Chess.com cheat engine — v16.1 (ACPL governor + rating-drift fixes)
REXXX.MENU v16 — Chess.com analysis + auto-play userscript
Big update. v16 is mostly about behavioral realism — making the script play less like an engine and more like a tilted 1800. UI got a real polish pass too.
What's new in v16
Critical-position bias — errors no longer fire uniformly across all moves. Quiet positions get played near-perfectly, tactical/critical positions get the human treatment (occasional misses).
Shallow-depth picks — sometimes prefers a shorter-PV alternative when the engine's best move requires 8+ ply of forcing calculation that no 1800 would actually find.
Tactical motif blindness — specifically misses zwischenzug, deflection sacrifices, backwards knight moves, and long-diagonal threats more often. These are the patterns humans drop the most.
Endgame slips — elevated error rate in technical endgames. K+P endings get +45% errors, R+P +30%, simple endings +20%. Mirrors how 1800s actually botch winning positions.
Book-exit pause — first move after leaving prep takes ~2.4x normal think time. Real players stop and think when they're out of memorized lines.
Time-bank curve — bell-shaped think-time distribution peaking around move 22 (the critical middlegame), tapering off in the opening and endgame. No more uniform timing.
Forced-move recognition — when there's only one good move (others losing 80+cp), plays it in 220-750ms like a human would.
King-safety asymmetry — thinks ~45% longer when our eval drops (defending) versus when we're attacking. Human defense > human offense.
Idle mouse behavior — during long thinks, the cursor produces tiny gaussian drifts and occasional piece hovers instead of sitting perfectly still. No more dead cursor between moves.
Right-click annotations — ~6% chance per long think to draw a red circle or arrow on the board, like real players do when they're calculating.
Post-game review — after a game ends, ~18% chance to linger on the board "reviewing" before re-queueing. Sometimes peeks at the opponent's profile.
Engine eval cache — caches FEN evaluations for 10 minutes so we don't ping the API for repeated positions (transpositions, takebacks, etc.). Cuts network fingerprint significantly.
Stealth mode — opt-in console-log silencing for users worried about console monitoring.
UI overhaul
Panel widened to 360px so the tab bar doesn't wrap anymore.
Pill-style sliding toggles instead of tiny checkboxes. The whole row is now the click target — you don't have to aim at the pill.
Collapsible sections in the SAFETY tab. The v15.1 Account section folds up by default so the v16 toggles are immediately visible.
Active-features badge in the header showing something like 28/33 — at-a-glance read on how much humanization you currently have enabled. Tints red below 50%, yellow below 80%, green above.
Eval bar in the MAIN status box. Gradient strip with an animated marker showing position eval mapped to ±5 pawns.
Bigger fonts, stronger section dividers, custom scrollbar, cleaner reset button.
Full feature list (everything the script does)
Engine chess-api.com (Stockfish 18.1) primary, stockfish.online (SF16) fallback, local SF10 as last resort. Per-game engine rotation so the tactical fingerprint varies. Adjustable depth with dynamic depth based on position complexity.
Move sources Lichess opening book, Syzygy tablebase for ≤7-piece endgames, Lichess explorer for human-played moves in given positions.
Auto-play Auto-move execution with human mouse physics (bezier curves, gaussian jitter, hardware-persona-aware drag speed). Auto-queue between games with configurable break delay and session caps. Auto-resign and auto-lose for safety.
Humanization (v15 base) Streak limiter, accuracy clustering, weakness profile (per-account), player-move database, timing-accuracy coupling, anti-correlation cap on top-move rate, change-of-mind hesitation, telemetry noise (fake hovers, canceled premoves, UI clicks during opponent's turn), random AFK pauses, time-pressure accuracy drop, opponent rating adaptation.
Account-level (v15.1) Warmup mode (-350 ELO for first 20 games, ramps up). Lifetime winrate targeting (default 52%) across sessions. Hard repertoire (each account picks 1-2 openings per color and sticks to them). Tilt after loss (worse play + slower timing for 2-4 games following a defeat). Smart premove gating (only premoves on forced replies or obvious recaptures). Time-control lock (refuses to auto-queue a different TC mid-session). Messy resignation (hesitates, sometimes blunders first, sometimes holds lost positions). Hardware persona (per-account mouse/trackpad/tablet personality).
Behavioral realism (v16) All 13 features listed above.
Visuals Best-move arrows with configurable opacity, threat highlighting, eval bar, real-time engine badge.
Recommended setup
Use an alt account. Always. No script is undetectable, this one included.
Target rating ≤ 1800. Higher ratings invite review faster.
Cap at 4 games/hour, 5 games/session.
Leave all v15.1 + v16 toggles ON. They're the reason this lasts longer than scripts that just send best-move.
When you switch Chess.com accounts, click RESET in the SAFETY tab to clear stored repertoire/hardware/win-history.
Vary play across days. Don't grind 50 games in one sitting.
Hotkeys
A — toggle auto-play X — toggle stealth (hide panel) R — reset to defaults
Known caveats
Chess.com's fair-play system runs offline analysis after games. With aggressive use (high rating target, max games/hour, all toggles off) accounts get flagged in days. With conservative defaults, weeks. There is no "safe forever" — only "less suspicious."
Bug reports welcome. Feature suggestions also welcome but I won't add anything that increases detectability for the sake of convenience.
