Y
You
Choose your opening0 / 71safe cells
Click: reveal · Right-click / F: flag · C / double-click: chord
HUMAN VS STRUCTURED INTELLIGENCE
Minesweepervs JEV
Two independent boards. The same starting square. First to clear wins.
MATCH MODEUnofficial practice
Click a square on your board to open the same starting coordinate on both boards.
J
Opponent
Waiting for your opening0 / 71safe cells
Independent mine layout. No hidden information is shared.
YOUR COMPLETION0%
SAFE-CELL LEAD0
YOUR COMMANDS0
OPPONENT COMMANDS0
VERIFICATIONNot started
Progress, move by move
Only visible information is shown during an active match.
YouOpponent
Accessible progress data
Sealed until game over. Layout analysis, flag accuracy, safety-forecast calibration, and mine-based heatmaps are available only after the match ends.
Post-game analytics
Complete a match to unlock the replay-derived report.
Player comparison
Opponent quality & reliability
Spatial analysis
You
Opponent
Darker cells indicate lower values; brighter cells indicate higher values. Full numeric matrices are included in the JSON export.
Safety-forecast calibration
Model safety probabilities only. An unconfigured or purely deterministic opponent has no model calibration samples.
Per-action journal
Structured opponent decisions
Complete metric explorer
Your match history
Owned records only. Filtered by the selected board and reasoning difficulty above.
Verified leaderboards
Ranked after 20 eligible matches. Versions, board sizes, and difficulties are never mixed.
Two boards. One race.
Choose your first square. That coordinate opens on both boards, with no mines in its surrounding neighborhood. Layouts are independent, so an opponent reveal does not reveal your mines. Clear every safe cell first to win. Flags are optional annotations, not points.
Controls
Click or press Enter/Space to reveal. Right-click or press F to set or remove a flag. Use the Reveal/Flag switch on touchscreens. Press C or double-click a revealed number to chord: its adjacent flag count must match that number. Incorrect flags can make a chord explode. Use arrow keys to move keyboard focus.
Race outcomes
A mine ends that board, not automatically the race. The other player must still clear their board. Two clears in the same 100-millisecond finish interval are a draw. Two exploded boards, or neither board clearing by 15 minutes, are a draw. Resignation or 30 seconds disconnected from a started server match is a loss. Ready matches expire after two minutes.
JEV and local play
JEV receives visible board facts and bounded candidate actions, never hidden mines. The solver handles proofs and exact arithmetic; JEV supplies structured candidate judgments. Difficulty changes reasoning depth, not hidden information. Without a TypeSafe key, the opponent is explicitly a local heuristic. Offline practice is entirely browser-local and never ranked.
Trust and rankings
Server games keep seeds, mine layouts, official time, and outcomes on the server. Discord identifies the player; a verified slash-command launch establishes server and channel context. Ranked matches require Discord, configured JEV, verified replay, and no fallback or material scheduling miss. Clear-win rate is wins divided by all eligible matches, including draws. A loss or draw breaks the win streak.
Analytics and privacy
The report records accepted commands, progress, timing, deduction coverage, flag precision and recall, action yield, board structure, 3BV, lead changes, heatmaps, provider reliability, token usage, and calibration where forecasts exist. It does not record keystrokes outside the game, device fingerprints, or mouse trails. Raw operational logs omit credentials and IP addresses. Post-game truth-derived metrics never appear in an active match.
A legal replay proves rule consistency, not unaided human play. Board randomness is not a guarantee of equal puzzle difficulty. “Uncertain” means unresolved by the bounded solver, not proof that guessing was logically unavoidable.