AI Model v7: Performance Report
We've upgraded our core algorithm. The new v7 model utilizes advanced statistical probability scoring to identify high-value opportunities.
*Figures represent the results of a backtested simulation using 2025/26 historical odds data. Flat 1-unit stake applied to all generated selections.
Inside the Black Box: The v7 Math
Our model doesn't care about team names, reputation, or media hype. It sees football matches as a series of probabilities derived from raw performance data. While the exact weighting of our algorithm is proprietary, the core logic relies on three pillars:
1. Poisson Distribution
We calculate the likelihood of every possible scoreline (0-0, 1-0, 2-1, etc.) by analyzing attack strength vs. defensive solidity. This generates a pure "True Probability" percentage for a Home Win, Draw, or Away Win.
2. Expected Value (EV)
We compare our "True Probability" against the Bookmaker's Odds. If the Bookmaker implies a 40% chance, but our math says 55%, the EV is positive. We only bet when the math says the price is wrong.
3. Decay Factors
Recent form is weighted heavier than older form. A goal scored 4 weeks ago is mathematically less significant than a goal scored yesterday. The model applies a time-decay curve to ensure data is current.
4. xG Variance
We look beyond actual goals. If a team won 2-0 but had an xG of 0.4, the model flags them as "lucky" and may bet against them next time, anticipating a regression to the mean.
Strategy Update: Efficiency Protocol
In previous iterations, we used a strict "Diversity Rule" which prevented the same team from appearing in multiple accumulators on the same day. While this spread risk, it also diluted quality.
The v7 Update removes this restriction.
🚀 Unleashing Statistical Efficiency
If the model's math identifies a team as a "Diamond Value" pick (High Probability + High EV), it will now utilize that team across multiple bet types (3-Fold, 4-Fold, and 5-Fold).
The Trade-Off: This increases short-term variance. If that key team loses, multiple accumulators will fail on that day. However, our 2025/26 backtest data (below) demonstrates that this aggressive stacking strategy yields a significantly higher theoretical ROI over the course of a full season.
Simulated Performance Breakdown
The following data is based on a backtest of the v7 model. These figures represent the theoretical performance of the model if it had been active from August 2024.
💰 The Value Stream
The "Value" bets seek mispriced odds where the model detects the bookmaker has underestimated a team.
🛡️ The Safe Stream
The "Safe" stream targets raw probability, prioritizing high strike-rates over high odds.
⚠️ Risk Warning
Higher Variance: The 6-Fold Accumulators failed to generate a profit in our simulation (-100% ROI). The probability of predicting 6 correct outcomes is exponentially lower.
We advise users to treat all statistical analysis as a tool for information, not a guarantee of winning. Even the best mathematical model cannot account for red cards, injuries, or luck.
Summary: The v7 Formula
- ✅ Back the Value 3-Fold & 4-Fold: Highest simulated yield.
- ✅ Back the Safe 5-Fold: Strong correlation in testing.
- ❌ Avoid 6-Folds: High risk, negative expected value.