STAT CHECK: ACS round 1

Recently, I made a post about my predictions for the ACS for every player at Champions
Here is an analysis of how well my predictions have been so far.
This only counts stats from the opening series (not FNC/VS which has already been played).
Because each player has only played a single series (in some cases a single map) I expected a high variance.
The ACS listed is my predictions before champions, Round 1 is the ACS after the first round, Diff is the ACS difference, Rank is their rank sorted by ACS, and R Diff is the difference in my predicted rank and the actual after Round 1.
I over predicted the total ACS in the tournament. The average of my predictions is 206.9 while the average of the actual is 199.9.

tldr: I think I predicted most of the players fairly well with some glaring exceptions, but variance is expected with only 1 series played.
Some standouts: I predicted PTC and keznit to do way better than they did, and Derke, Mazino, k1Ng, and L1NK to do way worse.
Sacy was even higher than I predicted even though I had him way higher than most people while I predicted ade's second to last place perfectly but he still did way worse than I expected.

    Player  ACS Round 1 Diff    Rank    R Diff
1   ScreaM  280.0   233.0   -47.0   19  -18
2   TenZ    275.0   266.3   -8.7    4   -2
3   mwzera  271.0   228.0   -43.0   22  -19
4   cNed    270.0   266.0   -4.0    5   -1
5   heat    259.0   251.7   -7.3    9   -4
6   Sacy    255.0   287.0   32.0    2   4
7   PTC     246.0   148.0   -98.0   73  -66
8   zeek    245.0   227.7   -17.3   23  -15
9   yay     245.0   244.5   -0.5    14  -5
10  Pati    245.0   230.5   -14.5   21  -11
11  keznit  241.0   165.0   -76.0   64  -53
12  nAts    240.0   234.0   -6.0    18  -6
13  SicK    238.0   183.7   -54.3   49  -36
14  sheydos 235.0   226.0   -9.0    25  -11
15  JohnOl  234.0   209.5   -24.5   32  -17
16  leaf    233.0   193.0   -40.0   44  -28
17  saadhak 231.0   250.0   19.0    10  7
18  crashie 230.0   204.5   -25.5   35  -17
19  BuZz    230.0   286.0   56.0    3   16
20  Jamppi  230.0   179.0   -51.0   51  -31
21  frz     229.0   163.5   -65.5   67  -46
22  xand    226.0   217.0   -9.0    27  -5
23  ShahZ   225.0   203.7   -21.3   36  -13
24  Rb      225.0   199.5   -25.5   38  -14
25  starxo  223.0   167.7   -55.3   63  -38
26  xeta    223.0   185.7   -37.3   46  -20
27  Munchk  221.0   197.5   -23.5   39  -12
28  gtnziN  220.0   233.0   13.0    20  8
29  MaKo    220.0   178.0   -42.0   52  -23
30  dapr    219.0   177.0   -42.0   53  -23
31  Witz    219.0   175.7   -43.3   55  -24
32  Nivera  219.0   249.5   30.5    11  21
33  Medusa  218.0   248.0   30.0    12  21
34  murizzz 215.0   185.3   -29.7   47  -13
35  foxz    215.0   173.0   -42.0   57  -22
36  Chronic 214.0   246.7   32.7    13  23
37  Bazzi   214.0   216.0   2.0 28  9
38  Xeppaa  214.0   224.7   10.7    26  12
39  Nozwerr 213.0   210.3   -2.7    31  8
40  DubsteP 211.0   195.7   -15.3   41  -1
41  Derke   210.0   291.3   81.3    1   40
42  Lakia   210.0                  
43  d3ffo   205.0   258.3   53.3    7   36
44  Victor  205.0   185.0   -20.0   48  -4
45  Khalil  202.0   226.3   24.3    24  21
46  LAMMY   201.0   213.5   12.5    30  16
47  stax    196.0   171.5   -24.5   59  -12
48  Doma    195.0   197.3   2.3 40  8
49  Fisker  195.0   241.0   46.0    16  33
50  sScary  194.0   176.5   -17.5   54  -4
51  Klaus   192.0   171.5   -20.5   60  -9
52  Marved  190.0   207.5   17.5    33  19
53  JessieV 188.0   216.0   28.0    29  24
54  BORKUM  187.0   143.0   -44.0   74  -20
55  soulcas 187.0   238.5   51.5    17  38
56  Quick   187.0   159.3   -27.7   69  -13
57  BONEC   186.0   165.0   -21.0   65  -8
58  Mazino  186.0   264.0   78.0    6   52
59  neth    184.0   143.0   -41.0   75  -16
60  mitch   184.0   172.3   -11.7   58  2
61  NagZ    183.0   190.5   7.5 45  16
62  Sushib  181.0   138.5   -42.5   77  -15
63  sutecas 180.0   182.0   2.0 50  13
64  ChAlala 179.0   127.5   -51.5   80  -16
65  Redgar  178.0   136.3   -41.7   78  -13
66  k1Ng    178.0   252.5   74.5    8   58
67  dispens 177.0   170.0   -7.0    61  6
68  zombs   175.0   163.3   -11.7   68  0
69  SuperBu 175.0   157.0   -18.0   71  -2
70  Boaster 175.0   168.0   -7.0    62  8
71  Crws    172.0   195.5   23.5    42  29
72  vanity  171.0   175.0   4.0 56  16
73  Kiles   170.0   194.7   24.7    43  30
74  Magnum  170.0   202.3   32.3    37  37
75  L1NK    169.0   243.0   74.0    15  60
76  Mistic  165.0   149.7   -15.3   72  4
77  FNS     165.0   206.5   41.5    34  43
78  v1xen   165.0   139.3   -25.7   76  2
79  delz1k  165.0   135.5   -29.5   79  0
80  Mazin   165.0   164.0   -1.0    66  14
81  ade     160.0   108.0   -52.0   81  0
82  Jhow    145.0   157.7   12.7    70  12

