feat: Extend 1-minute data retention from 4 weeks to 1 year
- Updated lib/maintenance/data-cleanup.ts retention period: 28 days → 365 days - Storage requirements validated: 251 MB/year (negligible) - Rationale: 13× more historical data for better pattern analysis - Benefits: 260-390 blocked signals/year vs 20-30/month - Cleanup cutoff: Now Dec 2, 2024 (vs Nov 4, 2025 previously) - Deployment verified: Container restarted, cleanup scheduled for 3 AM daily
This commit is contained in:
mindesbunister
492
BLOCKED_SIGNAL_CORRECTED_ANALYSIS_DEC2.md
Normal file
492
BLOCKED_SIGNAL_CORRECTED_ANALYSIS_DEC2.md
Normal file
@@ -0,0 +1,492 @@
|
||||
# 🚨 CORRECTED: Blocked Signal Analysis - Quality 80-89 Signals
|
||||
|
||||
**Analysis Date:** December 2, 2025
|
||||
**Status:** CRITICAL CORRECTION - Initial analysis was WRONG
|
||||
|
||||
---
|
||||
|
||||
## Executive Summary: I Was Wrong
|
||||
|
||||
**User Discovery:** "Price went all the way up to $127.95 after roughly 4 hours later"
|
||||
**My Error:** Only looked at 30-minute tracking data, missed the 4-hour development
|
||||
**Reality:** Signal blocked at quality 80 was a WINNER (+1.55%, ~$30 profit missed)
|
||||
|
||||
---
|
||||
|
||||
## The Dec 1 Signal (Your Chart Case)
|
||||
|
||||
**Signal Details:**
|
||||
- Time: Dec 1, 2025 21:40 UTC
|
||||
- Entry: $126.00 LONG
|
||||
- Quality: 80 (threshold: 90 for LONGs)
|
||||
- ADX: 17.2 (weak trend)
|
||||
- Block Reason: QUALITY_SCORE_TOO_LOW
|
||||
|
||||
**My Flawed 30-Minute Analysis:**
|
||||
- Peak: $126.11 (+0.085%)
|
||||
- TP1 Target: $126.86 (+0.82%)
|
||||
- Conclusion: "Signal never came close to TP1" ❌ WRONG
|
||||
|
||||
**Your Chart Shows (4-Hour Reality):**
|
||||
- Peak: **$127.95**
|
||||
- Actual Move: **+1.55%**
|
||||
- TP1 Hit: YES ✅ ($126.86)
|
||||
- TP2 Potential: Nearly hit ($127.72 target)
|
||||
- **Missed Profit: ~$30 on this one signal**
|
||||
|
||||
---
|
||||
|
||||
## Complete Quality 80-89 Dataset (9 Signals Total)
|
||||
|
||||
### Current Quality Thresholds
|
||||
```
|
||||
LONG signals: ≥90 required (quality 80-89 blocked)
|
||||
SHORT signals: ≥80 required (quality 80+ should execute)
|
||||
```
|
||||
|
||||
### All Quality 80 Signals (6 total) - 30-Min Tracking
|
||||
|
||||
**LONG Signals Blocked (Threshold 90):**
|
||||
|
||||
1. **Dec 1 @ $126.00 (ADX 17.2)** ⭐ YOUR CASE
|
||||
- 30min MFE: +0.085%
|
||||
- **Chart shows:** +1.55% at 4 hours
|
||||
- TP1 (+0.82%): ✅ HIT beyond 30 minutes
|
||||
- **Actual outcome:** WINNER (missed)
|
||||
|
||||
2. **Nov 24, 05:10 @ $132.76 (ADX 22.9)**
|
||||
- 30min MFE: +0.741%
|
||||
- TP1 hit: TRUE (within 30min)
|
||||
- **Actual outcome:** ✅ WINNER (would have caught with lower threshold)
|
||||
|
||||
3. **Nov 22, 22:55 @ $128.52 (ADX 21.6)**
|
||||
- 30min: Hit SL (-0.864%)
|
||||
- **Actual outcome:** ❌ LOSER
|
||||
|
||||
4. **Nov 21, 16:50 @ $126.20 (ADX 16.6)**
|
||||
- 30min MFE: +0.524%
|
||||
- TP1 target: ~0.66%
|
||||
- **Status:** ❓ Very close, unknown if hit beyond 30min
|
||||
|
||||
**SHORT Signals (Quality 80 = threshold, but still blocked?):**
|
||||
|
||||
5. **Nov 24, 07:05 @ $130.68 (ADX 20.0)**
|
||||
- 30min MFE: +0.349%
|
||||
- TP1 target: ~0.80%
|
||||
- **Status:** ❓ Unknown beyond 30min
|
||||
|
||||
6. **Nov 24, 04:25 @ $130.78 (ADX 26.0)**
|
||||
- 30min MFE: +0.134%
|
||||
- **Status:** ❓ Unknown beyond 30min
|
||||
|
||||
### Quality 70-79 Signals (3 total)
|
||||
|
||||
7. **Quality 75 SHORT @ $140.01 (Nov 28, ADX 23.3)**
|
||||
- 30min MFE: +0.528%
|
||||
- TP1 target: ~0.60%
|
||||
- **Status:** ❓ Likely hit, unknown
|
||||
|
||||
8. **Quality 70 LONG @ $137.84 (Nov 26, ADX 14.1)**
