Change Log

/opt/rook-shared.env rook-bot.service

Set PAPER_MODE=false in rook-shared.env. Restarted rook-bot service. Bot now reports mode: LIVE.
Live strategies executing real orders on BYDFi: S7 Mode B (MA200 bounce), S10 (SFP 4H/12H/1D), S14 (Turtle Soup), BOJAN (key zones).
Engine config confirmed: live_mode: true, paper: false, all four strategies in live_strategies[].

Week 1 live results: +59.16 net, 67% win rate, 9 trades. System proved itself — time to run all approved strategies with real capital.

New script intercepts /pause and /resume Discord commands. Sends command to bot API, then polls /botstat up to 5× (2s apart) to verify state actually flipped.
Args: pause | resume | status. Returns ✅ Verified or ❌ FAILED with live mode confirmation.

Previous pause/resume had no feedback loop — bot showed conflicting state across health and command endpoints. This resolves ambiguity and ensures every toggle is confirmed before reporting back.

s5-fvg.js scalpbot.js

TP1 now uses 0.618 Fibonacci retrace of the 30-candle prior move (Golden Pocket).
Long: structLow + range × 0.618 | Short: structHigh - range × 0.618
Priority: Volume Profile node → Golden Pocket → nearest swing fallback.
TV alert payloads updated: sfp_type: "low_sfp" → "bullish", "high_sfp" → "bearish"

Institutional money exits at the 0.618 level — it's where smart money was positioned before the move. "Nearest pivot" was catching micro-resistances $22 away. Golden Pocket targets $200-400 moves on the same entries.

S11 Range Fade + S12 Quartile Swing specs added. AltFINS confirmed provides SUPPORT, RESISTANCE, ATR, Bollinger Bands per asset — Phase 3 fuel.

trade-manager.js engine.js s1–s9 strategy files data-state.js

New modular architecture: Rook Engine = entry manager, Trade Manager = exit manager.
Every strategy now exports evaluateExit(trade, liveState) — strategy owns its own exit logic.
Trade Manager runs every 15s, calls each strategy's exit function per open trade.
All exits are real-time tick-based — no candle close dependency.
Volume Profile added to data pipeline: POC, VAH, VAL calculated from MMT candle data.

1. TP direction inverted on shorts — TP was above entry on a short. Fixed: direction guard rejects bad TPs.
2. Stale SL — SL wasn't recalculating per trade. Fixed: 30-candle structural lookback, fresh every signal.
3. Spent FVG re-entry — Bot re-entered the same exhausted FVG zone. Fixed: 4-hour zone tracking.
4. RR floor missing — Trades with RR < 1.0 now rejected.

Was using EMA12/26 — never matched the agreed MA9/MA13 strategy. Fixed to simple MA9/MA13.
Also fixed fired → fires field name — S6 was never registering as fired in the engine.
Exit logic: dual trigger — MA9 crosses MA13 OR price crosses MA13 (whichever first).

All EMA calculations replaced with simple MA to match S6 and agreed strategy. Same dual trigger exit as S6.

bot.js s13-blood-diamond.js

Entry: 4H blood diamond + 1H red diamond within 4h = SHORT at 50% position (Leg 1).
Leg 2: MACD H2_DOWN + RSI >65 within 8h window = add remaining 50%.
Bear trap guard: don't add during counter-move candle still open.
Exit: POC → VWAP → VAL stacked targets | 24h time decay.

3 confirmed signals post-March 14: 3/3 hit rate. $1k+ move missed 2026-03-26 during downtime.

rook-docs/index.html rook-dashboard/server.js

Live NOC status bar on docs page. 8 components monitored: Rook Engine, Trade Manager, Main Bot, ScalpBot, Signal Router, Screener Bot, MMT WebSocket, Database. Green/yellow/red dots, auto-refresh every 30s. No manual updates.

