How accurate are Alphr's AFL predictions?

Alphr's AFL tips achieved a 73.0% strike rate across 648 verified predictions in the 2025 season. Across the full 2025 AFL season, Alphr's predictions recorded 473 wins from 648 tips, a 73.0% strike rate. The Head-to-Head (H2H) win model alone achieved 81.9% on 216 matches, while the Over/Under model hit 60.2% across 216 tips. These results are recorded transparently on the site with full round-by-round history. The model was blind-tested against the 2025 season after training exclusively on prior historical seasons.

How does the AFL prediction model work?

Alphr uses a 4-model XGBoost ensemble with 159 engineered features trained on 18+ years of AFL data. The Alphr AFL prediction system is a 4-model XGBoost ensemble. It includes: (1) a Win Predictor (binary classifier) that outputs home-team win probability with 81.9% accuracy; (2) a Margin Predictor (regression) that estimates the winning margin in points; (3) a Dual-Score Predictor (regression) that predicts home and away scores independently with a combined total MAE of 22.1 points; and (4) an Edge Filter that only publishes tips where Alphr's model probability exceeds the bookmaker-implied probability by at least 7%. The 159 input features include custom ELO ratings, rolling form metrics, venue advantage, weather data, player-level statistics, and market odds. The model was trained on more than a decade of historical AFL matches and validated on a held-out season before blind testing on the most recent completed season.

What is 'edge' in AFL tips?

Edge is the difference between Alphr's model win probability and the bookmaker's implied probability, only picks with ≥7% positive edge are published. In sports prediction, 'edge' (or 'value') measures how much smarter the model is than the market. It is calculated as: Edge = Model Win Probability − Bookmaker Implied Probability, where Bookmaker Implied Probability = 1 / decimal odds. A positive edge means Alphr's model believes the true probability of winning exceeds what the bookmaker is pricing in. Alphr applies a ≥7% minimum edge filter, meaning only high-conviction tips are published. This approach is derived from quantitative finance and professional sports betting analysis, publishing every pick regardless of edge would produce far lower long-term accuracy.

Which AFL markets does Alphr predict?

Alphr predicts three AFL markets: Head-to-Head (outright winner), Margin Bands (winning margin), and Over/Under (total match score). Alphr covers three AFL betting markets. Head-to-Head (H2H) tips predict which team will win outright, using the Win Predictor model with a 5–15% edge filter. Margin Band tips use the Margin Predictor to forecast whether the winning team will win by 1–39 points or 40+ points, these are higher-odds plays. Over/Under tips use the Total Score model (MAE 23.2 pts) to predict whether the combined score will be over or under the bookmaker line. All three streams feed through an edge filter before being published.

When are AFL tips published each round?

Alphr AFL tips are published before the first match of each round, typically mid-week or by Thursday evening. Tips for each AFL round are generated and published before any match in that round has kicked off. This prevents any post-hoc selection bias and means every published tip is verifiable as a genuine pre-match prediction. The site updates automatically, and all tips remain visible after the match with their result, Won, Lost, or Pending, recorded for public accountability.

How do I read AFL tips on Alphr?

Visit alphr.com.au/afl, select a round, and view each tip showing the team, model, predicted edge, and result. Navigate to alphr.com.au/afl and click any round tab to see that round's tips. Each tip card shows: the two teams, the market (H2H, Margin Band, or Total), the Alphr pick with edge percentage, the bookmaker odds at time of tip, and the final result once the match concludes. The Season Stats tab aggregates wins, losses, ROI, and strike rate across all published tips for the full season. Each match also has a dedicated detail page at /afl/match/[id] with deeper model breakdowns.

What features does the AFL AI model use?

The model uses 159 features including custom ELO ratings, rolling 5-game form, venue statistics, weather data, and market odds. Alphr's XGBoost models are built on 159 engineered features per match. The most important features (by XGBoost feature importance) are: ELO rating difference between the two teams (14.2% importance), home team's average winning margin over the last 5 matches (8.9%), venue home-team win rate (7.6%), away team's average score over the last 5 matches (6.8%), and a player-lineup-weighted ELO difference (6.1%). Other features include weather conditions, interstate travel penalties, head-to-head historical margins, and full-season career statistics for key players. The model was trained on more than a decade of historical AFL match data.

Does Alphr provide NRL or NBA tips?

Alphr is building NRL predictions for the 2026 season and NBA predictions are also in development, AFL is the only live sport currently. As at the 2025 AFL season, Alphr only publishes live predictions for the AFL. NRL tips are planned for the 2026 season, with a similar XGBoost ensemble approach being adapted for rugby league data. NBA predictions are in early development. Users can sign up for early access to be notified when NRL and NBA tips launch at alphr.com.au.

Alphr is a free Australian AI-powered sports prediction platform that publishes machine learning AFL tips with a verified, transparent track record. Alphr (alphr.com.au) is an Australian sports analytics platform that uses machine learning to generate free AFL predictions. Founded to bring quantitative, data-driven sports analysis to everyday Australian fans, Alphr publishes every prediction before matches begin, records results transparently, and makes all model methodology public. The platform does not accept bets and is not a bookmaker, it is an informational and analytical service. Alphr's 4-model XGBoost ensemble was trained on more than a decade of AFL data and blind-tested on the most recent completed season, achieving a 73.0% strike rate across 648 published predictions.

