We’d like to thank everybody who entered our Betfair Brownlow Datathon.

Final results of all submissions can be found below. All Prize Winners will be contacted ASAP.


Final Leaderboard

Check out the Final Leaderboard of our Datathon Competition below.

Position SUBMISSION_NAME MSE
0 MD_BetFair_Brownlow 0.951619048
1 shevy_tothelevy_SF 0.959768707
2 funk_all 0.964432245
3 yosemite_sam 0.964954286
4 corey_AFLp 0.982013605
5 keesh-AFLCA 0.997278912
6 woo_HS 1.01244898
7 JMod 1.017918367
8 gluteus_maximus 1.019218912
9 C 1.027210884
10 SG1 1.038975137
11 CF Brownlow Datathon 1.057312925
12 dude_wheres_my_karla 1.084547347
13 Brownlow-What-Of-It 1.085168299
14 craigs a-list 1.08537415
15 Skipper 1.11292517
16 Fazer 1.132503401
17 Marcello Finity avg3 external 1.142630494
18 rompingwins_brownlow 1.164305714
19 RowanMarshall3Votes 1.196017955
20 johnny’s gto strats 1.210678147
21 fagan4eva 1.211852573
22 Zeus 1.237999142
23 Confused Roberta 1.242294347
24 CurryBoyModel 1.257211156
25 brownlowGIANT 1.275328299
26 Brownlo Magic 1.276275238
27 KingDangerPredicts 1.291955918
28 PoliteSlon 1.300680272
29 CA_Brownlow 1.30454433
30 Pharmacy Raider 1.305428586
31 ClintonKing 1.309783392
32 toby_greened 1.33527912
33 Euclid_AFL_Model_Express 1.356462585
34 Neil-Shyer 1.383836735
35 Model79 1.387755102
36 ordinal_logistic 1.39534214
37 PaulMcGrath_Brownlow2019 1.399319728
38 Pondmaster4000 – Copy 1.40140956
39 ChuggyChoices 1.412244898
40 PayneAndGain 1.488435374
41 Brownlow Datathon Entry – Michael Doyle 1.538285714
42 Edward Tasker’s Model 1.554686995
43 Lucas Betfair-Brownlow-Datathon-Submission 1.567346939
44 DangerZone 1.595918367
45 Athomas 1.615731293
46 Posner 1.654802721
47 statlow84 1.656996527
48 Foxy 1.667772109
49 Magnificent 1.672132321
50 Roadrat 1.687115517
51 BF 1.703995835
52 The Brownlow Oracle 1.733645183
53 Paul-Brownlow-Datathon-Submission 1.755102041
54 Brownlow Submission AHenry 1.76462585
55 Wasted Youth 1.807961958
56 Linear 1.836734694
57 IG Brownlow Predictor 1.854323129
58 AnthonyMelino_Brownlow2019 1.895238095
59 HuRdle Predictions 1.895679574
60 LosTigres 1.905599314
61 Brownlow Prophecy 1.993197279
62 BucketsBest 1.997841294
63 DangerVsField 2.005996599
64 boosted ballers 2.088860544
65 Dustin Martinez 2.140136054
66 Sportpunter 2.147189092
67 BillSkeltonBrownlow2019Predictions 2.491156463
68 FuansForest2 2.538710131
69 Cabbage 2.58513238
70 The Statz Machine 2.614679835
71 Bontlow 2.729668203
72 MiddyMedal23 2.786653061
73 Bilsko affle red line 2.80170068
74 The Weekend Preview 2.8243999
75 Brownlow19-SJ 2.876190476
76 Double Up Danger 2.884095238
77 chaRlie_Betfair_Preds_2019 2.912256735
78 gb13faithless 2.922661712
79 prophit 2.948704529
80 submission_ric_porteous 2.981008193
81 maRtin_model 3.045081004
82 Its Lachies Year 3.123006531
83 Panj 3.157823129
84 Betfair-Brownlow-Datathon-Submission-GLM 3.253061224
85 API 3.364129949
86 DNN_preds 3.403061224
87 Poms Brownlow Data Model 3.465952344
88 Brownlow_Datathon_BZ 3.508843537
89 brownhigh 3.825076356
90 Smiths_Fault 4.010884354
91 Last Minute Model 4.617656882
92 CarlosModel 5.06122449
93 MultivAriate_Regression-Trained_brownlow_INdicator 5.146938776
94 Basic Regression LS with negatives removed 6.622194133
95 plant_power 6.926530612
96 fst_model 10.70641826
97 Betfair-Brownlow-Datathon-Submission-Squires 10.78321272
98 BocceGouger 14.11493692
99 LP Tiser 19.06258503

The Specifics

Below is information on what data will be provided and how the prize money will be divided and allocated.

Click here for the Terms and Conditions.

Player data dating back to the 2010 AFL season will be provided upon entering. This includes:

  • Contested and uncontested possessions
  • Effective disposals and disposal efficiency
  • Brownlow Votes
  • AFL Fantasy and Supercoach Scores
  • Time on Ground

The metric used to determine the winner will be the ‘Mean Squared Error‘ over all predictions, based on the actual Brownlow votes received compared to predicted Brownlow Votes received.

For example, the Betfair Data Scientists’ Brownlow Model predicted Tom Mitchell to poll 35.48 votes last season, while he only actually polled 28. Thus, the Squared Error for this individual prediction would was (35.48-28)^2=56.04. On the other hand, if our model had correctly predicted votes, the Squared Error would have been 0. The winner of the Datathon will be the participant with the lowest Mean Squared Error over all predictions.

It is important that answers are submitted in the format that is provided in ‘Betfair Brownlow Datathon Submission.csv’.

We will assume 0 predicted votes for any missing values. For example, if someone doesn’t provide a prediction for ‘P Dangerfield’ we will assume they predict 0, and mean squared error will be calculated accordingly.

Winners will be announced via email and on this Betfair Hub page after the Competition concludes on or around 24 September 2019.

Prizes will be awarded to the top 10 entries in the form of a loaded Mastercard/Visa gift card.

Betfair entries that are in the Datathon are not eligible for prizes.

RankPrize
1st$2,000
2nd$1,000
3rd$750
4th$500
5th$250
6th$100
7th$100
8th$100
9th$100
10th$100

Winners will be announced via email and on the Betfair website after the Competition concludes on or around 24 September 2019.

Final Submissions are due by 17:59 AEST on 23 September 2019.

The name of your prediction model will be published on the Betfair website and will be visible to members of the public. Please bear this in mind when naming your model. Offensive names will not be published.

Submissions must be in a CSV format, any submissions that are not in the correct format will not be accepted;

To submit your output, email your final submission to datathon@betfair.com.au.


FAQ's

  • Some of the players are missing. i.e. there is only one listing of A Brayshaw, when there should be two Brayshaws.

Player names are in the format {firstname initial}{surname} as some small merging of datasets was required to create the dataset for this competition. This means some players with identical names are not uniquely identifed. Don’t worry about providing a submission for these players; you can provide a prediction or no prediction at all and you will not be penalised as we will have to exclude these players from scoring. However, do not make changes to the submission file to try and distinguish between these players.



Go where the value is on the Exchange. Don’t settle for bad odds. Join Betfair.


Related Articles

How to use Machine Learning for the World Cup Datathon

Another ELO approach to modeling the World Cup from the Betfair Data Scientists.

How To Make A Predictive Model In 5 Easy Steps

Want to learn how to create your own predictive model using sports or racing data, but you don’t know ...

How to Build an AFL Model

A quant take us through how to build an AFL Model in an easy step by step guide.