APBRmetrics

Posted: **Sat Feb 16, 2019 3:00 pm**

I missed big on the Bucks. More than some but everybody missed at least pretty big. Missed big on Jazz. May improve but injury to Rubio, terrible start by Mitchell and poor rotation management by Synder (at least early, haven't checked recently) explains. Rockets hurt by Paul injury and loss of depth and defense hurt more than expected by most. Didnt see Kings fast rise coming. Underestimated the coach and young guys playing for team. Orlando has won more than I expected but probably ends up closer to my expectation. Was too high of Wizards. Didnt think highly of them but didnt see a lot of reason to expect big change. Wall mediocrity and injury was a lot of it but Brooks coaching and loss of leadership control was a lot of it too. 6 big misses. The leaders kept big misses (8 wins plus or minus) to 3. I was off a little more generally. One leader was within 2 wins on 19 teams. I had too many misses by 3-5. More exact hits than anybody I think. But just 4 of those didnt help much. Being within 1 point of the lead on average error is ok to me given limited time devoted and some gambling to try to stand out. It is a big and generally tough field. Bigger and tougher than it used to be. More using RAPM, more blending probably. I used to be have less company on doing those things.

Posted: **Sat Feb 16, 2019 7:17 pm**

Seems like everybody missed on Bucks/Kings.

Teams I'm noticeably different from the average on:

Hits: Pacers (Dipo injury could blow it up after the AS break), Wolves (maybe factored in team chemistry more than most?), Bulls (seemed like a clear worst in league contender to me, not sure why the avg was so high)

Misses: Rockets, Spurs (should've had more faith), Nets, Pelicans (Damn it AD), Magic (I still kinda expect them to tank it)

Overall quite happy with my guesses so far this season, not surprised I'm notably higher in MAE than RMSE, built to minimize MAE. Surprising I'm so close to avg with so many more big misses than big hits, must be eking out a lot of small edges.

Wouldn't expect a late season run this year unfortunately.

Posted: **Sat Feb 16, 2019 7:37 pm**

Top 4 and within a half point of lead for you is still strong. We've seen leader changes this season and late in previous years.

Posted: **Sat Feb 16, 2019 7:42 pm**

Best and worst predictions for each team. The +/- column is how much better or worse the team is doing (via b-r.com) than the avg error from our 16 or so projectionists.

Code: Select all

East +/-   Best               Worst
Atl  +1   0 emin,Crow,bbst    6 Rd11
Bos  -2   0 Cica              5 shad
Brk  +7   3 KPel             12 Crow
Cha   0   0 AnJo,KPel        10 Cica
Chi  -5   1 Cica             13 J-TF
Cle -15   7 J-TF             21 Rd11
Det   0   0 emin,shad,J-TF    8 Rd11
Ind  +6   0 Cica              8 Rd11
Mia  -4   1 Cica,J-TF         7 Rd,KP
Mil +14   6 Rd11             18 gold
NYK  -6   1 kmed             12 KPel
Orl  +8   3 AnJo             12 Crow
Phl  -1 1 em,Jx,RR,go,ca,AJ,JT 8 Cica
Tor  +1   0 Rd11              6 AnJo
Was -11   6 Rd11             15 kmed

West +/-   Best               Worst
Dal  +7   3 trzu,AnJo        16 Rd11
Den  +4   1 KPel              7 Crow
GSW  -6   1 emin,KPel        16 Rd11
Hou -10   6 Rd11,bbst,KPel   16 Ci,JT
LAC  +7   1 Crow             10 sndi
LAL  -5   2 Cica,KPel         9 Rd11
Mem   0   0 Jx,RR,go,km,KP    5 J-TF
Min  -6   0 emin             19 Rd11
NOP  -6   2 trzu             13 emin
OKC  +4   0 Jinx,cali        11 Cica
Phx  -7   1 Rd11             12 AnJo
Por  +8   1 Crow             18 J-TF
Sac +19  15 emin,trzu        25 Rd11
SAS  +5   0 Rd11             11 kmed
Uta  -6   2 shad,AnJo,J-TF   12 Rd11

Sixers are projecting to 51 wins. Five guessed lower and 11 higher, but none right at 51.
Including ties, total bests and worsts at the allstar break:

