Page 9 of 11
Re: 2019-20 team win projection contest
Posted: Thu Dec 05, 2019 11:55 am
by Mike G
A contest suddenly tight
Code: Select all
. avg err rmse
lisp 5.45 7.6
Crow 5.61 8.0
avg. 5.78 7.9
bbst 5.90 7.8
cali 5.92 7.6
trzu 6.00 7.6
emin 6.02 9.7
sndi 6.09 8.5
gold 6.12 7.9
RyRi 6.13 8.1
shad 6.30 8.5
KPel 6.33 8.4
.538 6.70 8.8
vegas 6.72 8.3
dtka 6.75 8.9
ncs. 7.48 9.3
Rd11 7.58 9.6
Mike 8.45 10.2
19re 8.58 10.8
Up to exponent 2.23, liminal space is in 1st; from 2.24 to 4.5 it's tarrazu; 4.6 to 11.9 is caliban
At the 2.2 power, I am out of last place; at e=12 and up, I am the leader: My biggest error is on GS, and everyone else's is bigger.
Current performance projects the Bucks to win 68 games -- 19-3 record with 12.0 SRS -- while the regressed b-r.com system says merely 63.
Dubs looking at 17 or 21 respectively.
Code: Select all
LAL 60 Mil 65
LAC 57 Tor 59
Dal 57 Bos 55
Hou 52 Mia 53
Den 51 Phl 50
Uta 44 Ind 49
Phx 41 Orl 41
Min 38 Brk 38
OKC 38 Det 37
Por 37 Was 33
Sac 34 Chi 31
NOP 32 Cha 28
SAS 32 Cle 26
Mem 26 Atl 24
GSW 19 NYK 22
Re: 2019-20 team win projection contest
Posted: Thu Dec 05, 2019 11:13 pm
by Crow
I might be able to beat lisp. Cavs going 1 - 10 recently is helping with perhaps my biggest gamble.
Re: 2019-20 team win projection contest
Posted: Wed Dec 11, 2019 12:33 pm
by Mike G
Going just by performance to date, with no accounting for future strength of schedule, Crow is in 1st.
But with half of the weight going to b-r.com's projections, which are regressed toward .500 and include coming SOS, it's like this:
Code: Select all
. avg err rmse
lisp 5.79 7.8
Crow 5.88 8.2
cali 5.95 7.9
bbst 6.07 8.0
avg. 6.12 8.3
trzu 6.20 7.8
emin 6.32 9.9
RyRi 6.35 8.4
sndi 6.37 8.8
gold 6.38 8.2
shad 6.48 8.7
vegas 6.74 8.4
KPel 6.87 8.7
.538 6.88 8.9
dtka 7.12 9.2
ncs. 7.79 9.8
Rd11 8.15 10.0
Mike 8.62 10.6
19re 8.88 11.1
In the last 3 weeks, average errors have worsened by a full 1.0 . The bottom 5 individuals are at or very near worst of the season.
As b-r.com relaxes their regression to the mean, teams look like they really are that good or bad:
Code: Select all
LAL 62 Mil 67
Dal 57 Bos 56
LAC 57 Tor 56
Hou 53 Mia 54
Den 50 Phl 53
Uta 43 Ind 47
Phx 43 Orl 40
OKC 41 Brk 40
Por 37 Det 38
Sac 37 Was 32
Min 35 Chi 30
SAS 33 Cha 27
NOP 29 Atl 26
Mem 26 Cle 23
GSW 20 NYK 19
With Griffin back, the Pistons look serious.
Kings and Blazers move up as Wolves sink. Thunder seem to be real.
Re: 2019-20 team win projection contest
Posted: Wed Dec 11, 2019 5:09 pm
by Crow
I like my numbers in the East compared to lisp. He might do better in West but I think my edge in the East will be bigger. Someone might pass us both. I haven't looked close at others yet.
