Page 6 of 11

Re: 2019-20 team win projection contest

Posted: Wed Oct 23, 2019 8:15 pm
by tarrazu
Thanks for adding these Mike G.

Code: Select all

tm    avg     highest     lowest 
Phl   51.8   trzu   55   43   Rd11
SAS   41.4   trzu   47   37   emin
Cha   27.5   Rd11   35   24   trzu
Dal   42.4   trzu   47   37   ncs.
Brk   39.2   shad   43   36 snd,trz
Thought it was interesting to see how my 'outliers' compared to APBR averages and betting markets, using Circa/Westgate (Las Vegas) and Pinnacle.

Code: Select all

Team	Circa  Westgate Pinnacle Trzu
PHI: 	55.5	55	55.5	55
SAS:	45.5	45.5	43.5	47
CHA:	23.5	22.5	21.5	24
DAL:	42.5	43	43.5	47
BKN: 	43.5	42.5	42.5	36
PHI: I'm in line with the market but we are collectively much lower.
SAS: About as much higher than market as our average is lower.
CHA: Despite having the lowest projection, market was even lower.
DAL: My highest 'outlier' relative to market. Our average about in line.
BKN: My lowest 'outlier' relative to market, but also our average much lower than the market.

APBR Average Picks Over/Under (+/-3 or more games):
ATL Under
BKN Under
CHA Over
IND Under
OKC Over
PHI Under
PHX Over
POR Under
SAS Under
UTA Under
WAS Over

Re: 2019-20 team win projection contest

Posted: Wed Oct 23, 2019 9:25 pm
by Crow
I dunno how I will do on average overall picking as many extremes as I did but on those extremes I expect to be more right on them than the other extreme at least. Trying to pick right in the middle is often the way to win but if you do that you have to do it for most or all. I went pretty straightforward, what I thought they will win, not what was the most strategic contest guess.

Clippers, I talked myself into lowering several times at the end for imbalances, redundancies and possible injuries. I probably went too far. Tried to be somewhat different than the norm.

Re: 2019-20 team win projection contest

Posted: Sat Oct 26, 2019 12:06 pm
by Mike G
After 1 or 2 games per team. Yes, this is just for fun.

Code: Select all

Rd11    9.9      ncs.   10.8
Mike    9.9      trzu   10.8
cali   10.2      .538   10.8
re19   10.2      RyRi   11.0
lisp   10.5      gold   11.3
shad   10.5      sndi   11.4
avg.   10.6      Crow   11.6
vegas  10.7      emin   12.1
Based on current W and L, and SRS according to b-r.com
Currently the east looks stronger than the west by about 1.7 ppg

The re19 entry is just [last year's wins] (W19 +41)/2
Likely I will use this template until b-r.com starts their playoff forecast page, increasing the weight on SRS as the evidence mounts.

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 11:43 am
by Mike G
After 3 or 4 games, team SRS seem to have stabilized. It kinda looks like this:

Code: Select all

west   W      east   W
Phx   55      Phl   60
Uta   54      Atl   56
LAC   53      Bos   53
LAL   52      Tor   51
Min   52      Mil   50
Den   47      Mia   50
SAS   46      Was   44
Dal   44      Orl   44

OKC   42      Det   37
Hou   42      Brk   35
Por   39      Cle   32
NOP   28      NYK   28
GSW   24      Cha   24
Mem   23      Ind   24
Sac   19      Chi   23
I just scaled the SRS back toward zero until projected wins range from ~21 to ~61.
This tends to pack the projections around .500 and favors those who didn't predict anything drastic.

Code: Select all

. avg err         avg err
trzu   8.0      emin   8.8
cali   8.3      RyRi   8.9
lisp   8.3      sndi   8.9
avg.   8.4      ncs.   9.1
538    8.6      gold   9.2
shad   8.7      re19   9.5
vegas  8.7      Mike   9.5
Rd11   8.8      Crow   9.9

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 4:49 pm
by nbacouchside
Mike G wrote: Tue Oct 29, 2019 11:43 am After 3 or 4 games, team SRS seem to have stabilized. It kinda looks like this:

Code: Select all

west   W      east   W
Phx   55      Phl   60
Uta   54      Atl   56
LAC   53      Bos   53
LAL   52      Tor   51
Min   52      Mil   50
Den   47      Mia   50
SAS   46      Was   44
Dal   44      Orl   44

OKC   42      Det   37
Hou   42      Brk   35
Por   39      Cle   32
NOP   28      NYK   28
GSW   24      Cha   24
Mem   23      Ind   24
Sac   19      Chi   23
I just scaled the SRS back toward zero until projected wins range from ~21 to ~61.
This tends to pack the projections around .500 and favors those who didn't predict anything drastic.

