180 correct / 213 model calls
| Direction rule | Accuracy | Correct |
|---|---|---|
| Model | 84.5% | 180/213 |
| Always Under | 85.0% | 181/213 |
| Always Over | 15.0% | 32/213 |
- Games
- 26
- 95% game-cluster interval
- 79.7%–89.6%
Evidence-first sports intelligence
Forecasts you
can inspect.
One clear view of sportsbook prices, prediction markets, player projections, and model evidence—built to show what is live, what is stale, and what is still unknown.
Evidence unavailableNFL
—games in view
—linked Kalshi props
—two-sided markets
Checking model scanAwaiting capture
NBA + NFL + CFBSportsbooks · Kalshi · PrizePicksPregame research only
Retrospective accuracy in context84.5%.
84.5%.
Then 82.0%.
Compare with constant directions
The model trailed always choosing Under in 2025 and scored higher in 2026 on the same selected records. These historical comparisons do not establish a profitable edge.
100 correct / 122 model calls
| Direction rule | Accuracy | Correct |
|---|---|---|
| Model | 82.0% | 100/122 |
| Always Under | 77.0% | 94/122 |
| Always Over | 23.0% | 28/122 |
- Games
- 15
- 95% game-cluster interval
- 72.8%–89.8%
Rule: at least 60.0% probability on the called side of a fixed 4.5 reception threshold, using only prior games.
Both constant directions use the same retrospectively selected records and fixed threshold. Archived sportsbook prices are unavailable; prospective profitability remains unproven.
Inspect every forecastSee the market.
Understand the gap.
Keep the proof.
linework / live markets
L
Last published NFL coverageEvery state has a reason.
Games—
Markets—
Two-sided—
VENUE STATUS
Kalshi Checking
Sportsbooks Checking
PrizePicks Checking
EVIDENCE POLICYNo price without provenance.
Source, time, identity, fees, and settlement logic stay attached.
Try a real past-game replayA clear forecast.
A clear forecast.
A result you can check.
See the player, direction, model evidence and verified final stat together. Compare Week 1 replays, including a miss, without leaving the card.
PAST-GAME DEMO · NFL WEEK 1
The forecast. The final result.
SF vs LAR · 2026-09-10
Brock Purdy
Over 200.5passing yards
A fixed research threshold, evaluated against the final game stat.
Model baseline70.0%Above the research threshold
History behind it24 gamesBefore the target game date
EvidenceFinal stat verifiedExact player and game match
WHY THE MODEL LEANED OVER
Brock Purdy exceeded 200.5 yards in 17 of 24 prior games. The median was 268.5 yards, 68.0 above the threshold.
See the history behind the forecast
17 above / 7 below
70.8% observed rate across the full sample.
3 of 5 most recent games above
Recent results were weaker than the full history.
| Date | Opponent | Passing yards | vs 200.5 |
|---|---|---|---|
| 2026-01-03 | SEA | 127 | Below |
| 2025-12-28 | CHI | 303 | Above |
| 2025-12-22 | IND | 295 | Above |
| 2025-12-14 | TEN | 295 | Above |
| 2025-11-30 | CLE | 168 | Below |
The model smooths the observed rate to 70.0% using (17 + 0.5) / (24 + 1). This is a historical-frequency baseline; it did not add an opponent, injury or teammate adjustment to this replay.
Inspect published prior-game statsReplay method and limitations
The retained replay used games from the 2024 season onward, an exclusive game-date cutoff and up to 40 prior games. The target game's final stat was used only for grading. This is a retrospective research threshold, not a recorded pregame sportsbook recommendation.
Historical sportsbook prices, a retained uncertainty interval and timestamped injury evidence are absent, so this replay cannot establish betting value, stake, ROI or CLV. Demo records remain separate from live picks and performance totals.
Retained final-stat source commitment1cefdae759629bef7ec6345d4f9e87edcdee907a934f0c04503297c414ebff13How it works
Signal without the theatre.
The interface stays calm because the system underneath is strict. Each stage can block a result instead of filling the screen with false certainty.
01
Observe the offer
Capture the line, price, event, player identity, and provider timestamp as one piece of evidence.
02
Resolve the market
Match the exact outcome across sportsbooks, Kalshi contracts, and projection boards.
03
Estimate honestly
Use only history available before the event and retain the model version and uncertainty.
04
Gate the result
Freshness, fees, liquidity, rules, and settlement coverage decide what reaches the dashboard.
Unknown is a valid answer.
Linework does not turn missing outcomes into performance, stale quotes into opportunities, or public projections into invented payouts.
No order execution Immutable evidence Explicit provider health
Ready when the data is.Read the market
Open the dashboard