HUMAN INTUITION. STRUCTURED DECISIONS.
Three in a row.
A window into JEV.
A familiar game. A measurable opponent. Every move has a record.
Choose your settings and start a game.
Guest matches are unranked.
POSTGAME LAB
Every move, measured.
Exact minimax labels are calculated after selection. They are not supplied to JEV.
| Ply | Player | Cell | Source | Before → after | Regret | Optimal | Latency |
|---|
Raw requests, responses and hash-linked event timeline
VERIFIED RESULTS ONLY
Leaderboard
| Rank | Player | Result rate | Wins | Draws | Losses | Games |
|---|
Loading verified results…
THE MEASUREMENT LAYER
Nothing hidden.
Nothing invented.
Your completed server matches. Live JEV, forced moves and fallback are reported separately.
Decision quality by source
Optimality accepts every equally optimal move. Regret measures loss of game-theoretic value.
| Source | Moves | Optimal | Regret | Avoidable loss |
|---|
Placement map
Counts, not win probabilities. Sources are never silently combined.
Latency & reliability
Server wall time includes transport. Human dwell includes network delay and time away from the game.
Outcome calibration
JEV-profile candidate predictions vs exact minimax labels. A repeated board is another observation, not an independent game.
Quality by move number
Combined sources; use the source table for attribution.| Ply | Moves | Optimal | Mean regret | Missed win / opportunities | Missed block / opportunities |
|---|
Configuration cohorts
Model, profile, code revision and policy are fingerprinted. Compare like with like.
Exhaustive reference audit
This packaged report traverses the complete reachable rules space. It is not a live JEV benchmark.
Open the generated reference report ↗REPRODUCIBLE BY DESIGN
Your match record.
Inspect a replay, download the original decision evidence, or verify the event chain offline.
| Started | Profile | You | Result | Eligibility | Moves | Replay |
|---|