CRITICAL: Cooldown was global across ALL symbols, causing missed opportunities
Example: ETH trade at 10:00 blocked SOL trade at 10:04 (5min cooldown)
Changes:
- Added getLastTradeTimeForSymbol() function to query last trade per symbol
- Updated check-risk endpoint to use symbol-specific cooldown
- Each coin (SOL/ETH/BTC) now has independent cooldown timer
- Cooldown message shows symbol: 'Must wait X min before next SOL-PERP trade'
Result: Can trade ETH and SOL simultaneously without interference
Example: ETH LONG at 10:00, SOL SHORT at 10:01 = both allowed
CRITICAL BUG: Position Manager was using current price to determine exit reason,
but on-chain orders filled at a DIFFERENT price in the past!
Example: LONG entry $184.55, TP1 filled at $184.66, but when Position Manager
checked later (price dropped), it saw currentPrice < TP1 and defaulted to 'SL'
Result: Profitable trades incorrectly labeled as SL exits in database
Fix:
- Use trade.tp1Hit and trade.tp2Hit flags to determine exit reason
- If no TP flags set, use realized P&L to distinguish:
- Profit >0.5% = TP1 filled
- Negative P&L = SL filled
- Remove duplicate P&L calculation
This ensures exit reasons match actual on-chain order fills
- Added qualityScore field to Execute Trade node JSON body
- Pulls value from Check Risk response: .item.json.qualityScore
- This enables quality score to be saved in database and displayed on analytics dashboard
Added Maximum Favorable/Adverse Excursion tracking:
- Track maxFavorableExcursion: best profit % reached during trade
- Track maxAdverseExcursion: worst loss % reached during trade
- Track maxFavorablePrice and maxAdversePrice
- Update every price check (2s interval)
- Save to database on trade exit for optimization analysis
Benefits:
- Identify if TP levels are too conservative (MFE consistently higher)
- Determine if SL is too tight (MAE < SL but trade recovers)
- Optimize runner size based on how often MFE >> TP2
- Data-driven exit strategy tuning after collecting 10-20 trades
Display in monitoring logs: Shows MFE/MAE % every 20 seconds
CRITICAL: After TP1 closes 75%, the on-chain stop loss orders were NOT being updated
- Position Manager was tracking new SL price internally but not updating Drift orders
- Old SL orders (e.g., $181.69) remained active even after TP1 at $185.28
- This prevented the 'move SL to breakeven after TP1' logic from working
Fix:
- After TP1 hits, cancel ALL old orders on-chain
- Place new SL orders at updated price (breakeven + configured %)
- Place remaining TP2 order for the 25% runner position
- Maintains dual-stop system if enabled
Result: SL will now actually move up on Drift UI after TP1 fires
- Convert closePosition.closedSize (base asset) to USD when updating trade.currentSize
- Fix conversion when position.size detected from Drift: set currentSize = position.size * currentPrice
- Prevent trade.currentSize from being reduced to tiny values due to unit mismatch
- Execute Trade node was trying to access qualityScore from Check Risk node
- This caused syntax error in n8n when Check Risk blocks the trade
- Backend API calculates qualityScore from the provided metrics (atr, adx, rsi, etc.)
