Page 5 of 13
Re: 2023-24 team win projection contest
Posted: Tue Oct 31, 2023 8:20 pm
by Crow
Contest leader ahead of all outsiders. Somewhat mixed after that.
Re: 2023-24 team win projection contest
Posted: Thu Nov 02, 2023 12:19 pm
by Mike G
After an 0-3 night vs the West, the median East SOS is -3.9
https://www.basketball-reference.com/fr ... _prob.html
Sixers have played 3 games, everyone else 4 or 5.
Code: Select all
. avg err rmse r^2 . avg err rmse r^2
eWin 5.23 6.65 .11 DRKO 5.11 6.44 .28
ncs. 5.63 7.11 .21 LEBR 5.19 6.59 .21
avgA 5.63 7.19 .21 EBPI 5.34 7.09 .21
trzu 5.93 7.37 .18 4141 5.61 6.61
Crow 5.97 7.44 .20 KPel 5.76 7.42 .11
dtka 5.91 7.73 .15 vegas 6.12 7.63 .22
DQin 6.52 8.21 .18 EExp 6.41 8.00 .23
vzro 6.53 8.51 .16 ChKl 6.66 8.56 .17
emin 6.70 8.62 .25 NuFi 6.95 9.03 .18
medi 7.12 8.68 .21 IanL 7.00 8.59 .24
A couple of 'dummies' have been added: the average of all 9 APBR direct-submissions; and one that predicts all teams go 41-41.
Re: 2023-24 team win projection contest
Posted: Thu Nov 02, 2023 12:24 pm
by v-zero
The 4141 dummy prediction really shows just how little these early standings mean.....or at least I hope so!
Re: 2023-24 team win projection contest
Posted: Thu Nov 02, 2023 1:24 pm
by Mike G
Yeah, kinda. The 41 guess looks like the best guess for a few teams -- Mil, NYK, Mem, Por, Uta -- and worst on none.
At 0-5, the Grizz are thus far the worst team in the west; everyone predicted them to be >41, and a few to be best in the west.
So the first 4-5 games may or may not well predict the remaining 77-78; but within 15 games or so, it historically does. That is, our predictions can be thrown out the window as future projections.
The current b-r.com regression to 41 still improves the error for everyone in the contest; I can de-regress, but it helps nobody's avg error or rmse.
Teams' SOS still seem pretty wack, but should stabilize soon. One blowout score can turn things upside down right now.
Re: 2023-24 team win projection contest
Posted: Thu Nov 02, 2023 10:19 pm
by Crow
41-41 finished last in previous year and it is extremely likely to do so again.
Might look early but that should steadily fade even with moderate amount of parity.
Re: 2023-24 team win projection contest
Posted: Fri Nov 03, 2023 11:34 pm
by Crow
Bref's projected standings currently declare 5 teams have a 98.5+% chance of making playoffs and 5 teams already with 20% or less chance.
Some people might disagree with giving the Magic a 99.5% / virtual lock at making the playoffs and / or assigning the Thunder to that 20% or less group and at only 7.7%. Thunder given very very little chance of seeding above the play-in round.
Until the measuring stick gets pretty solid, the contest standings are pretty speculative.
Re: 2023-24 team win projection contest
Posted: Sat Nov 04, 2023 12:09 pm
by Mike G
Yes, b-r.com has this feature under "frivolities" still; but they also link to it right up top.
The assumption seems to be that teams ARE this good (or bad) after just a few games; and if so, then it's likely they continue as such. And what constitutes a decent sample size? Five games -- or 500 possessions on offense and defense?
I find it interesting to see how the timid predictions fare next to the bold ones, as the season progresses. Pretty sure I'm the most timid this year.
Code: Select all
. avg err rmse r^2 . avg err rmse r^2
eWin 5.27 6.62 .12 DRkO 5.10 6.36 .29
avgA 5.45 7.09 .23 LEBR 5.11 6.52 .23
ncs. 5.52 7.11 .21 EBPI 5.16 7.05 .22
trzu 5.74 7.28 .20 4141 5.54 6.65 #/0!
Crow 5.75 7.42 .20 KPel 5.59 7.37 .12
dtka 5.86 7.68 .15 vegas 5.97 7.54 .24
vzro 6.42 8.49 .17 EExp 6.29 7.93 .24
DQin 6.59 8.10 .19 ChKl 6.74 8.55 .17
emin 6.60 8.26 .30 NuFi 6.89 8.96 .19
medi 6.88 8.58 .22 IanL 7.04 8.61 .24
The Magics are thus far the most overachieving team relative to our prediction avg -- according to b-r.com:
Code: Select all
over tm avgA b-r. under tm avgA b-r.
11.8 Orl 36.3 48.1 -17.2 Cle 49.3 32.1
10.7 SAS 28.7 39.4 -16.0 Mem 44.7 28.6
9.8 Det 26.1 35.9 -10.4 Mil 50.3 39.9
9.7 Por 30.1 39.8 -8.1 NYK 47.2 39.1
9.6 LAC 42.6 52.1 -6.6 Mia 42.4 35.7
5.2 GSW 45.8 50.9 -5.4 Chi 37.7 32.3
5.1 Sac 42.5 47.6 -4.6 NOP 45.2 40.6
4.7 LAL 43.2 47.8 -4.6 Bos 56.5 51.9
4.4 Uta 37.4 41.8 -3.2 Phx 46.9 43.7
2.4 Phl 47.9 50.3 -1.7 OKC 40.3 38.5
2.4 Hou 31.9 34.2 -1.7 Brk 39.9 38.2
2.1 Dal 41.0 43.1 -1.3 Min 46.0 44.7
1.4 Tor 40.4 41.7 -0.6 Den 50.7 50.1
1.1 Atl 42.0 43.1 -0.2 Ind 35.0 34.7
0.8 Cha 32.3 33.1
0.7 Was 29.8 30.5
Re: 2023-24 team win projection contest
Posted: Sat Nov 04, 2023 5:39 pm
by Crow
4 biggest early over-achievers were projected low achievers. 8 biggest underachieving were projected to be strong.
