Model predictions
|
Adelaide v Carlton
|
|
09 April @ Adelaide Oval
|
|
+76
|
HGA
|
|
|
+88
|
Elo rating
|
−81
|
|
80.34%
|
Win prob. (Elo)
|
19.66%
|
|
24.7
|
xSS
|
24.1
|
|
57.54%
|
Win prob. (xSS)
|
42.46%
|
|
+1.382
|
Team elite tier
|
+1.136
|
|
+0.093
|
Team standard
|
+0.047
|
|
−0.765
|
Team subpar tier
|
−0.780
|
|
60.75%
|
Win prob. (TiCo)
|
39.25%
|
|
5.2
|
Forward strength
|
5.0
|
|
6.3
|
Midfield strength
|
5.8
|
|
5.1
|
Ruck strength
|
6.0
|
|
7.4
|
Backline strength
|
6.4
|
|
49.65%
|
Win prob. (APP)
|
50.35%
|
|
57.99%
|
Win prob. (Ensemble)
|
42.01%
|
|
Collingwood v Fremantle
|
|
10 April @ Adelaide Oval
|
|
+68
|
HGA
|
|
|
+69
|
Elo rating
|
+73
|
|
59.08%
|
Win prob. (Elo)
|
40.92%
|
|
25.6
|
xSS
|
27.3
|
|
41.19%
|
Win prob. (xSS)
|
58.81%
|
|
+1.179
|
Team elite tier
|
+1.354
|
|
+0.093
|
Team standard
|
+0.031
|
|
−0.942
|
Team subpar tier
|
−0.801
|
|
61.65%
|
Win prob. (TiCo)
|
38.35%
|
|
5.3
|
Forward strength
|
6.9
|
|
3.4
|
Midfield strength
|
6.0
|
|
5.2
|
Ruck strength
|
3.8
|
|
5.7
|
Backline strength
|
5.3
|
|
45.15%
|
Win prob. (APP)
|
54.85%
|
|
36.83%
|
Win prob. (Ensemble)
|
63.17%
|
|
North Melbourne v Brisbane
|
|
11 April @ Barossa Oval
|
|
+69
|
HGA
|
|
|
−122
|
Elo rating
|
+160
|
|
22.66%
|
Win prob. (Elo)
|
77.34%
|
|
22.5
|
xSS
|
23.3
|
|
56.07%
|
Win prob. (xSS)
|
43.93%
|
|
+1.147
|
Team elite tier
|
+1.112
|
|
+0.072
|
Team standard
|
+0.104
|
|
−0.962
|
Team subpar tier
|
−0.724
|
|
45.61%
|
Win prob. (TiCo)
|
54.39%
|
|
5.0
|
Forward strength
|
5.4
|
|
6.4
|
Midfield strength
|
5.8
|
|
8.1
|
Ruck strength
|
5.7
|
|
3.4
|
Backline strength
|
5.9
|
|
46.05%
|
Win prob. (APP)
|
53.95%
|
|
37.89%
|
Win prob. (Ensemble)
|
62.11%
|
|
Essendon v Melbourne
|
|
11 April @ Adelaide Oval
|
|
−6
|
HGA
|
|
|
−190
|
Elo rating
|
−16
|
|
26.23%
|
Win prob. (Elo)
|
73.77%
|
|
20.7
|
xSS
|
21.9
|
|
44.06%
|
Win prob. (xSS)
|
55.94%
|
|
+1.030
|
Team elite tier
|
+1.376
|
|
+0.078
|
Team standard
|
+0.065
|
|
−0.860
|
Team subpar tier
|
−0.710
|
|
47.73%
|
Win prob. (TiCo)
|
52.27%
|
|
4.0
|
Forward strength
|
5.4
|
|
5.2
|
Midfield strength
|
8.5
|
|
0.9
|
Ruck strength
|
9.9
|
|
4.2
|
Backline strength
|
5.9
|
|
46.84%
|
Win prob. (APP)
|
53.16%
|
|
33.32%
|
Win prob. (Ensemble)
|
66.68%
|
|
Sydney v Gold Coast
|
|
11 April @ Norwood Oval
|
|
+45
|
HGA
|
|
|
+67
|
Elo rating
|
+109
|
|
50.52%
|
Win prob. (Elo)
|
49.48%
|
|
27.3
|
xSS
|
27.0
|
|
51.93%
|
Win prob. (xSS)
|
48.07%
|
|
+1.375
|
Team elite tier
|
+1.235
|
|
+0.080
|
Team standard
|
+0.118
|
|
−0.813
|
Team subpar tier
|
−0.750
|
|
48.02%
|
Win prob. (TiCo)
|
51.98%
|
|
5.8
|
Forward strength
|
6.0
|
|
5.8
|
Midfield strength
|
5.7
|
|
8.8
|
Ruck strength
|
5.8
|
|
6.6
|
Backline strength
|
5.3
|
|
48.89%
|
Win prob. (APP)
|
51.11%
|
|
52.24%
|
Win prob. (Ensemble)
|
47.76%
|
|
Hawthorn v Footscray
|
|
11 April @ Adelaide Oval
|
|
+4
|
HGA
|
|
|
+144
|
Elo rating
|
+172
|
|
46.41%
|
Win prob. (Elo)
|
53.59%
|
|
28.0
|
xSS
|
31.1
|
|
32.02%
|
Win prob. (xSS)
|
67.98%
|
|
+1.161
|
Team elite tier
|
+1.308
|
|
+0.091
|
Team standard
|
+0.119
|
|
−0.767
|
Team subpar tier
|
−0.958
|
|
44.19%
|