- No need to pass it explicitly in the request body
- Lower ATR threshold from 0.6% to 0.15% (allows low volatility breakouts)
- Increase volume bonus: +15 for very strong volume (1.5x+), was +10 for 1.2x+
- Add volume breakout logic: High volume (1.4x+) at 95%+ range gets +5 instead of -15 penalty
- Add volume compensation: +10 bonus when volume >1.8x and ATR <0.6%
- Example: SOL signal with 0.18% ATR, 1.74x volume at 95.6% range now scores 70/100 (PASS) instead of 25/100 (BLOCK)
- This signal moved +0.97% and would have hit TP1 (+1.5%) - proves quality scoring was too conservative
- Changes apply globally to all symbols (SOL, ETH, BTC) using same scoring algorithm
- Position Manager now detects TP1/TP2 fills by monitoring position size reductions instead of entry price mismatches
- When position size reduces by ~75%, marks TP1 as filled and updates currentSize
- When position size reduces by ~95%, marks TP2 as filled and activates trailing stop for 5% runner
- Entry price mismatch check now skipped after TP fills (Drift shows weighted average entry price after partial closes)
- Fixes bug where runners were incorrectly closed after TP1/TP2 fired on-chain
- Adds grace period for new trades (<30s) to avoid false positives during blockchain propagation delays
- This unblocks Phase 1 data collection for signal quality optimization (need 10+ trades with MAE/MFE data)
- Add qualityScore to ExecuteTradeResponse interface and response object
- Update analytics page to always show Signal Quality card (N/A if unavailable)
- Fix n8n workflow to pass context metrics and qualityScore to execute endpoint
- Fix timezone in Telegram notifications (Europe/Berlin)
- Fix symbol normalization in /api/trading/close endpoint
- Update Drift ETH-PERP minimum order size (0.002 ETH not 0.01)
- Add transaction confirmation to closePosition() to prevent phantom closes
- Add 30-second grace period for new trades in Position Manager
- Fix execution order: database save before Position Manager.addTrade()
- Update copilot instructions with transaction confirmation pattern
- Changed regex from /\.P\s+(\d+)/ to /\b(buy|sell)\s+(\d+|D|W|M)\b/i
- Matches new format: 'ETH buy 15' instead of 'SOL buy .P 15'
- Supports all timeframes: 5, 15, 60, D (daily), W (weekly), M (monthly)
- Updated comment to reflect new format example
- Changed 'SOL buy .P 15' to 'SOL buy 15' (cleaner format)
- timeframe.period is already dynamic (no conversion needed)
- Works for any timeframe: 5, 15, 60, 240, D, etc.
- Format: 'ETH buy 15' or 'BTC sell 5' or 'SOL buy 60'
- Use syminfo.ticker to dynamically get symbol name
- Strip USD/USDT/PERP suffixes to get base currency
- Works for ETH, SOL, BTC, and any other symbol
- Alerts now correctly show 'ETH buy' for Ethereum, 'BTC buy' for Bitcoin, etc.
This fixes the bug where ETH triggers sent 'SOL buy' alerts.
- Updated to Set node v3.4 with proper assignments format
- Explicitly reference Parse Signal Enhanced for rawMessage
- Use $json for Check Risk output (reason, details, score, reasons)
- Properly formatted message with all data fields populated
- Added seconds to timestamp for better tracking
- Removed ugly escaped syntax with $('Parse Signal').item.json references
- Use $json directly (cleaner and works correctly)
- Issues now display as bullet points instead of comma-separated
- Proper line breaks and formatting
- Professional looking blocked trade notifications
- New button in analytics page to clear orphaned trades
- API endpoint /api/trading/clear-manual-closes
- Intelligently checks Drift positions before deleting
- Only removes trades with no matching position or mismatched entry price
- Safe operation: keeps trades on error (false positives better than deletions)
- User-friendly confirmation dialog
- Added getConnection() method to DriftService
- Added proper transaction confirmation in openPosition()
- Check confirmation.value.err to detect on-chain failures
- Return error if transaction fails instead of assuming success
- Prevents phantom trades that never actually execute
This fixes the issue where bot was recording trades with transaction
signatures that don't exist on-chain (like 2gqrPxnvGzdRp56...).
- Added minQualityScore to TradingConfig (default: 60)
- Updated settings UI with slider control (0-100, step 5)
- Updated check-risk endpoint to use config value
- Made scoreSignalQuality function accept minScore parameter
- Updated API to read/write MIN_QUALITY_SCORE env variable
- Allows users to adjust quality threshold from settings page
- Extended MarketConfig with optional positionSize and leverage fields
- Configured ETH-PERP at @ 1x leverage for minimal-risk data collection
- Created getPositionSizeForSymbol() helper function in config/trading.ts
- Integrated symbol-specific sizing into execute endpoint
- Added comprehensive guide in docs/guides/SYMBOL_SPECIFIC_SIZING.md
Purpose: Enable ETH trading for faster signal quality data collection
while preserving SOL's profit-generation sizing (0 @ 10x)
Next: Create ETH alert in TradingView and restart bot
Enhanced 'Format Risk' node in n8n workflow to display:
- Specific blocking reason (duplicate, drawdown, cooldown, quality, etc.)
- Details about what triggered the block
- Quality score if low quality was the reason
- Quality issues breakdown (ATR too low, weak ADX, etc.)
