trading_bot_v4/docs/analysis/BLOCKED_SIGNALS_TRACKING.md

# Blocked Signals Tracking System

**Date Implemented:** November 11, 2025  
**Status:** ✅ ACTIVE

## Overview

Automatically tracks all signals that get blocked by the trading bot's risk checks. This data allows us to analyze whether blocked signals would have been profitable, helping optimize the signal quality thresholds over time.

## What Gets Tracked

Every time a signal is blocked, the system saves:

### Signal Metrics
- Symbol (e.g., SOL-PERP)
- Direction (long/short)
- Timeframe (5min, 15min, 1H, etc.)
- Price at signal time
- ATR, ADX, RSI, volume ratio, price position

### Quality Score
- Calculated score (0-100)
- Score version (v4 = current)
- Detailed breakdown of scoring reasons
- Minimum score required (currently 91, raised Nov 21, 2025)
- **Why 91?** Perfect separation in 7 v8 trades:
  - ALL winners: quality ≥95 (95, 95, 100, 105)
  - ALL losers: quality ≤90 (80, 90, 90)
  - Would have prevented 100% of losses (-$624.90 total)

### Block Reason
- `QUALITY_SCORE_TOO_LOW` - Score below threshold
- `COOLDOWN_PERIOD` - Too soon after last trade
- `HOURLY_TRADE_LIMIT` - Too many trades in last hour
- `DAILY_DRAWDOWN_LIMIT` - Max daily loss reached

### Future Analysis Fields (NOT YET IMPLEMENTED)
- `priceAfter1Min`, `priceAfter5Min`, `priceAfter15Min`, `priceAfter30Min`
- `wouldHitTP1`, `wouldHitTP2`, `wouldHitSL`
- `analysisComplete`

These will be filled by a monitoring job that tracks what happened after each blocked signal.

## Database Table

```sql
Table: BlockedSignal
- id (PK)
- createdAt (timestamp)
- symbol, direction, timeframe
- signalPrice, atr, adx, rsi, volumeRatio, pricePosition
- signalQualityScore, signalQualityVersion, scoreBreakdown (JSON)
- minScoreRequired, blockReason, blockDetails
- priceAfter1Min/5Min/15Min/30Min (for future analysis)
- wouldHitTP1/TP2/SL, analysisComplete
```

## Query Examples

### Recent Blocked Signals
```sql
SELECT 
  symbol, 
  direction,
  signalQualityScore as score,
  minScoreRequired as threshold,
  blockReason,
  createdAt
FROM "BlockedSignal"
ORDER BY createdAt DESC
LIMIT 20;
```

### Blocked by Quality Score (60-64 range)
```sql
SELECT 
  symbol,
  direction,
  signalQualityScore,
  ROUND(atr::numeric, 2) as atr,
  ROUND(adx::numeric, 1) as adx,
  ROUND(rsi::numeric, 1) as rsi,
  ROUND(pricePosition::numeric, 1) as pos,
  blockDetails
FROM "BlockedSignal"
WHERE blockReason = 'QUALITY_SCORE_TOO_LOW'
  AND signalQualityScore >= 60
  AND signalQualityScore < 65
ORDER BY createdAt DESC;
```

### Breakdown by Block Reason
```sql
SELECT 
  blockReason,
  COUNT(*) as count,
  ROUND(AVG(signalQualityScore)::numeric, 1) as avg_score,
  MIN(signalQualityScore) as min_score,
  MAX(signalQualityScore) as max_score
FROM "BlockedSignal"
GROUP BY blockReason
ORDER BY count DESC;
```

### Today's Blocked Signals
```sql
SELECT 
  TO_CHAR(createdAt, 'HH24:MI:SS') as time,
  symbol,
  direction,
  signalQualityScore,
  blockReason
FROM "BlockedSignal"
WHERE createdAt >= CURRENT_DATE
ORDER BY createdAt DESC;
```

## Analysis Workflow

### Step 1: Collect Data (Current Phase)
- Bot automatically saves blocked signals
- Wait for 10-20 blocked signals to accumulate
- No action needed - runs automatically

### Step 2: Manual Analysis (When Ready)
```sql
-- Check how many blocked signals we have
SELECT COUNT(*) FROM "BlockedSignal";

-- Analyze score distribution
SELECT 
  CASE 
    WHEN signalQualityScore >= 60 THEN '60-64 (Close Call)'
    WHEN signalQualityScore >= 55 THEN '55-59 (Marginal)'
    WHEN signalQualityScore >= 50 THEN '50-54 (Weak)'
    ELSE '0-49 (Very Weak)'
  END as score_tier,
  COUNT(*) as count,
  ROUND(AVG(atr)::numeric, 2) as avg_atr,
  ROUND(AVG(adx)::numeric, 1) as avg_adx
FROM "BlockedSignal"
WHERE blockReason = 'QUALITY_SCORE_TOO_LOW'
GROUP BY score_tier
ORDER BY MIN(signalQualityScore) DESC;
```

### Step 3: Future Automation (Not Yet Built)
Create a monitoring job that:
1. Fetches `BlockedSignal` records where `analysisComplete = false` and `createdAt` > 30min ago
2. Gets price history for those timestamps
3. Calculates if TP1/TP2/SL would have been hit
4. Updates the record with analysis results
5. Sets `analysisComplete = true`

## Integration Points

### Code Files Modified
1. `prisma/schema.prisma` - Added `BlockedSignal` model
2. `lib/database/trades.ts` - Added `createBlockedSignal()` function
3. `app/api/trading/check-risk/route.ts` - Saves blocked signals

### Where Blocking Happens
- Quality score check (line ~311-350)
- Cooldown period check (line ~281-303)
- Hourly trade limit (line ~235-258)
- Daily drawdown limit (line ~211-223)

## Next Steps

### Phase 1: Data Collection (CURRENT)
- ✅ Database table created
- ✅ Automatic saving implemented
- ✅ Bot deployed and running
- ⏳ Collect 10-20 blocked signals (wait ~1-2 weeks)

### Phase 2: Analysis
- Query blocked signal patterns
- Identify "close calls" (score 60-64)
- Compare with executed trades that had similar scores
- Determine if threshold adjustment is warranted

### Phase 3: Automation (Future)
- Build price monitoring job
- Auto-calculate would-be outcomes
- Generate reports on missed opportunities
- Feed data into threshold optimization algorithm

## Benefits

1. **Data-Driven Decisions** - No guessing, only facts
2. **Prevents Over-Optimization** - Wait for statistically significant sample
3. **Tracks All Block Reasons** - Not just quality score
4. **Historical Record** - Can review past decisions
5. **Continuous Improvement** - System learns from what it blocks

## Important Notes

⚠️ **Don't change thresholds prematurely!** 
- 2 trades is NOT enough data
- Wait for 10-20 blocked signals minimum
- Analyze patterns before making changes

✅ **System is working correctly if:**
- Blocked signals appear in database
- Each has metrics (ATR, ADX, RSI, etc.)
- Block reason is recorded
- Timestamp is correct

❌ **Troubleshooting:**
- If no blocked signals appear: Check bot is receiving TradingView alerts with metrics
- If missing metrics: Ensure TradingView webhook includes ATR/ADX/RSI/volume/pricePosition
- If database errors: Check Prisma client is regenerated after schema changes

---

**Last Updated:** November 11, 2025  
**Version:** 1.0  
**Maintained By:** Trading Bot v4 Development Team