ScoreboardEdge prices every NFL, NBA and NHL game itself, checks that price against the bookmaker's, and turns a genuine gap into a recommendation — reviewed, logged and graded, and delivered to members of the Client Portal. No money changes hands.
Each league gets its own Elo ratings, its own live board, and its own news — built and rolled out on its own timeline.
Elo ratings replayed from real nflverse results, 1999–2026, updating after every game. Full live board of upcoming games with model vs. market prices.
Pipeline is live, but historical ratings are currently anchored at Dec 2022 while we source a compliant 2023–2026 results bridge ahead of the Oct 3 tip-off.
NHL coverage is confirmed for ScoreboardEdge — the odds and ratings pipeline is next in line to be built, after NFL/NBA. Not live yet.
Every ScoreboardEdge market boils down to one of these two questions.
Who wins, straight up. The model produces a win probability for each team, compared against the market's own no-vig probability — the gap between the two is the "edge."
Wins by how much — the point spread in NFL/NBA, the puck line in NHL. Reference market lines are already being logged, but our margin model isn't calibrated enough yet to trust for live picks.
Tiers and pricing are being finalized. Check back soon for details.
Every ScoreboardEdge recommendation comes from the same three-step comparison — here's exactly what happens before a call is made.
Our Elo model converts each team's current strength into a win probability for every scheduled game — no bookmaker input involved yet.
That probability is checked against the bookmaker's own price, converted to a no-vig ("fair") probability so their built-in margin doesn't distort the comparison.
When the gap between the two — the "edge" — crosses a threshold, ScoreboardEdge issues a recommendation: which side, at what price, and how large the edge is.
Reviewed before it reaches you. Every recommendation the model produces is checked by a human before it's published to the Client Portal — see the Admin → Client flow in our portal design. Nothing here is a raw, unreviewed model output.
The model. Recommendations are scored with a bare-Elo model (margin-of-victory k-factor, seasonal regression) — the version that came out of our Phase 3 proof-of-concept backtest. A more complex three-model blend was also tested and did not make the cut: it underperformed on real backtests and isn't wired into live recommendations.
The honest headline. In backtesting on real historical data, bare Elo did not beat the closing market outright for either NFL or NBA — that's exactly where our first sports model started too. Every game is logged at four edge thresholds so the live sample can answer, honestly, whether that changes.
The data. Live odds and results come from public sportsbook feeds; historical ratings are bootstrapped from real play-by-play and results archives for each league as it's built.
The rules. No staking decision is pre-selected. All stakes are one notional unit. No money changes hands — this is a public research and validation project, not a betting product.