hou   1.3      Det   -1.5
Chi   1.2      Den   -1.5
Cle   1.1      NYK   -1.1
Orl   1.0      NOP   -0.8
Dal   0.8      Sac   -0.8
Tor   0.6      LAL   -0.5
SAS   0.5      LAC   -0.5
Por   0.5      Ind   -0.4
Mia   0.2         
Okl   0.2         

It may be coaching, or it may be a 'culture' of winning/losing that drives these things.
Three seasons is of course a small sample size; and beyond 3 years, things may have changed.
I'm considering adding 1/2 or 1/3 of these values to all teams' win projections.

It's probably just bad luck/randomness. It's probably not coaching, as they've had 3 different coaches in those 3 seasons. Also, we have a much longer history of Thibs being a good coach than a bad one.

[caliban]

Edit1: Updated to the latest Pelton minutes

Code: Select all

CLE	53,9
TOR	50,8
BOS	48,7
WAS	46,3
CHO	45,9
MIA	45,7
MIL	43,0
DET	37,1
ORL	35,3
PHI	34,9
IND	31,4
BRK	28,7
NYK	28,5
ATL	27,3
CHI	26,1

GSW	69,1
HOU	60,9
SAS	53,3
OKC	52,2
MIN	48,2
LAC	48,1
DEN	44,8
POR	42,2
UTA	42,1
NOP	39,4
MEM	35,9
DAL	34,3
LAL	26,6
PHO	24,9
SAC	24,6

[Crow]

I might tinker with it a bit in next few hours but here is my set:

ATL 26
BOS 50
BRK 26
CHI 24
CHO 46
CLE 58
DAL 29
DEN 48
DET 34
GSW 70
HOU 58
IND 32
LAC 51
LAL 28
MEM 38
MIA 45
MIL 48
MIN 45
NOP 43
NYK 32
OKC 55
ORL 33
PHI 35
PHO 23
POR 43
SAC 23
SAS 50
TOR 47
UTA 40
WAS 50

As usual, it is a custom weighted blend of other projections with a final subjective tweaking up & down.

[knarsu3]

I'll put this in code a bit later but here are my win projections: https://medium.com/krishna-narsu/win-pr ... fdcfa9f4df

And in tableau form: https://public.tableau.com/profile/knar ... rojections

Code: Select all

Team	SHACPM	adjRPM-PTPM	RPM-PTPM	adjRPM	RPM
ATL	36.3	26.9	27.8	21.4	22.9
BKN	28.7	26.1	26.7	24.9	26.1
BOS	45.6	48.3	47.5	48.3	46.8
CHA	44.1	46.5	47.3	44.9	46.5
CHI	28.4	23.8	24.8	22.9	24.7
CLE	53.6	59.4	58.2	61.1	58.7
DAL	43.1	33.9	33.8	31.4	31.8
DEN	33.5	45.1	43.8	48.4	45.9
DET	33.6	33.0	33.2	32.7	33.0
GSW	70.1	70.3	70.0	69.7	69.0
HOU	56.9	61.5	61.0	62.5	61.5
IND	38.6	31.2	31.8	28.8	30.0
LAC	43.8	52.2	51.7	50.8	49.8
LAL	24.2	25.7	25.5	28.3	27.9
MEM	39.1	33.4	33.8	31.6	32.3
MIA	49.0	44.7	45.4	45.3	46.5
MIL	43.8	46.8	46.4	50.5	49.7
MIN	42.6	51.1	49.3	53.3	49.9
NOP	44.9	44.0	44.7	43.3	44.7
NYK	25.7	30.3	29.7	26.6	25.4
OKC	47.2	51.2	50.1	55.4	53.5
ORL	31.9	31.4	31.2	30.1	29.5
PHI	40.2	34.5	35.3	35.2	36.7
PHX	24.5	22.6	22.3	23.8	23.1
POR	38.4	39.2	39.8	42.3	43.4
SAC	26.3	23.7	24.1	22.3	23.0
SAS	56.8	56.5	57.0	54.8	55.7
TOR	49.2	46.8	46.4	46.3	45.5
UTA	42.4	41.6	43.4	43.5	47.0
WAS	47.7	48.2	48.2	49.6	49.5

[Kevin Pelton]

Minutes projections updated with final rosters for anyone still looking to get projections in:
https://drive.google.com/file/d/0By9ByT ... BHZEU/view

[kmedved]

Here's my entry:

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Also visually:



Let me know if another format would be preferable.