Code: Select all

. avg err         avg err
trzu   8.0      emin   8.8
cali   8.3      RyRi   8.9
lisp   8.3      sndi   8.9
avg.   8.4      ncs.   9.1
538    8.6      gold   9.2
shad   8.7      re19   9.5
vegas  8.7      Mike   9.5
Rd11   8.8      Crow   9.9
tarrazu kicking ass again. god damn lol.

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 6:06 pm
by Crow
Outlier projections have plenty of opportunity to be wrong on their own but they will also look bad against a measuring stick regressed to mean and early when noise competes heavier with signal.

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 7:09 pm
by Mike G
nbacouchside wrote: Tue Oct 29, 2019 4:49 pm tarrazu kicking ass again...
If I lower the weight on current SRS, to 3/4 of this arbitrary level, caliban takes the top spot.
At 60%, it's Rd11 on top. At this point, the spread in wins is 26 (Sac) to 54 (Phl). Rd11 also had a narrow range of projections, with only Mil outside the 35-50 range.

At 25% more weight on SRS, I capture last place.
At 150% the current SRS weight, tarrazu still leads, Sixers project to 73 wins, Kings looking at 6.

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 7:57 pm
by Crow
At some point later in season, if you want to cut it different in another way, score the predictions by conference. I tend to do less well in the East, which I know and care less about. If I do badly this season, it will likely be in the East (at least).

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 8:02 pm
by Crow
About 6 teams are doing significantly different than I expected in broad terms. W-L not a good guide yet.

Re: 2019-20 team win projection contest

Posted: Tue Oct 29, 2019 11:55 pm
by Mike G
The West seems to be less in line with most of our predictions. On avg, we are off by 8.3 on East teams and 10.7 in the West.
You (Crow) are less imbalanced than most, at 9.3 and 10.5
sndesai is extreme at E 7.8 and 10.0 W
By contrast, I am worst in the east, not so bad on the west: 10.1 and 9.0

Projected over-unders relative to our avg predictions:

Code: Select all

GSW  -22      26   Atl
Phx   21     -20   Ind
Sac  -17     -14   Chi
Min   14      13   Was
Hou  -13       8   Phl
NOP  -13       8   Cle
Mem   -7      -6   Mil
OKC    5       6   Mia
SAS    4       6   Bos
Den   -4      -4   Brk
Por   -4       4   Tor
Uta    3      -3   Cha
LAL    3       1   Orl
Dal    2       0   NYK
LAC    2       0   Det

Re: 2019-20 team win projection contest

Posted: Wed Oct 30, 2019 2:56 am
by Crow
Thanks for the info. I am less imbalanced in my errors than others by conference but I am modestly better in west than east for myself and compared to the competition.

Re: 2019-20 team win projection contest

Posted: Wed Oct 30, 2019 12:42 pm
by Mike G
It's a bit early to think of them as "errors", though that is the mathematical term. Our errors on the Hawks dropped by 10 overnight with one bad game, while the Heat outstripped most of us by 6.

With no inter-conference games, idle teams all lost SRS in the East while gaining in the West. Now the projections indicate an average of 43.6 wins per West team vs 38.4 in the East -- almost double last year's imbalance.

Re: 2019-20 team win projection contest

Posted: Wed Oct 30, 2019 4:36 pm
by Crow
Yes, it is only truly "error" at the end. Variance til then.

Re: 2019-20 team win projection contest

Posted: Thu Oct 31, 2019 4:41 am
by eminence
Dang, lose watching a great player and that may ice my fate in this contest as well. Gives me a better excuse for the crazy bad prediction though right?

Re: 2019-20 team win projection contest

Posted: Thu Oct 31, 2019 1:41 pm
by Rd11490
eminence wrote: Thu Oct 31, 2019 4:41 am Dang, lose watching a great player and that may ice my fate in this contest as well. Gives me a better excuse for the crazy bad prediction though right?
SMH, Gonna win on an injury. Real bad predictions come from bad modeling :D