Home for all your discussion of basketball statistical analysis.
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Sun Jan 08, 2017 2:14 pm
BasketDork wrote: Am I seeing this correctly ? This latest has me in the lead ?
Yes, well you may pick your projection and then pick your exponent.
Your avg error from b-r.com is 3.71, and from 538 it's 3.77 . Yet everyone else is even more worse at 538.
Being that we are at the top of a page, here are today's b-r.com projections vs our gang
Code: Select all
. RMSE avg abs err SMRE
RyRi 4.84 lnqi 3.61 nrfo 2.84
lnqi 4.92 Mike 3.67 lnqi 2.85
amp5 4.99 RyRi 3.67 kmed 2.88
trzu 5.01 BaDo 3.68 BaDo 2.90
sbs. 5.09 sbs. 3.76 sbs. 2.91
- nrfo 3.77 Mike 2.92
BaDo 5.21 amp5 3.78 amp5 2.98
Mike 5.24 kmed 3.85 RyRi 2.99
ncsD 5.28 trzu 3.94
vegas 5.28
ncsB 5.28 shad 4.04 trzu 3.19
ATCt 5.30 vegas 4.09 Crow 3.25
shad 5.35 Crow 4.15 jg34 3.29
taco 5.37 ATCt 4.16 shad 3.35
nrfo 5.45 jg34 4.19
sndi 5.47 vegas 3.52
kmed 5.50 ncsD 4.30 ATCt 3.54
Crow 5.51 sndi 4.30 ncsD 3.54
- taco 4.31 sndi 3.63
cali 5.61 ncsB 4.36 cali 3.64
jg34 5.66 cali 4.36 taco 3.71
- yoop 3.75
-
- yoop 4.61 ncsB 3.86
-
yoop 6.08
GK5. 6.14
- 538 4.92 538 4.17
538 6.28 GK5. 5.04 GK5. 4.34
http://www.basketball-reference.com/fri ... f_prob.cgi
3 exponents, 3 leaders
BasketDork
Posts: 201 Joined: Thu Dec 04, 2014 12:58 pm
Post
by BasketDork Sun Jan 08, 2017 7:50 pm
So if I'm to understand, I've made less errors than most, but my errors are also of a larger discrepancy. When I hit, I'm almost dead on, but when I miss, it's pretty substantial.
The Bearded Geek
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Mon Jan 09, 2017 12:57 pm
You're hardly an extreme case; your errors aren't especially large, and you have more than your share of very close predictions.
Today's table -- Notice I'm just worse than you in RMSE, then better when the exponent is reduced. So my errors are both fewer
and worser than yours.
Code: Select all
. RMSE avg abs err SMRE
RyRi 4.94 lnqi 3.61 lnqi 2.77
lnqi 4.99 Mike 3.69
trzu 5.03 BaDo 3.70 Mike 2.93
amp5 5.07 RyRi 3.74 BaDo 2.94
- nrfo 3.81 nrfo 2.94
sbs. 5.17 amp5 3.82 amp5 2.99
BaDo 5.23 sbs. 3.87 RyRi 3.03
Mike 5.28 kmed 3.06
vegas 5.31 kmed 3.96 sbs. 3.14
ATCt 5.36 trzu 3.97
ncsD 5.39 jg34 3.26
shad 5.42 vegas 4.06 trzu 3.27
ncsB 5.45 shad 4.13 Crow 3.30
nrfo 5.47 Crow 4.18
taco 5.48 ATCt 4.19 shad 3.42
Crow 5.50 jg34 4.22 vegas 3.47
sndi 5.51 sndi 4.28 ATCt 3.50
kmed 5.57 taco 4.36 sndi 3.56
- ncsD 4.40 cali 3.66
cali 5.68 cali 4.41 taco 3.72
jg34 5.74 ncsB 4.47 ncsD 3.72
- yoop 3.75
- yoop 4.65
- ncsB 3.93
yoop 6.13 ...
-
GK5. 6.24 538 5.02 538 4.29
- ...
