Page 8 of 13
Re: 2017-18 team win projection contest
Posted: Mon Nov 13, 2017 3:41 pm
by shadow
Standings based on updated win projections derived from my power rankings:
Code: Select all
Entry Avg Error
shad 4.96
sbs. 5.39
RyRi 5.46
gold 5.49
538_ 5.51
kmed 5.63
Josh 5.65
sndi 5.76
cali 5.79
lnqi 5.81
AJBk 5.83
GK5. 6.08
Crow 6.11
Nath 6.28
ATCt 6.66
Mike 6.68
emin 6.89
ncs. 6.91
knar 7.28
FWIW, 4.96 would have only beat 2 entries (538 & Regressed-Pythag) at the end of last year in average error.
Re: 2017-18 team win projection contest
Posted: Mon Nov 13, 2017 5:36 pm
by Mike G
kmedved wrote:As you'd expect with square errors, a lot of Shadow's lead is a function of being higher on big outperformers like Detroit, Boston, and Memphis..
Except that his unsquared-error lead is bigger than his squared error lead. At least relative to b-r.com projections.
Last year on Nov. 22, our errors ranged from 5.08 to 6.28
So we are actually better than at the same time last year.
But on Nov 26, the range was 4.32 to 5.27
On Nov 29 we were 3.99 to 4.98
Dec 3 -- 4.55 to 5.53
Dec 8 -- 3.89 to 5.26
...
http://www.apbr.org/metrics/viewtopic.php?f=2&t=9188
Re: 2017-18 team win projection contest
Posted: Thu Nov 16, 2017 12:47 pm
by Mike G
Before last night's games, both Atl and Sac were projected 26-win teams.
Now Atl looks like 30 W and Sac is at 22. A 45-pt blowout at this point changes your MOV by 3 or so.
Code: Select all
_avg + err RMSE avg + err RMSE
shad 5.3 6.4 sndi 6.1 7.2
gold 5.5 6.3 Josh 6.3 7.3
sbs. 5.7 6.7 Crow 6.4 7.5
RyRi 5.7 6.7 GK5. 6.5 7.5
kmed 5.8 7.2 Nath 6.5 7.6
538_ 5.8 7.0 ATCt 6.9 7.9
AJBk 5.8 7.0 Mike 7.0 8.3
vegas 5.9 7.0 emin 7.2 8.4
cali 6.1 7.2 ncs. 7.2 8.7
lnqi 6.1 7.1 knar 7.6 8.8
Everyone looks about .20 worse, due to that one gnarly game.
Re: 2017-18 team win projection contest
Posted: Sun Nov 19, 2017 11:36 am
by Mike G
Yesterday, Orlando was projecting to exceed everyone's expectations by at least 7 wins, and an avg of 11.
And Utah was running at least 4, avg 8 wins short of our guesses. With a 40 point margin, both teams are "corrected" by 4 wins.
Here's a summary of yesterday's b-r.com projections and today's, for last night's contestants.
Code: Select all
tm Pts 11-18 11-19 diff
Uta 125 34.5 38.5 4.0
Orl 85 43.8 39.8 -4.0
Dal 111 24.5 27.7 3.2
Mil 79 43.1 39.4 -3.7
Hou 105 53.7 55.7 2.0
Mem 83 43.9 42.3 -1.6
Cha 102 38.2 39.2 1.0
LAC 87 39.9 38.4 -1.5
Por 102 45.1 44.5 -0.6
Sac 90 23.9 23.8 -0.1
Bos 110 56.7 56.8 0.1
Atl 99 30.3 30.1 -0.2
GSW 124 59.6 60.0 0.4
Phl 116 43.0 43.9 0.9
The changes aren't symmetrical in most cases, at this time of year, because previous-opponent strengths are being recalibrated.
Both the Sixers and the Dubs look better today, and we presume it's because both were under-rated prior to the game.
Blazers and Kings both look worse for their efforts -- esp. the winners!
Some teams which did not play had their forecast change more than some who notched a win or a loss:
Code: Select all
dnp 11-18 11-19 diff
Min 44.2 45.0 +0.8
Was 45.3 45.9 +0.6
Tor 53.2 53.6 +0.4
...
Det 47.9 47.6 -0.3
Cle 39.5 39.1 -0.4
NOP 38.5 38.1 -0.4
SAS 46.2 45.7 -0.5
After all this, our little race has tightened some, and we may say there are co-leaders:
Code: Select all
avg abs err RMSE avg abs err RMSE
shad 5.05 6.11 sndi 5.69 6.67
gold 5.12 6.07 Josh 5.80 6.79
RyRi 5.29 6.31 GK5. 5.94 7.07
538_ 5.38 6.63 Nath 5.96 7.16
sbs. 5.39 6.31 Crow 6.10 7.39
kmed 5.39 6.73 ATCt 6.54 7.69
AJBk 5.48 6.69 Mike 6.63 7.80
vegas 5.53 6.61 emin 6.68 7.69
lnqi 5.56 6.56 ncs. 6.73 8.41
cali 5.65 6.72 knar 7.31 8.40
Update 11-21
Code: Select all
avg abs err RMSE avg abs err RMSE
shad 5.54 6.71 lnqi 6.20 7.27
gold 5.63 6.70 Josh 6.25 7.39
sbs. 5.86 6.93 Nath 6.49 7.62
RyRi 5.89 6.93 GK5. 6.62 7.77
538_ 5.94 7.16 Crow 6.64 7.85
vegas 5.98 7.18 Mike 7.11 8.13
kmed 6.02 7.38 ATCt 7.12 8.37
cali 6.16 7.44 emin 7.13 8.27
AJBk 6.16 7.21 ncs. 7.40 8.89
sndi 6.18 7.20 knar 7.81 8.88
We've all taken a beating.