Example output:
TRADE BLOCKED
SHORT | ATR:0.30 | ADX:19.1 | RSI:46
Issues: ATR too low (0.30% - dead market), Moderate trend (ADX 19.1), RSI supports short (46.0)
14:23
Comprehensive guide covering:
- How ATR is captured and stored (entry value frozen)
- Static ATR approach (Phases 1-3): Use entry ATR for entire trade
- Dynamic ATR approach (Phase 5+): Real-time updates via TradingView or bot calculation
- Use cases: Dynamic TP/SL, trailing stops, scaling in/out decisions
- Implementation path: Start simple with entry ATR, add real-time later if data supports
- Code examples for all approaches
- Troubleshooting common ATR issues
- Database schema considerations
Explains why waiting for data is critical before implementing advanced ATR features.
Corrections:
- Runner system already exists (5% with 0.3% trailing stop)
- Current +41% trade is the runner in action!
- Phase 5 reframed: Optimize runner size (5% → 10-25% for high quality) and make trailing stop ATR-based
- Updated current state and trade example to reflect actual implementation
- 6-phase development plan: data collection → ATR-based → quality tiers → direction bias → runners → ML
- Each phase has clear prerequisites, implementation tasks, and success criteria
- Decision gates based on data validation (20+ trades for Phase 2, 30+ for Phase 3, etc.)
- Includes SQL queries for analysis and performance validation
- Documents current +41% trade as motivation for runner implementation
- Estimated 3-4 months timeline to complete Phases 1-5
- Add signalQualityScore field to Trade model (0-100)
- Calculate quality score in execute endpoint using same logic as check-risk
- Save score with every trade for correlation analysis
- Create database migration for new field
- Enables future analysis: score vs win rate, P&L, etc.
This allows data-driven decisions on dynamic position sizing
- Add getLastTrade() function to database service
- Create /api/analytics/last-trade endpoint
- Display last trade with full details on analytics page
- Show entry/exit prices, P&L, position size, targets
- Visual indicators for trade direction and exit reason
- Helps quickly diagnose where trades went (TP1, TP2, or SL)
Bug: Position Manager was comparing ANY position on the symbol to the trade being
tracked, without verifying entry price match. When a new position opened, it would
think the old tracked trade 'closed externally' and cancel ALL orders - including
the new position's exit orders.
Fix: Added entry price verification (0.5% tolerance). If position entry price doesn't
match the tracked trade, mark the old trade as 'lost tracking' and remove from
monitoring WITHOUT cancelling orders (they belong to the new position).
This prevents the catastrophic scenario where exit orders are repeatedly cancelled,
leaving positions unprotected.
Added 5 context metrics to alert messages:
- ATR% (volatility as % of price)
- ADX (trend strength)
- RSI (momentum)
- VOL (volume ratio vs 20-bar MA)
- POS (price position in 20-bar range 0-100%)
Changes to Pine Script:
- Always calculate ADX (needed for context even if filter disabled)
- Extract ta.rma() calls outside ternary operators (Pine Script requirement)
- Use alert() instead of alertcondition() for dynamic message support
- Changed to single-line string concatenation for compatibility
Alert message format:
OLD: 'Buy SOL 15 | Profile=Hours ATR=10 Mult=3.0'
NEW: 'SOL buy .P 15 | ATR:1.85 | ADX:28.3 | RSI:62.5 | VOL:1.45 | POS:75.3'
Next: Update n8n to parse these metrics, implement signal quality scoring in bot
- Updated minTimeBetweenTrades config to use minutes instead of seconds
- Changed default from 600 seconds to 10 minutes
- Updated Settings UI label from 'seconds' to 'minutes' and adjusted range (0-60 min)
- Updated .env comments to reflect new unit
- No functional change since cooldown enforcement not yet implemented (TODO in check-risk route)
Bug: MAE/MFE was tracked in memory during trades but not saved to database on exit
Cause: updateTradeExit() wasn't receiving or saving MAE/MFE parameters
Changes:
- Added MAE/MFE fields to UpdateTradeExitParams interface
- Modified updateTradeExit() to save maxFavorableExcursion, maxAdverseExcursion, maxFavorablePrice, maxAdversePrice
- Updated both updateTradeExit() calls in Position Manager to pass MAE/MFE values
- Enhanced exit logging to show final MAE/MFE percentages
Impact: Future trades will now properly save MAE/MFE data for analytics
Note: Past 2 trades (from before this fix) don't have MAE/MFE saved
- Created /analytics/optimization page with comprehensive UI
- Displays MAE/MFE analysis with percentiles
- Shows current TP/SL performance with hit rate bars
- Visualizes optimal recommendations vs current levels
- Projects impact of optimization (win rate, profit factor, P&L improvement)
- Provides reasoning for each recommended level
- Added navigation link from main analytics page
Dashboard features:
- Overview stats: total trades, win rate, profit factor, money left on table
- MAE analysis: avg, median, 25th/75th percentile, worst
- MFE analysis: avg, median, 25th/75th percentile, best
- Current config: TP1/TP2/SL hit rates with progress bars
- Recommendations: optimal levels with color-coded cards
- Reasoning cards: explanation for each recommendation
- Projected impact: win rate change, profit factor change, profit improvement
- Direct link to Settings page to apply recommendations
Access at: http://localhost:3001/analytics/optimization
Phase 1-4 Complete! System now tracks MAE/MFE, captures market context,
analyzes performance, and provides data-driven TP/SL recommendations.