Re: 2019-20 team win projection contest
Posted: Thu Dec 12, 2019 11:28 am
by Mike G
Crow picks up by .23 overnight
Code: Select all
. avg err rmse
Crow 5.65 8.1
lisp 5.69 7.7
cali 5.87 7.8
avg. 6.02 8.2
bbst 6.05 8.0
trzu 6.14 7.7
sndi 6.15 8.7
emin 6.22 9.8
gold 6.24 8.1
RyRi 6.34 8.3
shad 6.43 8.7
vegas 6.69 8.3
KPel 6.76 8.7
.538 6.83 8.9
dtka 7.04 9.1
ncs. 7.67 9.7
Rd11 8.04 9.9
Mike 8.65 10.5
19re 8.85 11.1
Others lead with different exponents:
Code: Select all
. exponent leader
0.01 - 0.34 eminence
0.35 - 1.10 Crow
1.11 - 2.11 liminal space
2.12 - 7.1 tarrazu
7.2 - 11.2 caliban
11.3 - Mike
emin is worst at e > 2.28, and I am worst below that point.
emin also has 7 guesses which are within 1.0 of current projections. Crow is within 2 on an astonishing 15 of 30.
emin within 5 on 19 teams.
EDIT Dec.16
With Sunday's games, lisp avg error jumps by 0.17, while Crow diminishes by the same amount.
Code: Select all
. avg err rmse
Crow 5.53 8.2
lisp 5.86 7.9
cali 5.91 8.0
avg. 6.10 8.3
bbst 6.27 8.2
trzu 6.30 8.0
gold 6.31 8.3
sndi 6.31 8.9
shad 6.36 8.8
RyRi 6.36 8.5
emin 6.52 10.0
KPel 6.55 8.8
vegas 6.61 8.4
.538 7.07 9.1
dtka 7.28 9.4
ncs. 7.83 9.8
Rd11 8.12 9.9
Mike 8.64 10.5
19re 8.82 11.1
Crow's error is his best of the season; worst mark so far for bbstats, dtkavana, and 538.
Separation between top and bottom is almost a season high.
These are currently projected team wins:
Code: Select all
LAL 63 Mil 67
LAC 57 Bos 54
Dal 56 Tor 54
Hou 53 Mia 54
Den 52 Phl 52
Uta 44 Ind 50
Phx 40 Brk 41
OKC 40 Orl 39
Sac 37 Det 39
Por 36 Chi 33
Min 34 Was 30
SAS 32 Cha 29
Mem 29 Cle 24
NOP 28 Atl 24
GSW 18 NYK 21
EDIT Dec 24
It's turning into a rout. Crow's lead over #2 is the largest of the season, as is the difference between top and bottom.
Code: Select all
. avg err rmse
Crow 5.37 7.72
lisp 5.84 7.70
bbst 6.02 8.02
cali 6.07 7.75
avg. 6.23 8.14
emin 6.25 9.76
gold 6.26 7.91
RyRi 6.26 8.18
trzu 6.45 7.73
sndi 6.47 8.66
shad 6.54 8.58
.538 6.71 8.91
vegas 6.75 8.22
KPel 6.84 8.81
dtka 7.22 9.13
ncs. 7.77 9.61
Rd11 8.64 10.0
Mike 8.88 10.6
19re 8.99 10.9
Crow had some pretty radical predictions on a number of teams: highest on 7 and low on 6. How has that worked out?
Ranking the teams by how much over their expectations -- defined as the avg of our guesses -- and Crow's departure from that avg:
Code: Select all
over tm Cro
14 Dal -1
12 Mil 5
10 LAL 1
10 Mia 9
10 Bos 6
9 Tor 7
8 Ind 7
6 LAC -5
5 OKC -5
3 Den 6
2 Cle -9
2 Phx -2
1 Brk 1
1 Cha 1
.3 Phl .4
-1 Hou -4
-1 Was -1
-2 Mem -3
-2 Sac 4
-3 Det -1
-4 Chi -3
-4 Por 4
-5 Orl -2
-5 NYK 3
-5 Uta -4
-7 SAS 2
-8 Min -9
-9 Atl -4
-12 NOP -4
-26 GSW .2
Crow's on the right side on 10 of 15 overachieving teams, and all but 4 of the unders.
Re: 2019-20 team win projection contest
Posted: Wed Dec 25, 2019 1:19 pm
by jgoldstein34
FYI to all, I replicated the sheet someone put together last year to track the contest:
https://docs.google.com/spreadsheets/d/ ... sp=sharing
Re: 2019-20 team win projection contest
Posted: Wed Dec 25, 2019 2:08 pm
by Mike G
Very cool. Does it update daily? What is the source of
current projections?