Update 11-26:
Code: Select all
avg abs err RMSE avg abs err RMSE
shad 5.46 6.62 Josh 6.01 7.19
sndi 5.62 6.83 Nath 6.05 7.37
vegas 5.68 6.87 AJBk 6.09 7.17
kmed 5.69 7.23 Crow 6.37 7.71
gold 5.69 6.86 GK5. 6.53 7.80
cali 5.77 7.12 emin 6.72 7.80
sbs. 5.79 6.82 Mike 6.76 7.79
RyRi 5.87 6.87 ATCt 6.81 8.25
lnqi 5.90 6.94 ncs. 7.13 8.75
538_ 5.99 7.15 knar 7.29 8.55
Re: 2017-18 team win projection contest
Posted: Mon Nov 27, 2017 3:43 pm
by Mike G
Here are teams ranked by how much better they are doing, than what we expected.
On the left are the over- and under-achievers, teams which are projecting entirely above or below all our guesses.
On the right, the middle has been pulled aside: These teams affect your avg error
relative to everyone else's.
Current projection = avg + off
Code: Select all
tm off avg tm off avg
Phl 14 36 Por 4 41
Ind 13 32 Orl 3 33
NYK 12 30 Mem 2 36
Det 10 36 Atl 1 27
Bos 10 48 Hou 1 58
LAL 7 28 NOP 0 42
Tor 6 48 Sac -2 26
Brk 5 29 Dal -2 33
. . . Den -3 47
Chi -5 26 Mia -3 43
Min -6 48 Was -3 48
Cha -6 45 Phx -3 27
GSW -7 67 Uta -3 43
Okl -8 51
SAS -8 53
Mil -8 46
LAC -10 48
Cle -11 55
So, when the Cavs win, or the Sixers lose, everyone's avg-err improves; and nothing changes in the contest.
With any win or loss by a team on the right, your place in the race is affected.
Re: 2017-18 team win projection contest
Posted: Tue Nov 28, 2017 10:41 am
by Mike G
Improvements by everyone in 2 days, and a general tightening:
Code: Select all
avg abs err RMSE avg abs err RMSE
shad 5.29 6.36 Nath 5.83 7.14
sndi 5.36 6.57 lnqi 5.91 6.76
vegas 5.51 6.71 AJBk 5.91 6.99
sbs. 5.54 6.59 Crow 6.10 7.49
kmed 5.58 7.05 GK5. 6.21 7.49
RyRi 5.61 6.61 Mike 6.39 7.45
Josh 5.68 6.86 emin 6.47 7.59
gold 5.70 6.77 ATCt 6.61 7.93
cali 5.74 6.94 ncs. 6.78 8.40
538_ 5.81 6.90 knar 7.06 8.28
Re: 2017-18 team win projection contest
Posted: Tue Nov 28, 2017 1:43 pm
by caliban
Not much change since the 10 game update. Very compact field. shadow still the boards best hope against Vegas.
Re: 2017-18 team win projection contest
Posted: Tue Nov 28, 2017 1:50 pm
by Mike G
You can get the same as (RMSE+MAE)/2 by taking the 1.45 root of the sum of the 1.45 power of all errors.
Re: 2017-18 team win projection contest
Posted: Sat Dec 02, 2017 12:07 pm
by Mike G
Whoa
Code: Select all
avg abs err RMSE avg abs err RMSE
sndi 5.27 6.53 Nath 5.83 7.10
gold 5.56 6.74 AJBk 5.92 7.19
cali 5.56 7.06 Josh 5.99 7.10
kmed 5.65 7.09 emin 6.32 7.46
RyRi 5.67 6.72 GK5. 6.37 7.70
shad 5.67 6.55 Crow 6.37 7.66
538_ 5.72 6.99 ATCt 6.69 8.06
sbs. 5.74 6.70 Mike 6.71 7.67
lnqi 5.76 6.76 knar 7.09 8.28
vegas 5.83 6.84 ncs. 7.12 8.53
sndesai now looking best (or tied) on Brk, Cha, Chi, and SAS; worst on none.
shadow still in virtual tie for the lead in RMSE. He went low on Hou, high on Mem.
