Page 4 of 11
Re: 2019-20 team win projection contest
Posted: Tue Oct 15, 2019 6:11 pm
by Kevin Pelton
Justin Jackson is "Justin A. Jackson" in my spreadsheet to differentiate from the 2018 Orlando second-round pick.
Andre Iguodala and Josh Jackson are both on the Grizzlies' roster, for the record, but I'm not projecting them to see any minutes at this point as Jackson is reporting to the G League and Iguodala is staying home.
Re: 2019-20 team win projection contest
Posted: Wed Oct 16, 2019 1:24 am
by RyanRiot
Mike G wrote: ↑Sun Oct 13, 2019 4:59 pm
Subject to revisions and reviews, based on K Pelton minutes linked upthread.
Code: Select all
west wins east wins
LAL 54.7 Phl 53.9
Hou 53.9 Mil 52.6
LAC 51.9 Orl 45.5
Uta 50.7 Ind 45.2
Por 46.7 Bos 42.3
SAS 45.5 Brk 41.7
Den 44.8 Det 41.4
NOP 41.9 NYK 40.5
GSW 41.0 Tor 37.9
Dal 40.6 Chi 36.3
Sac 39.5 Mia 36.2
Min 36.7 Cle 35.4
Phx 35.0 Atl 34.6
OkC 28.2 Was 27.7
Mem 21.5 Cha 26.1
Kevin P has some mighty funny minute projections, but I'll go with it.
I've made no attempt to account for near-identical superstars on the same team: Harden/RW, George/Kawhi
After scaling to 41 wins per team, the conference imbalance is extreme. So I added 1 win to each East team and dropped the West by 1.
I think that's the highest projection I've seen on the Knicks by about 8 wins. What players does your model like on the Knicks better than other models?
Re: 2019-20 team win projection contest
Posted: Wed Oct 16, 2019 3:40 am
by Rd11490
Using the Blend column for the contest:
Edit: edited 10/17/19
Re: 2019-20 team win projection contest
Posted: Wed Oct 16, 2019 12:32 pm
by Mike G
RyanRiot wrote: ↑Wed Oct 16, 2019 1:24 am
I think that's the highest projection I've seen on the Knicks by about 8 wins. What players does your model like on the Knicks better than other models?
I haven't looked at other models, but I have looked into the Knicks, noting the anomaly as you have.
According to the K Pelton minutes distribution which I used, only 3/8 of the playing time will be from players who won just 17 games last year. The rest is coming from an interesting assortment of pretty-good new acquistions.
Here are their 2018-19 stats from b-r.com along with my own entirely-boxscore eWins:
Code: Select all
. Knicks Min? e484 PER WS/48 BPM
Julius Randle 2340 1.65 21.0 .131 1.4
Kevin Knox 2184 .43 8.7 -.030 -6.0
R.J. Barrett * 2184 1.25 17.0 .125 1.0
Marcus Morris 1900 1.04 14.2 .104 -1.4
M Robinson 1872 1.31 22.0 .217 5.7
Bobby Portis 1824 1.28 15.3 .058 -2.8
Dennis Smith 1824 .81 13.5 -.014 -2.7
Elfrid Payton 1752 1.03 14.0 .058 -0.9
Allonzo Trier 1155 .52 12.2 .030 -4.2
Wayne Ellington 924 .57 11.8 .097 .2
Taj Gibson 624 1.14 17.8 .144 1.0
Reggie Bullock 590 .43 10.5 .078 -0.9
Frank Ntilikina 273 .12 6.0 -.045 -5.7
Damyean Dotson 234 .47 10.8 .037 -2.3
. total 19680 1.00 15.1 .079 -0.9
Of the returning players, only Robinson is remotely 'good' by any measure. I don't count him as the superstar others do, yet.
I have no clue why Knox should continue getting major minutes, while DD and Frank Niti get almost none. Or why Taj coming off his best year, according to some, should be 3rd string.
* Barrett numbers are made up from nothing but a formula (for eWins based on his being #3 in the draft) and rough equivalents in the others.
PER concludes they're an avg NBA team, which should make playoffs in the East. BPM says they're on the bubble. WS isn't that hot on them, but much better than last year.
Re: 2019-20 team win projection contest
Posted: Thu Oct 17, 2019 8:24 pm
by Mike G
Rd11490 wrote: ↑Wed Oct 16, 2019 3:40 am
Using the Blend column for the contest:
...
Edit: edited 10/17/19
Your current predictions bear little resemblance to your original.
What do "3 year" and "1 year" refer to?
Re: 2019-20 team win projection contest
Posted: Fri Oct 18, 2019 12:59 am
by Crow
3yr and 1 yr presumably are the projections using his 3yr and 1 yr RAPM estimates for players.
The blend appears slightly tilted towards more 1 yr weight.