- Created /api/analytics/tp-sl-optimization endpoint
- Analyzes historical trades using MAE/MFE data
- Calculates optimal TP1/TP2/SL levels based on percentiles
- Provides win rate, profit factor, and hit rate analysis
- Shows money left on table (MFE - realized P&L)
- Projects impact of optimal levels on future performance
Analytics calculated:
- MAE analysis: avg, median, percentiles, worst
- MFE analysis: avg, median, percentiles, best
- Current level performance: TP1/TP2/SL hit rates
- Optimal recommendations: TP1=50% of avg MFE, TP2=80%, SL=70% of avg MAE
- Projected improvements: win rate change, profit factor, total P&L
Requires 10+ closed trades with MAE/MFE data to generate recommendations
Test script: scripts/test-analytics.sh
Next: Phase 4 (visual dashboard) or wait for trades with MAE/MFE data
- Added getFundingRate() method to DriftService
- Capture expectedEntryPrice from oracle before order execution
- Capture fundingRateAtEntry from Drift Protocol
- Save market context fields to database (expectedEntryPrice, fundingRateAtEntry)
- Calculate entry slippage percentage in createTrade()
- Fixed template literal syntax errors in execute endpoint
Database fields populated:
- expectedEntryPrice: Oracle price before order
- entrySlippagePct: Calculated from entrySlippage
- fundingRateAtEntry: Current funding rate from Drift
Next: Phase 3 (analytics API) or test market context on next trade
- Detect on-chain TP2 fills in size mismatch logic and set tp2Hit flag
- Position size thresholds: <30% = TP1, <10% = TP2 (prevents runner from being closed)
- Ensures runner (5-20%) trails properly instead of being market-closed immediately
**ROOT CAUSE:** placeExitOrders() calculated position size using TP/SL prices instead of entry price
**Problem:**
- TP1 order size: 85 / TP1_price (00.746) = 2.914 SOL
- Actual position: 80 / entry_price (99.946) = 3.901 SOL
- TP1 should close: 3.901 * 75% = 2.926 SOL
- But it only closed: 2.914 SOL = 74.7% ❌ WRONG!
**Result:** TP1 closed ~25% instead of 75%, no runner left
**Fix:**
- Changed usdToBase() to use entryPrice for ALL size calculations
- Added entryPrice param to PlaceExitOrdersOptions interface
- Updated all API routes to pass entryPrice
**Testing:** Next trade will have correctly sized TP/SL orders
- Remove saveTradeState() call from addTrade() to avoid P2025 error
- Add initialization check in checkTradeConditions() to skip when Drift not ready
- Silence 'not initialized' errors during startup (expected behavior)
- Trade state is now saved only by API endpoint after DB record created
- Change takeProfit2SizePercent from 100% to 80% to leave 5% runner
- Fix cancelAllOrders() to detect trigger orders using orderId > 0
- Trigger orders (TRIGGER_MARKET, TRIGGER_LIMIT) now properly canceled
- Trailing stop will now activate on 5% runner position