Code: Select all

trzu  3  0
cali  2  0
emin  6  1
AnJo  5  2
KPel  7  2
shad  2  1
Jinx  3  0
RyRi  2  0
bbst  2  0
gold  2  1
sndi  0  1
Crow  3  3
kmed  2  2
J-TF  5  4
Rd11  6 11
Cica  5  4

Almost certainly I missed some. If not for ties counting as "wins", there would be a lot more zeroes.
All worst guesses are upper or lowest for a team. Many of those extremes are best right now: Brk, Cle, Ind, Mia, Mil, NYK, Orl, Was, Dal, Den, Hou, LAC, NOP, Phx, Por, Sac, SAS. More than half of teams are performing completely outside the range of our guesses, for better or worse.

Posted: **Sat Feb 16, 2019 8:39 pm**

A couple days later and the results have moved some. Appreciate the data updates.

Posted: **Sat Feb 16, 2019 9:49 pm**

The leader in absolute error and root mean squared error is trzu, but others are better in outlying exponents.

Code: Select all

From - To
.01 - .62  emin
.63 - .79  cali
.80 - 5.5  trzu
5.6 +     vegas

Eminence is tops on direct hits, Vegas at fewest super-errors.

Crow, at the end of Jan. you were just .77 off the lead and slightly better than avg.
Eminence held 2nd place in the 3rd week of Jan., from .32 to .38 out of 1st. Other than that, it's been Caliban at #2 since Dec. 20; getting within .08 on V-day.

These things can change rapidly. I'm very surprised at the stability of the rankings, and the continued outside-prediction of so many teams.

Posted: **Tue Feb 19, 2019 5:40 am**

Mike G wrote: ↑Sat Feb 16, 2019 7:42 pm Best and worst predictions for each team. The +/- column is how much better or worse the team is doing (via b-r.com) than the avg error from our 16 or so projectionists.

Code: Select all

East +/-   Best               Worst
Atl  +1   0 emin,Crow,bbst    6 Rd11
Bos  -2   0 Cica              5 shad
Brk  +7   3 KPel             12 Crow
Cha   0   0 AnJo,KPel        10 Cica
Chi  -5   1 Cica             13 J-TF
Cle -15   7 J-TF             21 Rd11
Det   0   0 emin,shad,J-TF    8 Rd11
Ind  +6   0 Cica              8 Rd11
Mia  -4   1 Cica,J-TF         7 Rd,KP
Mil +14   6 Rd11             18 gold
NYK  -6   1 kmed             12 KPel
Orl  +8   3 AnJo             12 Crow
Phl  -1 1 em,Jx,RR,go,ca,AJ,JT 8 Cica
Tor  +1   0 Rd11              6 AnJo
Was -11   6 Rd11             15 kmed

West +/-   Best               Worst
Dal  +7   3 trzu,AnJo        16 Rd11
Den  +4   1 KPel              7 Crow
GSW  -6   1 emin,KPel        16 Rd11
Hou -10   6 Rd11,bbst,KPel   16 Ci,JT
LAC  +7   1 Crow             10 sndi
LAL  -5   2 Cica,KPel         9 Rd11
Mem   0   0 Jx,RR,go,km,KP    5 J-TF
Min  -6   0 emin             19 Rd11
NOP  -6   2 trzu             13 emin
OKC  +4   0 Jinx,cali        11 Cica
Phx  -7   1 Rd11             12 AnJo
Por  +8   1 Crow             18 J-TF
Sac +19  15 emin,trzu        25 Rd11
SAS  +5   0 Rd11             11 kmed
Uta  -6   2 shad,AnJo,J-TF   12 Rd11

Sac +19 15 emin,trzu 25 Rd11

sndesai1 actually gets credit for best SAC projection at 29 wins

Other notes on teams important to my entry:
My biggest miss relative to average is OKC. Going to need Paul George to look the tiniest bit human down the stretch. Their schedule is certainly more difficult, but I was way too low on OKC.

While I wasn't very far off on the average entry, CLE's season turned for the worst almost instantly. If they care to play Love past the first timeout down the stretch, could minimize that error a bit.

Will hope Detroit can continue to play well despite trading Bullock. Ish Smith being back has been important.

Continued AD saga helps my NOP projection relative to average.