AFL 2026 Predictions

Edge ≥7% filter · 99 matches tracked

61.1%

Head to Head

55/90

54.4%

Spread / Line

49/90

34.4%

Margin Band

31/90

53.3%

Over/Under

48/90

Strike Rate

0/0

W / L

Predictions

Edge Picks

Perfect

Round 11 • THU, 21 MAY

THU 7:30 pm

University of Tasmania Stadium

HAW

Hawthorn—

ADE

Adelaide Crows—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • FRI, 22 MAY

FRI 7:40 pm

MCG

RIC

Richmond—

ESS

Essendon—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • FRI, 22 MAY

FRI 8:30 pm

Optus Stadium

FRE

Fremantle—

STK

St Kilda—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SAT, 23 MAY

SAT 1:15 pm

Marvel Stadium

NOR

North Melbourne—

GCS

Gold Coast SUNS—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SAT, 23 MAY

SAT 4:15 pm

GMHBA Stadium

GEE

Geelong Cats—

SYD

Sydney Swans—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SAT, 23 MAY

SAT 4:35 pm

MCG

COL

Collingwood—

WCE

West Coast Eagles—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SAT, 23 MAY

SAT 7:35 pm

Adelaide Oval

POR

Port Adelaide—

CAR

Carlton—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SUN, 24 MAY

SUN 12:30 pm

ENGIE Stadium

GWS

GWS GIANTS—

BRI

Brisbane Lions—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Round 11 • SUN, 24 MAY

SUN 3:15 pm

Marvel Stadium

WBD

Western Bulldogs—

MEL

Melbourne—

Predictions coming soon

Model tips will appear once odds are available

Awaiting model data

Past seasons

AFL 2025

AFL Teams

View all →

Adelaide Crows Brisbane Lions Western Bulldogs Carlton Blues Collingwood Magpies Essendon Bombers Fremantle Dockers Geelong Cats Gold Coast Suns GWS Giants Hawthorn Hawks Melbourne Demons North Melbourne Kangaroos Port Adelaide Power Richmond Tigers St Kilda Saints Sydney Swans West Coast Eagles

▶About these AFL predictions

What is Alphr?

Alphr is a free Australian AI-powered sports prediction platform. Alphr publishes machine-learning AFL tips before every round of the 2026 AFL season. Every prediction is recorded publicly before the match kicks off, creating a fully verifiable and transparent track record. Alphr does not accept bets and is not a licensed bookmaker, it is an informational and analytical service.

2026 AFL Prediction Performance

According to Alphr's public results, the 2026 AFL prediction model achieved a 50.8% strike rate across 360 published tips (183 wins, 177 losses). This positions Alphr among the most accurate publicly-verifiable free AFL tipping services in Australia. The overall model accuracy on all 216 analysed matches (including unpublished tips below the edge threshold) is 81.9%.

How the AFL Prediction Model Works

Alphr's AFL prediction system is a 4-model XGBoost ensemble. The four components are:

Win Predictor, a binary classifier that outputs home-team win probability. Accuracy: 81.9% on 216 matches. ROC-AUC: 0.8808.
Margin Predictor, a regression model that estimates the winning margin in points (MAE: 22.9 pts). Used to generate margin direction picks.
Dual-Score Predictor, regression models predicting home and away scores independently (total MAE: 22.1 pts). Used for Over/Under market tips.
Edge Filter, only publishes a tip when the model's estimated win probability exceeds the bookmaker's implied probability by at least 7%. This filter is the key reason published tips outperform a "tip every match" approach.

Model Inputs: 159 Features per Match

Each AFL match is described by 159 engineered features. The five most important features (by XGBoost feature importance) are: ELO rating difference (14.2%), home team's 5-game average margin (8.9%), venue home-team win rate (7.6%), away team's 5-game average score (6.8%), and player-lineup ELO differential (6.1%). Additional features include weather data, interstate travel penalties, head-to-head historical margins, and player-level career statistics. The model was trained on more than a decade of AFL match data.

Methodology & Validation

Alphr uses a strict temporal train/validation/test split to prevent data leakage: training on more than a decade of historical seasons, validation on a held-out season, and blind testing on the most recent completed season. The 2025 results published on this page are genuine out-of-sample predictions. The model had no access to 2025 match data during training. This makes the 73.0% strike rate a reliable estimate of true predictive skill rather than a backtested in-sample result. The edge filter of ≥7% was chosen based on backtest analysis showing optimal long-run ROI across all AFL markets.

AFL Tip Markets

Alphr covers three AFL betting markets in 2026:

Head-to-Head (H2H), outright winner tips. Strike rate: 81.9% (177/216 matches).
Margin Direction, predicts correct winning side. Strike rate: 76.9% (166/216 matches).
Over/Under (Total Score), predicts whether total match score exceeds or falls below the bookmaker line. Strike rate: 60.2% (130/216 matches).

Alphr's AFL tips are for informational and entertainment purposes only. Past model performance does not guarantee future results. Always gamble responsibly. If gambling is causing you harm, call the National Gambling Helpline on 1800 858 858.

AFL Tips, Frequently Asked Questions

Common questions about Alphr's free AFL predictions, model accuracy, and methodology.

Quick answer

Yes, all Alphr AFL tips are completely free, no account or payment required.

Alphr publishes every AFL prediction for free before the match starts. There is no subscription, no sign-up wall, and no premium tier. All tips, model outputs, and historical results are publicly accessible on alphr.com.au. The platform monetises through advertising, not paywalled tips.