Win prob. (TiCo)
|
55.81%
|
|
6.9
|
Forward strength
|
5.1
|
|
5.6
|
Midfield strength
|
7.3
|
|
4.0
|
Ruck strength
|
5.9
|
|
6.3
|
Backline strength
|
4.8
|
|
51.77%
|
Win prob. (APP)
|
48.23%
|
|
50.75%
|
Win prob. (Ensemble)
|
49.25%
|
|
Geelong v West Coast
|
|
12 April @ Norwood Oval
|
|
+69
|
HGA
|
|
|
+105
|
Elo rating
|
−234
|
|
91.27%
|
Win prob. (Elo)
|
8.73%
|
|
22.4
|
xSS
|
20.7
|
|
64.06%
|
Win prob. (xSS)
|
35.94%
|
|
+1.241
|
Team elite tier
|
+1.119
|
|
+0.092
|
Team standard
|
+0.033
|
|
−0.813
|
Team subpar tier
|
−0.785
|
|
60.76%
|
Win prob. (TiCo)
|
39.24%
|
|
6.7
|
Forward strength
|
5.4
|
|
5.3
|
Midfield strength
|
5.5
|
|
4.1
|
Ruck strength
|
3.1
|
|
6.6
|
Backline strength
|
4.4
|
|
58.32%
|
Win prob. (APP)
|
41.68%
|
|
77.54%
|
Win prob. (Ensemble)
|
22.46%
|
|
GWS v Richmond
|
|
12 April @ Barossa Oval
|
|
−57
|
HGA
|
|
|
+49
|
Elo rating
|
−211
|
|
76.22%
|
Win prob. (Elo)
|
23.78%
|
|
20.9
|
xSS
|
19.3
|
|
70.52%
|
Win prob. (xSS)
|
29.48%
|
|
+1.191
|
Team elite tier
|
+1.144
|
|
+0.091
|
Team standard
|
+0.057
|
|
−0.668
|
Team subpar tier
|
−0.919
|
|
56.58%
|
Win prob. (TiCo)
|
43.42%
|
|
4.9
|
Forward strength
|
4.0
|
|
6.7
|
Midfield strength
|
4.6
|
|
4.8
|
Ruck strength
|
5.9
|
|
6.7
|
Backline strength
|
4.4
|
|
56.37%
|
Win prob. (APP)
|
43.63%
|
|
68.88%
|
Win prob. (Ensemble)
|
31.12%
|
|
Port Adelaide v St Kilda
|
|
12 April @ Adelaide Oval
|
|
+81
|
HGA
|
|
|
−111
|
Elo rating
|
−32
|
|
50.19%
|
Win prob. (Elo)
|
49.81%
|
|
23.1
|
xSS
|
22.1
|
|
50.15%
|
Win prob. (xSS)
|
49.85%
|
|
+1.367
|
Team elite tier
|
+1.285
|
|
+0.061
|
Team standard
|
+0.090
|
|
−0.832
|
Team subpar tier
|
−0.850
|
|
42.81%
|
Win prob. (TiCo)
|
57.19%
|
|
5.9
|
Forward strength
|
5.1
|
|
5.5
|
Midfield strength
|
5.2
|
|
6.0
|
Ruck strength
|
6.1
|
|
5.9
|
Backline strength
|
5.7
|
|
53.60%
|
Win prob. (APP)
|
46.40%
|
|
62.38%
|
Win prob. (Ensemble)
|
37.62%
|
Tips are based on five models:
- A team rating model (Elo) that tracks performance over time. 1,500 is the league average Elo rating. Home ground advantage (HGA) represents (a) the difference in experience between the two teams at the game’s venue for the current season and the two before, and (b) how far the teams have to travel to the venue, penalising long distance travel.
- An expected scoring shots model (xSS) that predicts each team’s scoring shots based on past offensive and defensive performance, and uses the difference between teams to estimate win percent.
- A tiered contribution model (TiCo) that assesses the difference in how well elite (top 25%) players, standard players, and subpar (bottom 25%) players perform between teams. A higher standard provides more support for the elite players, and a larger and worse subpar tier is a drag on team performance.
- An aggregated player performance (APP) model that predicts the relative strength of each line against the opponent’s by modelling player performance and building team strength from the player level up.
- An ensemble that uses the four above models, and their confidence, to give an overall estimate. More accurate models are weighted more strongly.