Minor edit for the Mirotic thing.

[Nathan]

I'll throw my hat in the ring. These predictions are based on Vegas lines, some half-baked calculations I've done to try to predict which newcomers will be successful, and a lot of random personal biases and gut feelings

ATL 25
BOS 46
BRK 27
CHA 42
CHI 22
CLE 57
DAL 32
DEN 52
DET 36
GSW 65
HOU 60
IND 28
LAC 40
LAL 30
MEM 34
MIA 45
MIL 49
MIN 46
NOP 40
NYK 32
OKC 52
ORL 34
PHI 40
PHX 26
POR 45
SAC 28
SAS 57
TOR 48
UTA 42
WAS 50

EDIT: minor adjustment after Portis/Mirotic altercation

[lanqiu]

GSW 67.3
HOU 57.0
SAS 56.3
CLE 55.5
OKC 51.6
MIN 50.8
TOR 48.7
WAS 47.9
LAC 47.2
DEN 46.1
MIA 45.7
MIL 45.1
BOS 44.2
UTA 43.5
CHA 42.8
POR 40.4
NOP 38.7
DET 38.4
MEM 37.1
PHI 37.0
DAL 34.0
ORL 33.2
IND 32.5
NYK 30.2
LAL 28.7
BKN 28.2
PHX 25.9
ATL 25.8
SAC 25.1
CHI 25.1

[caliban]

Ugh, horrible stuff last night

Having a hard time accepting a submissions after the Hayward injury unless we all get a BOS update.

Edit: Hypothetical post Hayward BOS + CHI altercation update

[Mike G]

Since winning close games appears to be dictated more by random variance than skill, we'd expect the differentials between pythag and actual wins for teams to be roughly normally distributed. If you look at the graphs in the article linked below that appears to be the case.

https://fansided.com/2017/09/18/nylon-c ... tribution/

They found roughly 60% of teams were within two wins and 94% of teams were within five wins of their pythag total, with a bell curve like distribution...

Indeed we would expect it. Yet the Wolves own 3 of the 7 worst differentials (Wins - PythWins) in the last 16 seasons. And all in the last 10 years (since KG left).

The odds of one team owning those, and all 10 years being negative -- all by pure chance -- are about one in 30 billion.
Since I cherry-picked 10 years, ignore the first of those seasons, and it's just one in 13 billion.

Code: Select all

year   W  Pyth  diff  prob
2014   40   48   -8   .006
2017   31   38   -7   .008
2011   17   24   -7   .008
2015   16   19   -3   .192
2013   31   34   -3   .192
2009   24   27   -3   .192
2016   29   31   -2   .300
2012   26   28   -2   .300
2010   15   17   -2   .300
2008   22   23   -1   .431
product           .0000000
inverse     31,617,247,230

[Crow]

I did say a deadline of 5 pm yesterday or up to midnight Thursday WITH an excuse.

Lanqiu, was there a reason your entry was delayed? Did you know about the Hayward injury before submitting and was it a factor in your projection?

My inclination is to not have a wave of projection adjustments. My inclination is to recognize the top 3 finishers. The methods for calculating win "errors" varied last year. They may again this season. Who is "in" might vary in people's eyes. Let's wait a bit for a reply and see if other "late" entries knock on the door.

[Crow]

Lanqiu was slightly late last season but indicated the entry was not influenced by information after start of season. I let in then because he offered excuse and there was no uproar. viewtopic.php?f=2&t=9188&p=28161#p28161

[kmedved]

Sorry, was the deadline 5pm Eastern? I assume it was 5pm West Coast time, since that was tipoff, right?