Re: 2019-20 team win projection contest
Posted: Thu Dec 26, 2019 2:24 pm
by jgoldstein34
Updates daily. Used the sources the sheet last year did so an average of TeamRank, 538 and BBR as well as current wins + average of our preseason projection pace for remaining games.
Re: 2019-20 team win projection contest
Posted: Thu Dec 26, 2019 4:20 pm
by Crow
Every unit of error being equal in weight is how I approached the contest and see the scoring.
Re: 2019-20 team win projection contest
Posted: Thu Dec 26, 2019 11:44 pm
by Mike G
jgoldstein34 wrote: ↑Thu Dec 26, 2019 2:24 pm
Updates daily. Used the sources the sheet last year did so an average of TeamRank, 538 and BBR as well as current wins + average of our preseason projection pace for remaining games.
Does it enhance the accuracy of win projections to suppose the Dubs are gonna win 46 games? or the Mavs should win 42?
Or does this factor diminish over time? At what point in the season is the Celtics' remaining win% more likely to be 75% than 58% ?
Re: 2019-20 team win projection contest
Posted: Sat Dec 28, 2019 1:36 am
by jgoldstein34
Honestly didn't put any thought into it haha. Someone built this last year and used this method, just copied it over to this year and automated it.
Also want to be clear what I mean. So using the Warriors for example, this is how the sites included have them:
538 - 25 wins
BBall Ref - 22.3
TeamRankings - 24.7
The 4th method (current wins + average preseason win projection pace for remaining games) has them at 30.4.
It's not just straight preseason win projection being included. I haven't done any work into if it makes sense to include in the average with the other 3 but it's relatively similar to the more sophisticated methods.
Re: 2019-20 team win projection contest
Posted: Sat Dec 28, 2019 1:11 pm
by Mike G
Crow keeps seeing reduced error, now the lowest anyone's had all year.
Code: Select all
. avg err rmse
Crow 5.16 7.30
lisp 5.57 7.26
bbst 5.70 7.61
cali 5.80 7.26
avg. 5.85 7.66
gold 5.89 7.42
RyRi 5.99 7.69
sndi 6.06 8.18
emin 6.06 9.28
trzu 6.08 7.29
shad 6.16 8.12
.538 6.47 8.47
vegas 6.51 7.78
KPel 6.65 8.40
dtka 6.81 8.64
ncs. 7.41 9.17
Rd11 8.22 9.61
Mike 8.58 10.2
19re 8.69 10.5
... relative to these projected wins:
Code: Select all
LAL 59 Mil 67
LAC 57 Bos 59
Dal 57 Mia 54
Hou 53 Tor 54
Den 53 Phl 52
Uta 46 Ind 50
OKC 42 Brk 39
Por 38 Orl 39
Phx 35 Det 35
SAS 34 Chi 33
Sac 34 Was 30
Min 31 Cha 28
Mem 30 Cle 26
NOP 30 NYK 24
GSW 23 Atl 20
By performance to date, East teams are projected to avg 40.6 wins, vs 41.4 for the West. This is about 1/5 of the disparity we/avg predicted.
There's a .37 correlation between our E/W ratios and our avg error size.
Re: 2019-20 team win projection contest
Posted: Sat Dec 28, 2019 4:47 pm
by Crow
Thanks for the updates.
Re: 2019-20 team win projection contest
Posted: Wed Jan 01, 2020 12:12 pm
by Mike G
new year, new/old leader
Code: Select all
. avg err rmse
lisp 5.49 7.1
Crow 5.51 7.4
gold 5.71 7.3
bbst 5.72 7.4
cali 5.82 7.1
avg. 5.84 7.5
emin 5.87 9.1
RyRi 6.11 7.6
trzu 6.11 7.2
sndi 6.11 8.1
.538 6.19 8.2
shad 6.32 8.1
vegas 6.62 7.8
KPel 6.65 8.4
dtka 6.71 8.4
ncs. 7.05 8.8
Rd11 8.05 9.4
Mike 8.54 10.2
19re 8.69 10.5
Over and under-achieving teams relative to the avg of our predictions:
Code: Select all
over tm pre pro
12.7 Dal 42 55
11.6 Mil 56 67
10.8 LAL 49 60
+9.7 Bos 48 58
+9.1 Mia 44 53
+8.7 Tor 46 55
+6.5 Ind 43 49
+6.4 LAC 51 57
+6.4 OKC 38 44
+2.4 Cle 24 26
+2.2 Phx 34 36
+0.5 Den 51 52
+0.5 Mem 30 31
-0.6 Cha 28 27
-1.1 Phl 52 51
-1.3 Brk 39 38
-1.9 Hou 55 53
-1.9 Uta 50 48
-2.2 Was 32 30
-2.8 NYK 27 24
-3.1 Chi 37 34
-4.0 Sac 36 32
-4.5 Det 37 33
-5.5 Orl 43 37
-5.5 SAS 42 36
-5.8 Por 42 37
-7.1 Min 39 32
-8.1 NOP 41 33
-9.2 Atl 29 20
-23.1 GSW 46 23
Final column is projected wins.