Jhow be like:

derke is on fire

These code text tables are really hard to read. Unfortunate, but the info is there if you like staring lol.

Not too bad so far indeed. Let me try to illustrate.

I made a basic 0-1 loss function and ran kernel density estimates for three ACS distributions(prior,predictions,random). I also plotted the mean error of three methods over all the players with mean error being the absolute difference between the true ACS and predicted, random & lifetime ACS.

Kernel Density Estimates: https://i.imgur.com/xO8Kf9S.png

The kernel density estimates show, that your prediction distribution is overall closer to the true distribution, than the distribution of prior ACS. At least so far. The random ACS distribution is close in spread to the true ACS distribution, but shifted, because it is literally just 81 random values from 150-300.

0-1 loss function:

Predictions correct Truth Predicted Prior Absolute difference from truth(pred,prior)
0/1                 233.0 280       267   (47.0, 34.0)
1/2                 266.0 275       269   (9.0, 3.0)
1/3                 228.0 271       277   (43.0, 49.0)
2/4                 266.0 270       262   (4.0, 4.0)
3/5                 251.0 259       248   (8.0, 3.0)
3/6                 287.0 255       249   (32.0, 38.0)
3/7                 148.0 246       236   (98.0, 88.0)
4/8                 227.0 245       230   (18.0, 3.0)
5/9                 244.0 245       240   (1.0, 4.0)
6/10                230.0 245       249   (15.0, 19.0)
6/11                165.0 241       270   (76.0, 105.0)
7/12                234.0 240       230   (6.0, 4.0)
7/13                183.0 238       226   (55.0, 43.0)
8/14                226.0 235       232   (9.0, 6.0)
8/15                209.0 234       245   (25.0, 36.0)
8/16                193.0 233       245   (40.0, 52.0)
9/17                250.0 231       218   (19.0, 32.0)
9/18                204.0 230       210   (26.0, 6.0)
9/19                286.0 230       239   (56.0, 47.0)
9/20                179.0 230       224   (51.0, 45.0)
9/21                163.0 229       214   (66.0, 51.0)
10/22               217.0 226       246   (9.0, 29.0)
10/23               203.0 225       215   (22.0, 12.0)
10/24               199.0 225       232   (26.0, 33.0)
10/25               167.0 223       212   (56.0, 45.0)
10/26               185.0 223       198   (38.0, 13.0)
10/27               197.0 221       224   (24.0, 27.0)
11/28               233.0 220       208   (13.0, 25.0)
11/29               178.0 220       220   (42.0, 42.0)
11/30               177.0 219       201   (42.0, 24.0)
11/31               175.0 219       217   (44.0, 42.0)
11/32               249.0 219       202   (30.0, 47.0)
11/33               248.0 218       212   (30.0, 36.0)
11/34               185.0 215       215   (30.0, 30.0)
11/35               173.0 215       220   (42.0, 47.0)
11/36               246.0 214       225   (32.0, 21.0)
12/37               216.0 214       216   (2.0, 0.0)
13/38               224.0 214       225   (10.0, 1.0)
14/39               210.0 213       229   (3.0, 19.0)
15/40               195.0 211       238   (16.0, 43.0)
15/41               291.0 210       254   (81.0, 37.0)