538 6.37 GK5. 5.18 GK5. 4.54Everyone would either slope up or down, except that we're arranged relative to one another. That is, others displace you by being more 'extreme' one way or the other.
RyRi is still avoiding big misses but tends toward average in the close ones.
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Wed Jan 11, 2017 12:36 pm
Yesterday and today, just the top tens
Code: Select all
1-10 RMSE avg abs err SMRE
RyRi 4.91 Mike 3.60 Mike 2.78
lnqi 4.95 lnqi 3.61 BaDo 2.79
trzu 4.99 BaDo 3.62 lnqi 2.87
amp5 5.06 RyRi 3.69 kmed 2.94
sbs. 5.14 sbs. 3.76 sbs. 2.98
BaDo 5.22 nrfo 3.81 nrfo 2.98
Mike 5.23 amp5 3.81 RyRi 3.00
vegas 5.30 kmed 3.87 amp5 3.06
ATCt 5.32 shad 4.03 jg34 3.21
ncsD 5.36 trzu 4.03 shad 3.24
1-11 RMSE avg abs err SMRE
RyRi 4.91 BaDo 3.52 BaDo 2.69
trzu 4.91 lnqi 3.53 RyRi 2.79
lnqi 4.92 RyRi 3.61 Mike 2.79
amp5 5.05 Mike 3.62 lnqi 2.79
sbs. 5.14 sbs. 3.70 sbs. 2.80
BaDo 5.19 nrfo 3.80 nrfo 2.94
Mike 5.30 amp5 3.86 kmed 3.01
vegas 5.31 trzu 3.88 trzu 3.17
ATCt 5.33 kmed 3.92 amp5 3.18
ncsD 5.34 shad 4.00 jg34 3.21
BasketDork
Posts: 201 Joined: Thu Dec 04, 2014 12:58 pm
Post
by BasketDork Thu Jan 12, 2017 1:19 am
Pleasantly surprised to say the least. It's still anybody's competition at this point. Alot of entries clustered pretty close together.
The Bearded Geek
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Thu Jan 12, 2017 11:56 am
Code: Select all
RyRi 4.70 lnqi 3.39 BaDo 2.62
lnqi 4.70 BaDo 3.40 lnqi 2.69
trzu 4.72 Mike 3.44 Mike 2.72
- kmed 2.79
amp5 4.82 RyRi 3.55 sbs. 2.82
sbs. 4.90 sbs. 3.62 RyRi 2.89
- amp5 3.70 jg34 2.98
Mike 5.00 kmed 3.71 nrfo 3.01
BaDo 5.04 nrfo 3.80 amp5 3.04
vegas 5.09 trzu 3.83
ATCt 5.12 jg34 3.90 shad 3.17
ncsD 5.14
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Fri Jan 13, 2017 11:02 am
We now have a supreme leader at every level. RyRi roars back in the lower exponents.
Code: Select all
exp=2 RMSE e=1 abs err exp=.5 SMRE
lnqi 4.59 lnqi 3.33 lnqi 2.66
RyRi 4.61 RyRi 2.74
trzu 4.63 RyRi 3.43 BaDo 2.77
amp5 4.71 BaDo 3.46 Mike 2.83
- Mike 3.54 jg34 2.88
sbs. 4.86 amp5 3.59 sbs. 2.90
- sbs. 3.64 shad 2.92
BaDo 4.99 trzu 3.71 kmed 2.94
vegas 5.03 kmed 3.77 amp5 2.95
Mike 5.03 shad 3.80
ATCt 5.04 nrfo 3.82 nrfo 3.08
ncsD 5.05 vegas 3.82 trzu 3.09
sndi 5.10 jg34 3.86
- ATCt 3.89 vegas 3.20
ncsB 5.22 ATCt 3.25
kmed 5.22 sndi 3.99 Crow 3.31
taco 5.25 Crow 4.04 taco 3.34
shad 5.28 ncsD 4.07 sndi 3.34
nrfo 5.31 taco 4.08 cali 3.40
Crow 5.33 ncsD 3.42
jg34 5.41 cali 4.19
cali 5.44 yoop 3.56
- ncsB 4.33
- yoop 4.38
- ncsB 3.80
yoop 5.83 ... ...