Dec. 5 update:
Code: Select all
avg abs err RMSE avg abs err RMSE
sndi 5.11 6.34 vegas 5.79 6.75
cali 5.48 6.86 AJBk 5.84 7.18
538_ 5.51 6.79 Nath 5.89 7.09
sbs. 5.56 6.56 GK5. 6.09 7.47
shad 5.57 6.49 emin 6.37 7.31
RyRi 5.58 6.63 Crow 6.47 7.67
gold 5.63 6.74 ATCt 6.51 7.86
Josh 5.64 6.89 Mike 6.53 7.36
lnqi 5.66 6.63 knar 6.96 8.12
kmed 5.74 6.90 ncs. 7.04 8.38
Desai now nailing Uta and Hou, as well as the 4 named above.
Update, Dec. 9
Code: Select all
avg abs err RMSE avg abs err RMSE
sndi 4.96 6.28 Nath 5.66 6.95
cali 5.30 6.88 AJBk 5.68 7.17
RyRi 5.40 6.56 kmed 5.72 6.87
sbs. 5.47 6.53 GK5. 5.97 7.46
shad 5.48 6.48 emin 6.07 7.14
538_ 5.48 6.87 Mike 6.37 7.29
gold 5.48 6.72 Crow 6.41 7.68
lnqi 5.53 6.59 ATCt 6.44 7.89
vegas 5.55 6.70 knar 6.81 8.10
Josh 5.65 6.88 ncs. 6.84 8.30
Separating east from west teams, ranked by their over-projection, relative to our average prediction.
We guessed the Pacers would win 32.2, but they're headed for 45.8, so we're
off by almost 14, on avg.
Code: Select all
West off East off
LAL 5.1 Ind 13.6
Hou 3.5 NYK 11.6
Uta 2.0 Det 9.6
Por 1.0 Phl 9.3
Sac -0.6 Bos 9.0
Mem -0.6 Brk 7.5
Dal -0.8 Tor 7.3
Phx -1.8 Orl 2.3
NOP -2.0 Atl 0.1
SAS -4.3 Was -3.0
Min -4.4 Mil -3.6
Den -5.5 Chi -6.0
GSW -6.0 Mia -6.3
Okl -8.0 Cle -6.8
LAC -14.5 Cha -8.4
tot -36.7 tot 36.4
Lakers most surprising in the West, would only rank 8th in the East.
East still + 5 in inter-conference games.
Update Dec 10 -- improvements as much as .32 overnight!
Code: Select all
avg abs err RMSE avg abs err RMSE
sndi 4.82 6.14 AJBk 5.49 6.95
RyRi 5.13 6.34 Josh 5.58 6.75
cali 5.17 6.67 kmed 5.59 6.73
sbs. 5.21 6.28 GK5. 5.87 7.25
shad 5.25 6.23 emin 5.96 6.94
gold 5.29 6.55 Mike 6.09 7.13
lnqi 5.32 6.36 Crow 6.20 7.45
vegas 5.33 6.42 ATCt 6.25 7.68
Nath 5.40 6.73 ncs. 6.52 8.07
538_ 5.40 6.73 knar 6.55 7.86
Re: 2017-18 team win projection contest
Posted: Sun Dec 10, 2017 11:32 pm
by shadow
Too late to vote for RMSE as the official metric?
Re: 2017-18 team win projection contest
Posted: Mon Dec 11, 2017 2:20 am
by Mike G
shadow would be in the lead with any exponent between 2.4 and 7.6
Above 7.6, it's eminence.
Below 0.12 it goes to Nathan
sndesai leads with exponent between .13 and 2.3
Re: 2017-18 team win projection contest
Posted: Tue Dec 12, 2017 11:03 am
by Mike G
General improvement and tightening.
Code: Select all
avg abs err RMSE avg abs err RMSE
sndi 4.68 5.92 AJBk 5.18 6.47
RyRi 4.87 5.97 Josh 5.29 6.43
cali 4.93 6.34 kmed 5.33 6.41
sbs. 4.93 5.99 GK5. 5.58 6.87
shad 5.00 6.04 emin 5.74 6.72
538_ 5.02 6.42 Mike 5.83 6.84
lnqi 5.07 6.06 Crow 5.88 7.16
gold 5.07 6.36 ATCt 5.97 7.32
Nath 5.12 6.53 knar 6.14 7.49
vegas 5.17 6.29 ncs. 6.20 7.66
Improvements in 2 days range from .14 (sndi) to .41 (knarsu), avg .28
Re: 2017-18 team win projection contest
Posted: Tue Dec 12, 2017 1:50 pm
by nbacouchside
gonna be bringing up the rear the whole year, it seems
Re: 2017-18 team win projection contest
Posted: Wed Dec 13, 2017 11:24 am
by Mike G
For sure -- provided the Cavs' surge runs out, the Pistons dive is contained, the Celts maintain their inordinate high, and the Lakers and Sixers keep outdoing themselves.
If any of those don't happen -- or if Atl goes south, Cha gets it together, Chi keeps improving, Mil figures it out, Curry comes back, Min makes a playoff run, NOP flex their muscles, Por can't keep it up, Kawhi IS back -- so many ways things get better.
I sort of don't see the Clipps righting their ship; otherwise, most developments should be favorable.
Re: 2017-18 team win projection contest
Posted: Tue Dec 19, 2017 5:18 am
by Crow
Not surprising to me, all my big misses are in the east except for the Clips. That has happened before.