AFL Expert Tips, Graded: Every Public Model Ranked
Most footy tips are never checked. We grade them. This is every public AFL model on Squiggle with 80+ graded tips - 31 of them, including the betting market - ranked by 2026 home-and-away tipping accuracy against actual results. Our own model's back-test is shown separately, because it covers fewer games. No spin, just the scoreboard.
2026 AFL model leaderboard
Ranked by tipping accuracy across the 25 rounds of the home-and-away season (rounds 0-24). Finals are not included. "Bits" is the information-score metric the AFL analytics community uses to reward confident correct tips and punish confident wrong ones - a sharper measure than raw accuracy. Our model (highlighted, not ranked) is a walk-forward back-test: trained only on 2019-2025, then run over 116 games from rounds 0-14 after they were played. The public models tipped before each game, so the two are not like for like.
| # | Model / tipster | Correct | Accuracy | Bits |
|---|---|---|---|---|
| 1 | Drop Kick Data | 157/207 | 75.8% | 51.1 |
| 2 | Holy Grail Ratings | 157/207 | 75.8% | 35.3 |
| 3 | s10 | 155/207 | 74.9% | 53.8 |
| 4 | Punters (the betting market) | 154/207 | 74.4% | 52.6 |
| 5 | Glicko Ratings | 154/207 | 74.4% | 48.8 |
| 6 | ZaphBot | 154/207 | 74.4% | 51.6 |
| 7 | Wheelo Ratings | 154/207 | 74.4% | 51.1 |
| 8 | Aggregate | 153/207 | 73.9% | 51.2 |
| 9 | What Snoo Thinks | 153/207 | 73.9% | 48.8 |
| 10 | AFLalytics | 152/207 | 73.4% | 52.7 |
| 11 | Informed Stats | 152/207 | 73.4% | 49.9 |
| 12 | The Wooden Finger | 151/207 | 72.9% | 52.4 |
| 13 | Don't Blame the Data | 151/207 | 72.9% | 54.7 |
| 14 | Hyperion | 144/198 | 72.7% | 43.7 |
| 15 | Massey Ratings | 132/182 | 72.5% | 46.4 |
| 16 | The Cruncher | 147/203 | 72.4% | 50.0 |
| 17 | Cheap Stats | 149/206 | 72.3% | 40.0 |
| 18 | Matter of Stats | 149/207 | 72.0% | 52.7 |
| 19 | HBin | 149/207 | 72.0% | 48.7 |
| 20 | PlusSixOne | 148/207 | 71.5% | 47.1 |
| 21 | Stattraction | 148/207 | 71.5% | 50.5 |
| 22 | AFL Lab | 148/207 | 71.5% | 46.8 |
| 23 | The Footycast | 148/207 | 71.5% | 41.6 |
| 24 | footycharts | 148/207 | 71.5% | 50.1 |
| 25 | Winnable | 148/207 | 71.5% | 50.2 |
| 26 | Graft | 144/207 | 69.6% | 46.4 |
| 27 | Live Ladders | 144/207 | 69.6% | 46.3 |
| 28 | Elo Predicts! | 144/207 | 69.6% | 46.4 |
| 29 | In The Game | 144/207 | 69.6% | 48.8 |
| 30 | AFL Scorigami | 143/207 | 69.1% | 41.0 |
| 31 | Squiggle | 141/207 | 68.1% | 43.1 |
| - | TippingEdge AI ★ (back-test, rounds 0-14, not ranked) | 86/116 | 74.1% | - |
Model tip data via the Squiggle API (credit to Squiggle for aggregating the public AFL model community). This table is rebuilt only when we publish an AFL round, not automatically, so it can lag. "Punters" is Squiggle's tag for the betting-market favourite.
Why beating the market is the real test
Any tipster can look good in a season where favourites win. The honest benchmark is the betting market itself - the "Punters" line above, which tips whichever team is favourite. Beating it consistently is hard, because the market already prices in everything public. On the 116 games from rounds 0-14 that our back-test covers, the betting market got 83 right and our model got 86.
The other thing accuracy alone hides: price. A model that tips winners at $1.20 makes you nothing; one that finds winners at $2.00 prints. That is why our weekly tips publish a probability for every game, so you can compare it to the odds and bet only when there is value - use the value checker to do exactly that.
What about media expert tips?
Newspaper and TV panels - The Age, ESPN, Fox Footy, SEN and the rest - publish weekly tips for the office competition, but they rarely publish a graded season record, and they tip winners rather than prices. The model leaderboard above is the data-backed version of the same question: who actually tips well? When a media expert diverges from both the market and the models, that is the interesting tip - everything else is noise around the favourite.
More from TippingEdge
FAQ
Who is the most accurate AFL tipster in 2026?
Of the 31 ranked models above, Drop Kick Data and Holy Grail Ratings lead on raw accuracy through round 24 of the home-and-away season, on 75.8%. Our model is not ranked: its back-test scored 74.1% on 116 games from rounds 0-14. Rankings shift during the season, and we rebuild this table only when we publish an AFL round.
Are AI models better than expert tips?
The leaderboard is dominated by statistical models, and 3 of them finished ahead of the betting market in the 2026 home-and-away season. But accuracy is only half the story: a model that tips short-priced favourites adds no betting value. Look for tips that come with a probability you can compare to the odds.
How is your model graded?
By a walk-forward back-test: trained only on 2019-2025 data, then run over 116 games from rounds 0-14 of 2026 after they were played. That is not the same as the public models, which tipped before each game. It is detailed on our accuracy page.
Where does the data come from?
Public model tips and their grading come from the Squiggle API, which aggregates the AFL prediction-model community. Our model's back-test is computed in-house. The market line ("Punters") is Squiggle's favourite-based benchmark.
Get the model's tips by email
The model picks every NRL and AFL game, every round. Join free and we'll email its tips for each round of the 2027 season, the week each round opens.
Free. 18+ only. One email per round in season, unsubscribe any time. See our privacy policy.