15/42               258.0 205       219   (53.0, 39.0)
15/43               185.0 205       228   (20.0, 43.0)
15/44               226.0 202       194   (24.0, 32.0)
16/45               213.0 201       213   (12.0, 0.0)
16/46               171.0 196       214   (25.0, 43.0)
17/47               197.0 195       230   (2.0, 33.0)
17/48               241.0 195       237   (46.0, 4.0)
18/49               176.0 194       215   (18.0, 39.0)
18/50               171.0 192       197   (21.0, 26.0)
19/51               207.0 190       206   (17.0, 1.0)
19/52               216.0 188       189   (28.0, 27.0)
19/53               143.0 187       216   (44.0, 73.0)
19/54               238.0 187       201   (51.0, 37.0)
19/55               159.0 187       215   (28.0, 56.0)
19/56               165.0 186       187   (21.0, 22.0)
19/57               264.0 186       227   (78.0, 37.0)
19/58               143.0 184       217   (41.0, 74.0)
20/59               172.0 184       204   (12.0, 32.0)
21/60               190.0 183       231   (7.0, 41.0)
21/61               138.0 181       206   (43.0, 68.0)
22/62               182.0 180       187   (2.0, 5.0)
22/63               127.0 179       181   (52.0, 54.0)
22/64               136.0 178       172   (42.0, 36.0)
22/65               252.0 178       205   (74.0, 47.0)
23/66               170.0 177       208   (7.0, 38.0)
24/67               163.0 175       172   (12.0, 9.0)
25/68               157.0 175       185   (18.0, 28.0)
26/69               168.0 175       191   (7.0, 23.0)
26/70               195.0 172       196   (23.0, 1.0)
27/71               175.0 171       191   (4.0, 16.0)
27/72               194.0 170       196   (24.0, 2.0)
27/73               202.0 170       213   (32.0, 11.0)
27/74               243.0 169       188   (74.0, 55.0)
28/75               149.0 165       191   (16.0, 42.0)
28/76               206.0 165       181   (41.0, 25.0)
28/77               139.0 165       182   (26.0, 43.0)
28/78               135.0 165       200   (30.0, 65.0)
29/79               164.0 165       194   (1.0, 30.0)
29/80               108.0 160       178   (52.0, 70.0)
30/81               157.0 145       167   (12.0, 10.0)
Total prior err = 2582.0 Total pred err = 2436.0 Total rand err = 4183.0
Average prior err = 31.88 Average pred err = 30.07 Average rand err = 51.64
Correct predictions(pred): 30/81 = 37%
Correct predictions(prior acs): 24/81 = 29%
Correct predictions(random): 17/81= 20%

The basic premise with the 0-1 loss function is categorizing whether a prediction is correct or not based on whether it's within 20 ACS of the true ACS.

It shows that you made 6 more predictions classified as "correct" compared to lifetime ACS. And 13 more correct predictions compared to random ACS.

Mean error over players: https://i.imgur.com/hX36fpS.png

It shows that the mean errors of prior ACS are similar to ones from your predictions and that the mean error of random ACS approaches 50. I refrain from hypothesis testing here as these are averages of averages, the purpose is just to illustrate how well you are doing in relation to lifetime acs as a predictor and random acs as a predictor.