His Twitter says he found a bug that he had to fix.
Re: 2019-20 team win projection contest
Posted: Fri Oct 18, 2019 2:23 am
by mtpowell22
Hey everybody, I created this Win Totals league to track everyone's projections and follow the predictions leaders in-season as things play out. Join here and input your projections:
https://wintotals.com/leagues/b6ca8a6d
Cheers!
Miles
Re: 2019-20 team win projection contest
Posted: Fri Oct 18, 2019 2:50 am
by Rd11490
Mike G wrote: ↑Thu Oct 17, 2019 8:24 pm
Rd11490 wrote: ↑Wed Oct 16, 2019 3:40 am
Using the Blend column for the contest:
...
Edit: edited 10/17/19
Your current predictions bear little resemblance to your original.
What do "3 year" and "1 year" refer to?
Crow is correct, I found a bug in my code that was throwing away all rookie minutes.
3yr is built of 3 yr 4 factors, 1yr is based on 1 yr 4 factors
Re: 2019-20 team win projection contest
Posted: Fri Oct 18, 2019 6:57 am
by Mike G
So without rookie inputs, the Pels were going to win 42 games.
And with that fixed, they're now looking at 38 or 39
Maybe they shouldn't play any rookies?
Re: 2019-20 team win projection contest
Posted: Fri Oct 18, 2019 6:09 pm
by Rd11490
My model treats all rookies exactly the same. I'd say that the pels are one of the many blindspots with this model.
Re: 2019-20 team win projection contest
Posted: Sun Oct 20, 2019 12:54 am
by jgoldstein34
Good luck to everyone this year.
Code: Select all
Team Wins
ATL 27.19
BRK 36.71
BOS 49.36
CHI 39.25
CHO 26.21
CLE 23.13
DAL 45.51
DEN 52.36
DET 32.89
GSW 44.85
HOU 55.88
IND 41.96
LAC 50.15
LAL 49.34
MEM 31.51
MIA 43.41
MIL 59.95
MIN 39.23
NOP 43.64
NYK 26.27
OKC 37.54
ORL 41.74
PHI 53.51
PHO 29.55
POR 45.38
SAC 37.41
SAS 40.72
TOR 47.78
UTA 47.57
WAS 29.99
Re: 2019-20 team win projection contest
Posted: Mon Oct 21, 2019 1:31 am
by eminence
Will update mine tomorrow, been a long day, but is everyone else getting at or near record levels of conference imbalance? I think as of now mine's at 63/37.
Re: 2019-20 team win projection contest
Posted: Mon Oct 21, 2019 1:47 pm
by eminence
Alrighty, here we go. Switching it up a bit this year and going with a simple regression based on player grades (accuracy hopefully similar, just wanted to try something new). Still included my completely arbitrary no ties within conference rule.
Things not considered:
-Injury risk (other than downgrading guys who are currently injured)
-Trade likelihood
-Schedule strength (if I had to guess it'd probably be more accurate to shift the entire East up a win or two and the West down accordingly)
East
Bucks 53
Heat 49
Sixers 48
Raptors 45
Celtics 44
Pacers 42
Magic 41
Nets 38
Pistons 37
Bulls 36
Wizards 32
Hawks 29
Hornets 28
Cavs 24
Knicks 20
West
Warriors 59
Jazz 57
Clippers 56
Rockets 54
Nuggets 53
Lakers 52
Thunder 44
Blazers 42
Mavs 41
Wolves 40
Spurs 37
Pelicans 36
Kings 33
Suns 32
Grizzlies 28
Edit for Zion injury, Pels -3, Bucks/Pacers/Magic all +1.
Re: 2019-20 team win projection contest
Posted: Mon Oct 21, 2019 7:23 pm
by RyanRiot
Trying something new this year where I simulate the whole schedule instead of just making team-level projections and back that into 1230 wins. Also artificially increased the standard deviation of wins because the grouping was way too tight. Hopefully it works out.
Code: Select all
Team Wins
MIL 58.52
HOU 55.81
PHI 54.94
DEN 53.49
LAC 51.78
UTA 50.36
BOS 49.14
LAL 49.03
GSW 47.62
TOR 47.58
MIA 44.22
IND 43.98
ORL 42.95
POR 42.88
DAL 42.13
NOP 41.22
BKN 41.02
SAS 40.16
CHI 37.72
MIN 37.7
DET 36.5
OKC 35.73
SAC 35.3
PHX 30.54
ATL 29.75
WAS 28.8
MEM 28.09
NYK 25.94
CHA 25.23
CLE 21.87
Re: 2019-20 team win projection contest
Posted: Mon Oct 21, 2019 8:46 pm
by eminence
Dang it on Zion, looks like I'll need at least one small update yet this evening. Guess I'm glad it came out today?