Easiest for this to happen very early but greater parity is something to watch.
Re: 2023-24 team win projection contest
Posted: Sat Nov 04, 2023 10:34 pm
by Crow
BRef projected standings only has 2 over 50 wins and 2 under 30.
Last season it was 6 and 4 respectively.
But I went with 2 and 3 (and may not have thought about it much at overview level), so I should hope for the tight distribution.
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 1:49 am
by Mike G
Orlando's big improvement this year (thru 5 games) is on defense. They're +3 on steals and turnovers per game. They've held 5 opponents to .016 lower eFG% on avg, an improvement of .021 from last year.
Opp. 2FG% is down a whopping .047 from last year.
Unfortunately, their starting center is out of action for a while now.
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 2:57 am
by Crow
Carter may be missed as a part of a productive starting unit, mostly positive big pairs and the rotations leading rebounder on both boards.
His -5 BPM runs counter to that though.
Will want to see replacement and change.
Tonight they went with Bitadze and he got a double double and team best raw +/-.
They smashed the Lakers on the boards and he got 5 blocks.
Only ten previous minutes but it was amazing.
Seems like positive things are more likely to continue than not.
Team is above average on all factors except own efg%.
I did not see this coming but I have barely studied them.
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 11:01 am
by Mike G
Fell on dark o'days.
Code: Select all
. avg err rmse r^2 . avg err rmse r^2
eWin 5.61 7.10 .08 DRKO 4.80 6.18 .34
ncs. 5.65 7.27 .20 LEBR 5.09 6.52 .24
avgA 5.71 7.33 .21 EBPI 5.52 7.30 .20
trzu 5.95 7.52 .18 4141 5.59 6.85
dtka 5.98 7.71 .16 KPel 5.66 7.46 .12
Crow 6.23 7.77 .17 vegas 6.60 7.99 .19
vzro 6.67 8.64 .16 EExp 7.06 8.49 .18
medi 6.73 8.62 .22 ChKl 7.26 9.02 .13
emin 6.86 8.45 .28 NuFi 7.40 9.50 .14
DQin 6.93 8.46 .16 IanL 7.76 9.18 .17
The East had a good night, going 3-1 vs the West. Den>Chi
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 3:54 pm
by Mike G
Noticing some younger teams with big 'overachievment' projections. This is defined as b-r.com vs our Avg predictions.
Team age has a modest negative correlation of -.22
Code: Select all
tm avgA over age over?
OKC 40.3 -2.5 22.7 6.8
Det 26.1 9.7 23.1 6.1
SAS 28.7 10.7 23.7 5.0
Orl 36.3 13.7 24.3 4.0
Por 30.1 10.4 24.6 3.4
Cha 32.3 2.2 25.0 2.7
NYK 47.2 -7.8 25.0 2.7
Ind 35.0 -1.9 25.2 2.4
Mem 44.7 -16.8 25.2 2.4
Hou 31.9 6.7 25.3 2.2
Atl 42.0 4.8 25.6 1.7
Was 29.8 1.2 25.8 1.3
Uta 37.4 0.0 25.8 1.3
NOP 45.2 -7.2 26.0 0.9
Cle 49.3 -17.2 26.1 0.8
Tor 40.4 2.7 26.3 0.4
Dal 41.0 2.4 26.3 0.4
Sac 42.5 -1.3 26.6 -0.1
Den 50.7 0.3 26.7 -0.3
Brk 39.9 -1.0 26.8 -0.5
Min 46.0 3.9 27.5 -1.7
Phl 47.9 4.9 28.3 -3.1
Mia 42.4 -5.4 28.4 -3.3
Bos 56.5 -3.5 28.5 -3.5
LAL 43.2 0.7 28.5 -3.5
Chi 37.7 -6.8 28.8 -4.0
Phx 46.9 -5.2 29.1 -4.6
GSW 45.8 2.7 30.0 -6.2
Mil 50.3 -9.0 30.2 -6.5
LAC 42.6 8.6 30.5 -7.1
Clipps are the oldest team but are doing better than expected. And they haven't been bitten by injury yet.
I always think my predictions should include an age adjustment, but I again failed to do this. Final column is from a formula that 'predicts' overachievment based on this table:
over? = (26.7-Age)*1.78 - .30
https://www.basketball-reference.com/le ... _2024.html
Age is under
Advanced Stats
NBA avg court-time age is 26.6 this year.
I'd like to run this a couple more times during the season, esp. before tanking season starts.
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 5:31 pm
by Crow
"Final column is from a formula that 'predicts' overachievment based on this table:
over = (26.7-Age)*1.78 - .30."
Could you walk thru this?
Predicts overachievement? Compared to what? Positives and negative mean what here?
"Clippers are the oldest team but are doing better than expected."
In the second column. But what is the narrative for the 4th column for them? And for OKC?
Re: 2023-24 team win projection contest
Posted: Sun Nov 05, 2023 9:41 pm
by Mike G
OKC avg playing age this year is 22.7
26.7-22.7 = 4.0
4.0*1.78 = 7.12
7.12 - .30 = 6.82
So the formula predicts that they should do 6.8 wins better than they would based on last year's stats.
A straight-line proportion like this is probably not the best estimator; but it was simple.
An additional 7 wins seems like a lot, and they fall well short; but the next 4 teams significantly beat the formula, for now -- as do several others.