2020-21 team win projection contest

[Crow]

0.13 off MAE lead overall. Currently first among contest class.

Slight 16-14 lead compared to Vegas going team by team for closeness.

C Wood's injury and J Butler's absence hurt my average error considerably.

[Mike G]

The professional bettors looking better.

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  4.88   6.27      dtka   5.58   7.14
bbst   5.04   6.51      538R   5.59   7.67
EExp   5.07   6.12      KPel   5.59   7.31
cali   5.20   6.46      DSM1   5.69   7.35
shad   5.20   6.55      lisp   5.69   7.63
Crow   5.23   6.71      emin   6.34   7.87
avg.   5.24   6.76      538E   6.65   8.33
TeRa   5.34   6.63      EBPI   8.55  10.17
trzu   5.34   7.08              

These are worst MAE of the year for cali, bbst, trzu, lisp, 538e, and the avg of all. The overachieving teams keep winning, while the unders keep undering.

I'm not sure why there should be much difference between errors from the lisp list and from b-r.com projections, but for some there is significant variation.

Code: Select all

lisp  av err  rmse      b-r  avg err  rmse
vegas  5.00   6.39      vegas  4.92   6.30
EExp   5.02   6.19      bbst   5.09   6.51
bbst   5.14   6.62      EExp   5.14   6.16
shad   5.18   6.63      cali   5.24   6.48
Crow   5.21   6.80      shad   5.28   6.57
avg.   5.28   6.92      Crow   5.32   6.74
cali   5.34   6.59      avg.   5.32   6.77
TeRa   5.36   6.74      trzu   5.44   7.07
trzu   5.42   7.25      TeRa   5.44   6.66
dtka   5.51   7.25      KPel   5.61   7.24
538R   5.55   7.68      538R   5.65   7.63
DSM1   5.68   7.45      dtka   5.68   7.17
lisp   5.98   7.85      DSM1   5.76   7.36
KPel   5.99   7.57      lisp   5.78   7.62
emin   6.56   8.07      emin   6.34   7.88
538E   6.92   8.51      538E   6.55   8.25
EBPI   8.38  10.05      EBPI   8.63  10.21

The top 3 are the same, and the top 7, 9, 14 -- in different order.
I have been averaging these but would prefer to do just one.
https://www.basketball-reference.com/fr ... _prob.html


Update Mar. 19

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  4.99   6.27      538R   5.66   7.66
EExp   5.07   6.08      dtka   5.66   7.11
bbst   5.13   6.52      DSM1   5.72   7.34
Crow   5.26   6.75      KPel   5.81   7.37
shad   5.28   6.54      lisp   5.91   7.65
cali   5.33   6.42      emin   6.37   7.88
avgA   5.34   6.77      538E   6.93   8.55
TeRa   5.41   6.60      EBPI   8.55  10.29
trzu   5.47   7.10            

Yesterday, almost everyone had their worst numbers of the season.
Last night, 2 games -- Was>Uta and Min>Phx -- improved everyone by about 0.16


Update Mar. 24
Last night 6 games, 12 teams mostly doing more of what they aren't supposed to be doing -- winning or losing too much.
bad-bad: Den>Orl , NYK>Was , Phx>Mia , Brk>Por
good-bad: NOP>LAL, Phl>GSW
A few entrants were not 'bad' on quite every outcome; but everyone suffers their worst errors of the season.

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  5.22   6.60      dtka   5.83   7.31
bbst   5.35   6.84      KPel   5.93   7.60
EExp   5.38   6.45      538R   5.93   7.91
shad   5.47   6.87      DSM1   5.99   7.64
cali   5.55   6.68      lisp   6.02   7.87
Crow   5.57   7.06      emin   6.57   8.12
avgA   5.57   7.05      538E   7.12   8.79
TeRa   5.62   6.94      EBPI   8.78  10.53
trzu   5.69   7.38            

[Mike G]

Over- and under-achieving teams, ranked by the difference between their APBR-avg prediction and the projections by b-r.com

Code: Select all

over   tm   proj   avgA      over    tm   proj   avgA
13.1  NYK   35.6   22.5     -18.7   Hou   19.9   38.6
11.6  Cha   36.1   24.5     -13.3   Min   18.6   31.9
10.5  Uta   54.0   43.5      -9.6   Tor   32.2   41.8
10.4  Phx   47.6   37.3      -8.2   Mia   34.8   43.0
7.0   Phl   49.1   42.2      -7.0   Orl   23.4   30.4
4.6   OKC   28.8   24.2      -6.0   Was   26.6   32.6
4.4   Chi   32.0   27.6      -4.8   Bos   36.7   41.5
4.0   Cle   26.1   22.1      -4.1   LAL   43.7   47.8
3.6   SAS   34.9   31.3      -4.1   NOP   34.1   38.2
3.2   Brk   46.2   43.0      -2.8   Dal   40.4   43.2
2.2   Atl   37.2   35.1      -2.2   Ind   36.0   38.2
2.0   Det   25.0   23.0      -1.2   Por   38.2   39.4
1.5   Mem   36.5   35.0               
1.3   LAC   47.1   45.8               
1.1   Den   43.4   42.3               
1.1   GSW   35.8   34.7               
0.5   Mil   48.6   48.1               
0.1   Sac   31.2   31.1               

8 of the 12 biggest misses are in the East
This is right after the trade deadline and (I think) before many new players have suited up.


