About
How WhatIsMyTip uses AI to help you make better footy tips.
Our Approach
WhatIsMyTip combines eight machine learning and statistical models with three heuristic strategies to generate footy tips. Our system analyzes historical results, team and player form, head-to-head matchups, home advantage, injuries, weather, and value opportunities to provide informed predictions.
We use a large language model via OpenRouter to generate human-readable explanations for each tip, helping you understand the reasoning behind our predictions. The model is configurable, so we can swap in newer options as they become available.
Models
Elo
A rating system that tracks team strength based on historical performance. Teams gain points for winning and lose points for losing, with the amount adjusted based on the opponent's strength.
- Adapts to team strength over time
- Accounts for home advantage
- Best for measuring long-term team quality
Form
Analyzes recent team performance over the last few games to identify current momentum. This model looks at recent wins, losses, and scoring margins to determine which team is currently playing better.
- Focuses on short-term trends
- Captures team momentum
- Best for teams in good or poor form
Home Advantage
A simple but effective model that leverages the historical tendency for home teams to win more often. It analyzes venue-specific win rates to predict outcomes based on where the game is played.
- Uses venue-specific statistics
- Simple and reliable
- Best for games at strong home venues
Value
Identifies undervalued teams by comparing historical win rates against perceived odds. This model looks for teams that perform better than their reputation suggests, finding value opportunities.
- Targets undervalued teams
- Uses long-term win rate data
- Best for finding upset opportunities
Weather Impact
Weather can dramatically change how a game of AFL is played. This model classifies the forecast into tiers — from clear conditions to heavy rain and wind — then compares each team's historical win rate in similar conditions to predict who copes best.
- Uses match-day weather forecasts
- Accounts for rain, wind and temperature
- Best for games played in extreme conditions
Injury Impact
Missing key players can swing a result. This model pulls each team's current injury list, weighs every absent player's importance using their recent statistics, and penalizes the side carrying the heavier injury burden.
- Quantifies the value of injured players
- Updates as injury lists change
- Best when star players are ruled out
Matchup
Some teams just have another's number. This model studies the head-to-head history between the two teams, blending overall matchup results with how each side performs at the venue — all weighted so recent meetings count for more.
- Combines head-to-head history (60%) with venue record (40%)
- Applies time decay to favour recent clashes
- Best for long-standing rivalries
Player Form
Rather than looking at team results, this model dives into individual performances. It aggregates advanced stats — score involvements, contested possessions, metres gained and pressure acts — over each team's last few games to gauge who is genuinely playing the better football right now.
- Measures team quality from player-level data
- Focuses on advanced performance metrics
- Best for spotting teams in peak form
Heuristics
Best Bet
Conservative, consensus-based picks with high confidence. This heuristic aggregates predictions from all models and selects the most confident consensus pick.
YOLO
Aggressive picks based on the highest confidence model prediction. This heuristic goes all-in on the model with the strongest conviction.
Weighted Tip
A data-driven tip that combines model predictions with learned weights. The underlying model is retrained weekly to keep its picks current.
Data Source
Match fixtures, results and team data come from the Squiggle API — an excellent resource for AFL statistics that makes this project possible. Historical player statistics and injury lists are sourced from AFL Tables and FootyWire, and match-day weather forecasts come from Open-Meteo.
Open Source
WhatIsMyTip is an open source project. The code is available on GitHub and we welcome contributions from the community.