Roughly every 9th day rankings:
Code: Select all
date: N7 16 25 D5 14 23 J1
lisp 1 4 1 1 1 2 1
Crow 14 12 6 2 2 1 2
gold 6 7 2 9 10 6 3
bbst - - 3 4 5 3 4
cali 5 3 7 5 3 4 5
avg. 4 2 4 3 4 5 6
emin 3 5 10 7 8 8 7
RyRi 10 9 8 10 9 7 8
trzu 2 6 5 6 6 10 9
sndi 9 8 9 8 7 9 10
date: N7 16 25 D5 14 23 J1
.538 8 11 14 13 14 13 11
shad 7 10 11 11 11 11 12
vegas 12 14 13 14 12 14 13
KPel - - 12 12 13 12 14
dtka - - 15 15 15 15 15
ncs. 13 13 17 16 16 16 16
Rd11 11 1 16 17 17 17 17
Mike 15 15 18 18 18 18 18
Jan.2 quick update:
Code: Select all
. avg err rmse
Crow 5.50 7.4
lisp 5.57 7.1
gold 5.68 7.3
bbst 5.80 7.5
cali 5.81 7.1
avg. 5.82 7.5
emin 5.96 9.2
RyRi 6.01 7.6
sndi 6.09 8.1
trzu 6.12 7.2
shad 6.26 8.1
.538 6.31 8.3
vegas 6.59 7.8
KPel 6.65 8.4
dtka 6.79 8.4
ncs. 7.05 8.8
Rd11 8.07 9.4
Mike 8.47 10.1
19re 8.68 10.6
And the Spurs somehow find their way back to playoff ranking:
Code: Select all
LAL 60 Mil 67
LAC 57 Bos 58
Dal 55 Tor 55
Hou 53 Mia 53
Den 52 Phl 51
Uta 48 Ind 49
OKC 44 Orl 38
SAS 36 Brk 38
Phx 35 Chi 34
Por 35 Det 33
NOP 33 Was 29
Sac 32 Cha 27
Min 32 Cle 26
Mem 31 NYK 26
GSW 23 Atl 20
SA didn't play last night, but Phx and Por both lost.
Mavs now have the best SRS in the West.
Jan.4 update:
The online table linked by jgoldstein is broke due to Dal having #REF! projected wins and losses. This gives them a .667 chance of making the playoffs and puts me in 1st place!
Meanwhile, back on planet earth:
Code: Select all
. avg err rmse
lisp 5.56 7.1
Crow 5.59 7.5
gold 5.67 7.3
bbst 5.73 7.4
cali 5.75 7.2
emin 5.76 9.1
avg. 5.77 7.5
RyRi 5.93 7.6
trzu 6.03 7.1
sndi 6.06 8.1
shad 6.12 8.1
.538 6.17 8.2
vegas 6.42 7.8
KPel 6.57 8.4
dtka 6.82 8.4
ncs. 6.94 8.8
Rd11 8.05 9.4
Mike 8.44 10.1
19re 8.65 10.6
These are lowest MAE of the season for 538, vegas, and eminence.
Spurs back to #10 seed, Blazers #8
Re: 2019-20 team win projection contest
Posted: Sat Jan 04, 2020 2:50 pm
by eminence
Primary goal is always to beat vegas, secondary is to beat the average, so feeling great about how things are going so far
Entertaining season to date as well, think the race for the 2-6 seeds in each conference will be quite interesting.