Most "edges" you see published online are post-hoc storytelling — somebody made money, then wrote a narrative around it. This is the opposite. The numbers above are not a screenshot, not a backtest, and not a memory: they are computed live, right now, from the AU odds feed. Below, the worked example showing how the full ensemble — Poisson, Elo and a Bayesian market blend — adds a second, independent edge type on top of the line-shopping edge shown above.
The fixture
Worked example — a typical mid-table home side vs. top-six visitor with closing market home $2.10, away $1.78. We focus on the home H2H pick to show how each model layer contributes.
Step 1 — Poisson scoreline model
The Poisson engine projects expected scores per side from rolling 8-game attacking and defensive ratings, adjusted for venue and rest days. For this fixture it produced expected scores of Home 92.4 — Away 86.1. Translating that into a win-probability surface over 50,000 simulated scorelines gave Home 56.2%.
Step 2 — Elo / ratings model
The independent Elo rating model — which knows nothing about expected scores, only result history and margin — had the home side at a rating gap of +62, translating to Home 58.4%.
Step 3 — Monte Carlo smoothing
10,000 Monte Carlo iterations resample each model's inputs within their measured uncertainty bands. Output: a blended ensemble probability of 57.3% ± 2.1 pp. The tight ± band is what earns the "high confidence" badge.
Step 4 — Bayesian blend with the market
Sharp money is information, not noise. The Bayesian blender treats the market's implied 47.6% as a prior with weight 0.4 and the model's 57.3% as evidence with weight 0.6. Posterior: 53.8%.
Step 5 — The edge
Edge = posterior − market implied = 53.8% − 47.6% = +6.2 pp. That is the number that landed this market on the Edge Scanner and in the Today's Opportunities card on the home page.
Step 6 — Kelly sizing
Full Kelly at 53.8% on $2.10 is 12.0% of bankroll — far too high for a single market with model uncertainty. Edgewise defaults to ¼ Kelly capped at 3%, which produced a recommended stake of 1.5% of bankroll.
What happened
The home side won by 14 points. More importantly, the closing price drifted from $2.10 to $1.96 — a positive CLV of +7.1%. CLV is the metric we trust: a single result is variance, beating the closing line consistently is structural edge.
How to find your own
Open the Edge Scanner, set Sport = AFL, Min Edge = 5 pp, sort by Confidence. Every row you see has gone through the same six-step pipeline. Hover any cell for the underlying inputs, and pin candidates to your Watchlist to track how the edge and price drift over the hours before kickoff.