- - Update Mar. 28 -- the rankings have shuffled

Code: Select all

.   avg err   rmse       .  avg err   rmse
bbst   5.35   6.98      TeRa   5.71   7.11
cali   5.37   6.72      KPel   5.72   7.63
vegas  5.37   6.78      Crow   5.73   7.24
avgA   5.45   7.19      538R   5.79   7.99
EExp   5.49   6.60      DSM1   5.84   7.72
shad   5.51   7.01      lisp   6.00   8.06
trzu   5.60   7.55      emin   6.71   8.34
dtka   5.66   7.46      538E   7.01   8.71
.                       EBPI   8.63  10.60

Largest errors of the season for Vegas, EExp, shadow, TeRa, Crow, and eminence.

[Crow]

The slippage of last few weeks gets tougher and tougher to recover from.

But a lot still depends on how accurate the measuring sticks are / will be. Expect more jumbling, at least in the middle. Top performers may stay on top but the order could change further.

[Mike G]

Code: Select all

.   avg err   rmse       .  avg err   rmse
cali   5.28   6.73      KPel   5.67   7.71
vegas  5.34   6.82      Crow   5.71   7.24
bbst   5.36   7.04      TeRa   5.71   7.13
avgA   5.47   7.21      538R   5.81   8.09
EExp   5.48   6.62      DSM1   5.88   7.75
shad   5.52   7.06      lisp   6.02   8.08
trzu   5.60   7.59      emin   6.69   8.27
dtka   5.61   7.46      538E   7.11   8.84
.                       EBPI   8.56  10.59

There's a discrepancy in the RMSE given to caliban , between my spreadsheet and liminal's. Using the team projections on liminal's sheet, I get 6.71; and the online table currently shows 6.59 (ranking #1)

[Mike G]

19 of 30 teams are doing either better or worse than anyone expected; as such, a win or a loss by one of these teams makes no difference in the ranking of the contest's mean absolute errors.
Teams that are within 2 projected wins of our 'avg' predictions are here shown with their current win projection alongside everyone's guesses. The table is streamlined a bit to include APBR members and the 2 main competitors: the ESPN Experts and Circa/Vegas.
A Win by the -team- favors those on the left; a Loss benefits those to the right.

Code: Select all

tm   dtka  emin  EExp  shad  Circ  Crow  DSM1 -Mil-  bbst  lisp  trzu  cali  KPel
Mil   50    50    50   49.8  49.3   49   48.9  48.0  47.9  47.8  47.5  46.0  44.5
                          
tm   emin  Circ  EExp  dtka  shad  bbst -Por-  DSM1  trzu  cali  lisp  KPel  Crow
Por   44   41.3   41    40   39.8  39.8  39.6  39.5  39.5  38.5  38.4  38.2   36
                          
tm   emin  KPel  lisp  bbst -Cle-  trzu  Circ  shad  EExp  DSM1  cali  Crow  dtka
Cle   24   23.8  23.3  22.7  22.6  22.5  22.3  22.1   22   21.6  21.3   21    19
                          
tm   KPel  cali  dtka  DSM1  lisp -Sac-  emin  trzu  shad  bbst  Crow  EExp  Circ
Sac  35.0  33.5   33   31.5  31.5  30.5   30   30.0  29.4  29.4   28    28   27.3
                          
tm   cali  emin  KPel -Det-  Circ  dtka  shad  trzu  EExp  DSM1  Crow  lisp  bbst
Det  28.1   26   23.9  23.8  23.3   23   22.1  22.0   22   21.9   21   20.9  20.8
                          
tm   Crow  lisp -Mem-  trzu  KPel  cali  dtka  DSM1  EExp  bbst  Circ  shad  emin
Mem   40   37.3  36.0  36.0  36.0  35.4   35   34.4   33   32.7  32.3  32.0   31
                          
tm   dtka  cali -LAC-  shad  trzu  lisp  emin  EExp  bbst  Circ  DSM1  Crow  KPel
LAC   48   47.6  47.5  47.2  46.5  46.4   46    46   46.0  45.3  44.6   44   41.7
                          
tm   emin  lisp  shad  Circ  EExp  bbst  dtka  trzu  Crow -GSW-  KPel  DSM1  cali
GSW   43   37.6  36.4  36.3   36   35.7   35   35.0   35   32.8  31.2  30.0  28.1

The -tm- projections are those provided by the liminal_space table -- see first post on page 6 of this topic.