Every incoming TradingView webhook alert now posts to Discord #trade-signals. Format: name, symbol, timeframe, price. No secrets. One line per alert — readable at a glance.

weekend-pause.sh weekend-resume.sh crontab

Friday 1:00 PM AZ (20:00 UTC) → all bots pause. Active trades finish, no new entries.
Sunday 2:00 PM AZ (21:00 UTC) → all bots resume for Asia open.
Manual override: !pause / !resume in Discord.

Llama analysis confirmed: weekends have higher false signal rate, lower volume, no institutional participants. CVD Divergence, MA crosses, SFP, BB Squeeze all unreliable on weekends.

Gateway crashed mid-session (CLI command connected back to gateway causing SIGTERM). 5-hour rebuild: wiped 3 containers, redid Discord bot config, restored from GitHub. Now on OpenClaw v3.24.26 as systemd user service.

Running openclaw cron list or any CLI command from inside the container while live. Now a hard rule: never run OpenClaw CLI commands from inside the container.

Signal router had stale env path (/home/node/ → /home/Rook/). Mac mini node pairing restored via SSH tunnel over Tailscale. UFW rule added for loopback port 18789.

/opt/rook-shared.env rook-bot.service scalpbot.service

All env vars consolidated into /opt/rook-shared.env. Services use EnvironmentFile= instead of scattered inline Environment= lines. DevBot and DevBot2 deleted — were spamming MySQL with unclosed paper trades every 5 min.

MMT_API_KEY was missing from rook-bot service — rook-bot had no OI/CVD data. DevBot log pollution made trade analysis impossible.

formula-v1.json bot.js

(1) 5m/10m/15m weights zeroed — MainBot ignores sub-60m, routes to ScalpBot
(2) Short threshold raised to 55 (45 + 10 bonus) — shorts need higher confluence
(3) Short wave minimum: NAMI wave must be >30 for short entries
(4) Performance stats reset — fresh start post-audit
(5) Auto-execute threshold: 65→55→45 (lowered after backtest showed 47.9% WR at 45+)

Backtest showed 4H was the only profitable timeframe (75% WR). 5m had 27% WR and was bleeding. Shorts underperformed without higher bar.

mmt.js cvd.js oi.js scalpbot.js

Replaced AltFINS OHLCV + OKX WebSocket OI with MMT Binance Futures API.
10m timeframe now natively supported (AltFINS returned 400 on MINUTES10). 30m confirmed working.
CVD now polls MMT every 15s instead of Binance WebSocket aggTrade stream.
ScalpBot fetchHTFBias() + fetchMarketStructure() now use MMT candles.
10m candle maturity delay: 5min → 2min (actual 10m data, not 15m fallback).

Funding Rate Scoring (+12pts max, tiered). Liquidation Bias Scoring (+8pts max). Both from MMT.

HTF invalidation exit only fires if gross P&L > round-trip fees (~$22 on $35k notional at 0.06% taker × 2 sides)

Trade PAPER-1773548759033: closed at +$9.30 gross / −$12.18 net. HTF rule fired on a $37 move — fees ate the profit.

Removed weekend time multiplier. Now: 90min stall alert (notify only), 3hr hard close. Same all days.

30min weekend timer was too aggressive. Correct rule: stalling + HTF invalidation + in profit = exit. Not an arbitrary timer.

4 new post-entry exit reasons: conviction_decay (3hr no progress), htf_invalidation (1H candle against position), profit_protection_exit (was +$50, retraced to entry), stall alert at 90min

PAPER-1773509368379 SHORT: held 5.5 hours, peaked at +$71, bot did nothing — closed −$75. No concept of thesis expiry.

passCount < 2 → passCount < 1 — was requiring both CVD + OI, now requires at least 1 of 2

CVD was blocking all ScalpBot trades. OI alone sufficient for 5m scalper. CVD lags at reversal extremes.

db-logger.js rook-dashboard/server.js

All trades write to Cmelle MySQL Rook-Database.trade_logs in real time. Dashboard price source changed OKX → Kraken XBTUSD to match bot's internal P&L calculation.