Posted: **Tue Feb 19, 2019 7:49 am**

sndesai1 actually gets credit for best SAC projection at 29 wins

I have his guess at 24, just above avg. And his numbers sum to 41.00 per team.

On Caliban's table it has 29, but at this page it shows 24:
https://docs.google.com/spreadsheets/d/ ... 1029099383

Posted: **Tue Feb 19, 2019 7:40 pm**

Mike G wrote: ↑Tue Feb 19, 2019 7:49 am
sndesai1 actually gets credit for best SAC projection at 29 wins
I have his guess at 24, just above avg. And his numbers sum to 41.00 per team.

On Caliban's table it has 29, but at this page it shows 24:
https://docs.google.com/spreadsheets/d/ ... 1029099383

Oh ok nevermind, I see. Keep me with the best SAC projection, I need all of their wins!

Posted: **Tue Feb 19, 2019 10:42 pm**

Last year at this time, eminence was still .66 off the lead. A few weeks earlier, it was twice that much.
I don't remember a big mid-season tidal change in the NBA. Just a bunch of unrelated individual teams went his way.
Nobody ever came back from 1-3 in the Finals, until LeBron- Kyrie, et al.

Posted: **Wed Feb 20, 2019 2:03 am**

Mike G wrote: ↑Tue Feb 19, 2019 10:42 pm Last year at this time, eminence was still .66 off the lead. A few weeks earlier, it was twice that much.
I don't remember a big mid-season tidal change in the NBA. Just a bunch of unrelated individual teams went his way.
Nobody ever came back from 1-3 in the Finals, until LeBron- Kyrie, et al.

I feel like I'm Ty Lue in this analogy

Posted: **Sun Feb 24, 2019 2:07 pm**

Yesterday was quite a bad day for our projectors. There were 12 games, 24 teams that played.
For our gang to mostly look better, the overachieving teams (relative to most predictions) have to lose, while the underachieving teams must win. Margin of victory matters some, wins and losses matter more, at this time of year.

In the 12 games played, 4 were matchups between 2 "under" teams. Results -- NOP>LAL, Chi>Bos, Hou>GSW, and Cle>Mem -- Someone had to lose.
One game had two "over" teams: Sac>OKC. So these results had no effect on most contestants.

Then of the remaining 14 teams in 7 games, just one game went favorably, as Utah beat Dallas. Otherwise it was just more of the same -- overs being over, unders being under.
Showing overachievement to date and yesterday's win margin:

Code: Select all

over  tm   mov     under   tm   mov
14.6  Mil   12     -11.7   Was   -7
9.4   Por   15      -7.6   Phx   -8
7.2   Brk    2      -5.8   Min  -12
6.8   Ind    7      -5.6   Uta   16
6.5   Dal  -16      -5.0   Mia  -13
1.3   Atl    8      -1.7   Phl  -15
0.9   Det   13      -0.2   Cha   -2

Correlation between the columns is .44, and corr. between "over" and Winning was .60

Update Feb 25: Forgot yesterday to post the rankings:

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.61   6.78      bbst   6.37   7.55
cali   5.68   7.35      gold   6.57   8.03
emin   5.93   7.59      sndi   6.61   8.07
AnJo   5.97   7.51      538_   6.68   8.10
KPel   6.08   7.52      Crow   6.68   8.68
shad   6.21   7.44      J-TF   6.87   8.95
vegas  6.23   7.54      kmed   6.99   8.77
Jinx   6.27   7.78      Cica   7.58   9.21
RyRi   6.29   7.65      Rd11   7.63   9.88

Yesterday's 6.49 was the worst average error since Dec. 13 and worst leader error since the 11th. Dates to live in infamy.
update Feb. 28 -- trzu had the worst night, and cali had the best:

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.69   6.74      Jinx   6.33   7.71
cali   5.72   7.26      gold   6.49   7.95
AnJo   5.93   7.45      sndi   6.59   8.02
emin   5.97   7.57      Crow   6.67   8.59
KPel   6.14   7.47      538_   6.68   8.04
shad   6.17   7.39      J-TF   6.80   8.86
vegas  6.21   7.47      kmed   6.96   8.72
RyRi   6.25   7.59      Cica   7.56   9.11
bbst   6.33   7.52      Rd11   7.68   9.90