GK5. 5.91
- 538 4.78 538 4.13
538 6.01 GK5. 4.84 GK5. 4.23
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Sat Jan 14, 2017 10:29 pm
Code: Select all
1/14 RMSE avg err SMRE
lnqi 4.59 lnqi 3.25 lnqi 2.42
RyRi 4.67 ....
trzu 4.69
amp5 4.77 RyRi 3.53 RyRi 2.88
- amp5 3.60 sbs. 2.91
sbs. 4.91 BaDo 3.66 amp5 2.91
- sbs. 3.68 Mike 2.93
vegas 5.03 Mike 3.70 BaDo 2.96
ATCt 5.05 trzu 3.71 shad 2.98
ncsD 5.06 trzu 3.06
Mike 5.10 vegas 3.82 jg34 3.09
sndi 5.15 shad 3.88 kmed 3.11
BaDo 5.15 kmed 3.89 taco 3.14
Code: Select all
1/15 RMSE avg err SMRE
lnqi 4.58 lnqi 3.20 lnqi 2.29
trzu 4.61 - - - - -
RyRi 4.64 BaDo 3.49 BaDo 2.69
- RyRi 3.54
amp5 4.77 sbs. 3.61 sbs. 2.83
sbs. 4.85 Mike 3.64 Mike 2.90
- trzu 3.65 RyRi 2.94
vegas 4.98 amp5 3.69 trzu 3.03
ATCt 5.03 shad 3.03
Mike 5.06 vegas 3.81 nrfo 3.06
ncsD 5.06 kmed 3.85 taco 3.07
BaDo 5.06 shad 3.87 jg34 3.08Code: Select all
1/16 RMSE avg err SMRE
lnqi 4.55 lnqi 3.27 lnqi 2.46
trzu 4.60 BaDo 3.35 BaDo 2.50
RyRi 4.60 - -
- Mike 3.52 Mike 2.72
amp5 4.70 RyRi 3.56 - -
- sbs. 3.64 kmed 2.94
sbs. 4.81 trzu 3.68 sbs. 2.96
- amp5 3.71 RyRi 2.98
BaDo 4.92 kmed 3.73 shad 2.99
vegas 4.97 vegas 3.80 nrfo 3.02
ATCt 4.99 nrfo 3.84
Mike 5.02 shad 3.85 trzu 3.14
ncsD 5.03 Crow 3.87 jg34 3.14
sndi 5.11 amp5 3.14
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Tue Jan 17, 2017 11:54 am
Code: Select all
lnqi 4.51 BaDo 3.31 BaDo 2.39
trzu 4.53 lnqi 3.34 . .
RyRi 4.60
amp5 4.67 lnqi 2.72
- Mike 3.58 Mike 2.79
sbs. 4.79 trzu 3.60 kmed 2.87
BaDo 4.88 RyRi 3.64 sbs. 2.95
ATCt 4.92 sbs. 3.67 shad 3.02
vegas 4.94 amp5 3.68 Crow 3.06
ncsD 4.96 kmed 3.73 trzu 3.06
Mike 4.99 vegas 3.73 amp5 3.06
Crow 5.03 Crow 3.79 vegas 3.09
sndi 5.07 RyRi 3.11
kmed 5.13 shad 3.93 nrfo 3.19
taco 5.14 sndi 3.94 sndi 3.20
ncsB 5.18 nrfo 3.97 jg34 3.20
- ATCt 3.98
shad 5.29 jg34 4.02 cali 3.32
nrfo 5.30 taco 4.08 taco 3.33
cali 5.33 cali 4.11 ATCt 3.37
jg34 5.42 ncsD 4.12
- yoop 3.51
- yoop 4.27 ncsD 3.58
- ncsB 4.34 - -
GK5. 5.72 . . ncsB 3.83
yoop 5.73
- GK5. 4.70 GK5. 3.97
538 5.92 538 4.71 538 3.99
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Wed Jan 18, 2017 2:10 pm
It's unlikely these will be updated every day forever, but right now the daily changes are kind of fascinating.