[Crow]

Slipped in late March. Recovered halfway recently.

0.09 MAE off direct contest class lead. 0.22 off overall lead.

[Mike G]

Crow, you had a bad night; game results bumped your MAE up by .15
But everyone ahead of you took hits of .19 to .22
Everyone in the field, but for Crow and 538R, has their worst errors of the season.

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  5.43   6.85      TeRa   5.81   7.19
bbst   5.46   7.12      dtka   5.83   7.55
EExp   5.55   6.67      538R   5.91   8.11
cali   5.56   6.89      KPel   6.00   7.89
avgA   5.62   7.34      DSM1   6.01   7.86
shad   5.62   7.14      lisp   6.25   8.24
Crow   5.65   7.36      emin   6.81   8.37
trzu   5.74   7.70      538E   7.37   9.08
.                       EBPI   8.82  10.85

[Crow]

Another way to put it:

My MAE estimate is 4% higher than the Vegas lead. I consider that a small difference in the big picture.

The difference with current best in the direct contest class (Caliban) is even smaller at 1.6%.

[Crow]

0.06 MAE separate Caliban (1st), Shadow (2nd) and me (3rd) in contest class on liminal space sheet. Barely 1% difference.


As expected by some, ESPN BPI is the by far worst on MAE. Followed by 538 Elo.

538 Raptor is 13th and Team Rankings barely above average at 8th. Could argue to eliminate 538 Raptor weight in measuring stick and reduce or eliminate Team Rankings weight. 40% of weight of measuring stick is on these 2 measures. On their current measures I guess but perhaps somewhat discredited by their pre-season projections?

The end is coming fairly soon though and with it official results superseding the comparison to the measuring stick estimates.

[Mike G]

Ten days later, worst numbers of the season for cali, KPel, DSM1, 538 (both), and EBPI

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  5.32   6.93      dtka   5.85   7.79
EExp   5.40   6.72      trzu   5.87   7.89
bbst   5.43   7.25      KPel   6.06   8.11
shad   5.53   7.26      DSM1   6.15   8.09
Crow   5.60   7.52      lisp   6.37   8.43
cali   5.65   7.15      538R   6.38   8.37
avgA   5.68   7.52      emin   6.72   8.42
TeRa   5.71   7.29      538E   7.57   9.41
.                       EBPI   8.98  11.02

Crow, good news for you is that the Bulls have slipped into the broad contested zone between your lowest guess of 24 and Vegas' high guess of 30. They're now projecting at 29; should they slip to 25 -- and all else remaining equal -- you'd be ahead of Vegas (but behind bbst); and at 24 or lower, you would be in 1st.

[Crow]

Thanks for the tip.

I am looking at some specific teams but hadn't paid a lot of attention to the Chicago Bulls. I was expecting some slippage based on tougher schedule.

[Crow]

Mike G,

Are your recent updates just using B-Ref as the measuring stick?



By the liminal space measuring stick I am .01 MAE behind Shadow right now in the contest class.

[Mike G]

Crow: I am running numbers with the b-r.com projections and with the liminal-space projections, and then averaging them.
The liminal figures have shadow ahead of you (lower) by .01, and b-r.com has you ahead by .01. Essentially, you're tied for 4th.

Code: Select all

.   avg err   rmse       .  avg err   rmse
vegas  5.29   6.86      dtka   5.84   7.72
EExp   5.39   6.64      trzu   5.84   7.79
bbst   5.42   7.17      KPel   5.99   8.02
Crow   5.52   7.41      DSM1   6.14   8.02
shad   5.52   7.19      538R   6.26   8.27
cali   5.57   7.03      lisp   6.31   8.33
avgA   5.65   7.44      emin   6.75   8.42
TeRa   5.71   7.23      538E   7.45   9.31
.                       EBPI   8.94  10.99

UPDATE Apr 25
One week later, not much change in the order:

Code: Select all

   avg err   rmse       .  avg err   rmse
vegas  5.35   7.00      dtka   5.89   7.79
EExp   5.41   6.73      trzu   5.93   7.88
shad   5.50   7.28      KPel   6.06   8.03
bbst   5.56   7.28      DSM1   6.17   8.07
cali   5.56   7.13      lisp   6.26   8.38
Crow   5.56   7.53      538R   6.45   8.46
TeRa   5.68   7.32      emin   6.79   8.52
avgA   5.76   7.52      538E   7.68   9.47
.                       EBPI   9.02  11.11

Pelton with biggest MAE of the season; everyone else has been worse.

[Crow]

0.01 MAE behind Shadow on liminal space sheet. Creeping up a bit on Vegas. Each may be or could come down to difference on 1 team.