Atl>Min, Chi>Mem, SAS>Det all favor caliban; no result favored tarrazu.
-- update Mar.3 -- a good day for everyone, none more than fearless leader who overnight tripled his lead:

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.47   6.55      Jinx   6.18   7.63
cali   5.69   7.12      gold   6.25   7.80
AnJo   5.71   7.31      sndi   6.34   7.88
emin   5.82   7.41      Crow   6.42   8.36
KPel   5.92   7.37      538_   6.56   7.95
shad   5.98   7.27      J-TF   6.76   8.68
vegas  6.04   7.35      kmed   6.83   8.61
bbst   6.07   7.40      Cica   7.51   9.01
RyRi   6.10   7.47      Rd11   7.53   9.77

Mar. 6 -- Large improvements and a new #2:

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.26   6.38      Jinx   6.05   7.47
AnJo   5.50   7.07      gold   6.12   7.67
cali   5.59   6.92      Crow   6.27   8.15
emin   5.77   7.14      sndi   6.29   7.72
KPel   5.79   7.19      538_   6.41   7.79
shad   5.80   7.06      J-TF   6.68   8.48
vegas  5.83   7.12      kmed   6.77   8.50
RyRi   5.94   7.27      Cica   7.33   8.75
bbst   6.01   7.26      Rd11   7.50   9.72

Avg 6.16 is best since Feb. 11; leader error is lowest since Feb. 2
Mar. 13 update -- trzu lead keeps growing

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.38   6.51      Jinx   6.25   7.58
cali   5.75   6.96      gold   6.28   7.81
AnJo   5.77   7.21      Crow   6.43   8.30
emin   5.97   7.33      sndi   6.45   7.83
KPel   6.00   7.35      538_   6.61   7.87
shad   6.02   7.20      J-TF   6.65   8.44
RyRi   6.09   7.38      kmed   6.94   8.58
vegas  6.10   7.25      Cica   7.52   8.91
bbst   6.17   7.46      Rd11   7.71   9.77

Posted: **Fri Mar 15, 2019 1:02 am**

Thanks for the updates Mike G.

I like the win totals data viz from Caliban shown here: viewtopic.php?f=2&t=9503&start=105#p33637

If you are around for an update, would be much appreciated.

Posted: **Fri Mar 15, 2019 1:15 am**

As we near season's end, all projections should be converging, I would think. Some teams may be obviously tanking, or resting, but that's hard for an analytical system to predict.

In 4 days, your lead has shrunk from .43 to .23
Everyone else here considers that to be good news.

Posted: **Wed Mar 20, 2019 2:53 pm**

With tarrazu holding a comfy lead, everyone else should be okay with fewer things going his way.
About 18 teams are either above or below everyone's predictions. The other 12 are going to decide this contest. To foil the trajectory that trzu currently enjoys, these things (some or most) should happen:

Code: Select all

must win - favors:
Atl - AnJo, gold, kmed
Ind - emin, Cica; cali, shad
Phl - Cica; Crow, kmed, sndi
. . . Rd11, RyRi, bbst
OKC - everyone almost
 
must lose - favors:
Det - everyone but gold
Mia - J-TF, Cica, gold, kmed
Dal - everyone but AnJo
Mem - Cica, cali, emin
Uta - J-TF, shad, AnJo

either, just do it:
Chi
Tor
Phx

These last 3 are projecting right at trzu's number, so going either way would benefit some but not trzu.
If your guesses are very close to trzu's, there may be just a couple of teams you can hope will cooperate.
Current standings:

Code: Select all

avg abs err   rmse      avg abs err   rmse
trzu   5.50   6.54      gold   6.25   7.83
AnJo   5.80   7.30      Jinx   6.26   7.60
cali   5.83   6.97      Crow   6.34   8.16
KPel   5.98   7.44      sndi   6.45   7.84
emin   6.02   7.41      538_   6.58   7.98
shad   6.03   7.29      J-TF   6.76   8.45
RyRi   6.11   7.46      kmed   6.95   8.66
vegas  6.11   7.35      Cica   7.47   8.84
bbst   6.13   7.55      Rd11   7.68   9.69

APBRmetrics

2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest

Re: 2018-19 team win projection contest