Code: Select all
1-18 RMSE avg error SMRE
lnqi 4.32 lnqi 3.22 BaDo 2.40
trzu 4.37 BaDo 3.27 lnqi 2.63
RyRi 4.43 trzu 3.48 Mike 2.73
amp5 4.49 RyRi 3.52 sbs. 2.77
sbs. 4.64 Mike 3.53 kmed 2.88
BaDo 4.72 sbs. 3.54 shad 2.89
vegas 4.76 amp5 3.55 vegas 2.92
ATCt 4.78 vegas 3.59 trzu 2.95
ncsD 4.81 kmed 3.68 amp5 2.95
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Wed Jan 18, 2017 7:45 pm
Thanks for that. He'll appear in future renditions, but he's not in the top 10, except in RMSE.
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Thu Jan 19, 2017 1:53 pm
As you were ...
Code: Select all
1/19 RMSE avg abs err SMRE
lnqi 4.29 lnqi 3.13 lnqi 2.52
trzu 4.39 BaDo 3.33 BaDo 2.52
RyRi 4.40 amp5 3.40 trzu 2.76
amp5 4.42 trzu 3.42 vegas 2.76
sbs. 4.63 RyRi 3.43 amp5 2.79
AnJo 4.68 vegas 3.46 sbs. 2.81
vegas 4.72 sbs. 3.53 shad 2.81
BaDo 4.75 Mike 3.55 RyRi 2.88
ATCt 4.79 kmed 3.71 Mike 2.90
ncsD 4.83 sndi 3.74 kmed 2.94
Errors are at new lows currently. The progressive bests that have been recorded here:
Code: Select all
Date err Date RMSE Date SMRE
11/17 5.19 Mike 11/17 6.42 Mike 11/20 4.34 kmed
11/22 5.08 Mike
11/26 4.32 Mike
11/29 3.99 RyRi 12/03 5.54 Mike 12/03 3.80 RyRi
12/08 3.89 RyRi 12/13 4.94 RyRi
12/18 3.53 RyRi 12/21 4.59 RyRi
12/27 3.45 RyRi 12/27 4.49 RyRi
12/28 3.43 RyRi
12/31 3.24 RyRi
1/02 3.22 RyRi 1/08 2.84 nrfo
1/04 3.20 RyRi 1/09 2.77 lnqi
- 1/11 2.69 BaDo
- 1/12 2.62 BaDo
- 1/14 2.42 lnqi
1/15 3.20 lnqi 1/18 4.32 lnqi 1/15 2.29 lnqi
1/19 3.13 lnqi 1/19 4.29 lnqi
All but a handful of team projections now fall within the spread of our predictions.
Biggest outliers -- teams entirely above or below our guesses -- and our closest approach:
Code: Select all
tm min error
Hou 8 - yooper
Mil 7 - terrazu
Por 5 - yooper
Dal 4 - yooper
Tor 3 - ( 6 )
Mem 1 - Crow
Mia 1 - nrfo, Mike
SAS 1 - nrfo, cali, MikeRounded to nearest integer.
BasketDork
Posts: 201 Joined: Thu Dec 04, 2014 12:58 pm
Post
by BasketDork Fri Jan 20, 2017 4:12 am
Maybe I should just revert back to past year's contest for my answer, but is our winner based on average absolute error or RMSE ? Is working with a blend of the 3 redundant ?
The Bearded Geek
Mike G
Posts: 6144 Joined: Fri Apr 15, 2011 12:02 amLocation: Asheville, NC
Post
by Mike G Fri Jan 20, 2017 9:51 am
The prize money is split into 4ths, 1/4 going to each category winner and the person making the most posts.