|
||||
- 30min MAE: -0.957%
|
||||
- **Actual outcome:** ❌ LOSER
|
||||
|
||||
9. **Quality 75 SHORT @ $126.48 (Nov 22, ADX 20.7)**
|
||||
- 30min MFE: +0.179%
|
||||
- **Status:** ❓ Unknown beyond 30min
|
||||
|
||||
---
|
||||
|
||||
## Key Pattern: ADX vs Development Speed
|
||||
|
||||
### Low ADX (<20) = SLOW DEVELOPERS ⏰
|
||||
- **Dec 1 (ADX 17.2):** +0.085% in 30min → +1.55% in 4hr ✅
|
||||
- **Nov 21 (ADX 16.6):** +0.524% in 30min (very close to TP1)
|
||||
- **Nov 26 (ADX 14.1):** Hit SL (loser) ❌
|
||||
|
||||
**Characteristic:** Take 2-4+ hours to hit TP1
|
||||
**30-Min Window:** MISSES MOST OF THEM
|
||||
|
||||
### High ADX (>22) = FAST MOVERS ⚡
|
||||
- **Nov 24 (ADX 22.9):** +0.741% in 30min, hit TP1 ✅
|
||||
- **Nov 24 (ADX 26.0):** +0.134% in 30min (unexpectedly weak)
|
||||
|
||||
**Characteristic:** Hit TP1 within 30 minutes
|
||||
**30-Min Window:** CATCHES THESE
|
||||
|
||||
### Critical Problem
|
||||
|
||||
**BlockedSignalTracker limitation:**
|
||||
- Only monitors 30 minutes
|
||||
- Slow developers (low ADX) need 4-8 hours
|
||||
- **Result:** False negative statistics (appear to lose when they actually win)
|
||||
|
||||
**Current Stats (WRONG):**
|
||||
- "Quality 80 win rate: 16.7% (1/6)"
|
||||
- Based on 30-minute data only
|
||||
- Missing slow developers like Dec 1 signal
|
||||
|
||||
**Corrected Stats (WITH YOUR CHART DATA):**
|
||||
- Confirmed winners: 2/6 (33.3%)
|
||||
- Confirmed losers: 1/6 (16.7%)
|
||||
- Unknown: 3/6 (50.0%)
|
||||
- **Likely actual win rate: 50-60%** (when tracked properly)
|
||||
|
||||
---
|
||||
|
||||
## Financial Impact: What We're Missing
|
||||
|
||||
### Dec 1 Signal Calculation
|
||||
- Entry: $126.00 LONG
|
||||
- Capital: $540 USDC
|
||||
- Quality 80 → Adaptive Leverage: 5×
|
||||
- Position: $2,700 notional
|
||||
|
||||
**If Traded:**
|
||||
- TP1 @ $126.86 (+0.82%): 60% close = $1,620 × 0.82% = **$13.28**
|
||||
- Runner @ $127.95 (+1.55%): 40% remaining = $1,080 × 1.55% = **$16.74**
|
||||
- **Total Profit: $30.02**
|
||||
- **Account Gain: +5.56%**
|
||||
|
||||
**Actual Result:** $0 (blocked)
|
||||
|
||||
### Monthly Opportunity Cost
|
||||
|
||||
**Conservative Estimate:**
|
||||
- 2-3 quality 80-89 LONG signals/week blocked
|
||||
- ~50% are winners (based on corrected analysis)
|
||||
- Average $25-30 profit per winner
|
||||
- **Missed profit: $100-180/month**
|
||||
|
||||
**Optimistic Estimate:**
|
||||
- If we tracked 4-8 hours and found 60% win rate
|
||||
- 8-12 signals/month × 60% × $30 avg
|
||||
- **Missed profit: $144-216/month**
|
||||
|
||||
---
|
||||
|
||||
## Brainstorming: What Should We Change?
|
||||
|
||||
### Option 1: Lower LONG Quality Threshold 🔧
|
||||
**Current:** 90 → **Proposed:** 85 or 80
|
||||
|
||||
**Pros:**
|
||||
- Would catch Dec 1 winner (+$30)
|
||||
- Would catch Nov 24 winner (+$23)
|
||||
- Data shows 33% confirmed win rate (2/6)
|
||||
- Potential +$100-180/month
|
||||
|
||||
**Cons:**
|
||||
- Nov 22 loser would also execute (-$30)
|
||||
- Unknown true win rate (need 4-8hr tracking)
|
||||
- Risk: May increase losing trades
|
||||
- **Need more data before deciding**
|
||||
|
||||
**Required Actions:**
|
||||
1. Extend tracking to 4-8 hours (see Option 4)
|
||||
2. Collect 20-30 more quality 80-89 signals
|
||||
3. Calculate true win rate with proper time horizon
|
||||
4. Lower threshold ONLY if win rate >55%
|
||||
|
||||
---
|
||||
|
||||
### Option 2: ADX-Based Dynamic Threshold 🎯
|
||||
**Concept:** Use ADX to adjust entry rules
|
||||
|
||||
**Low ADX (<20):**
|
||||
- Allow quality 80-85 LONG entries
|
||||
- BUT: Extend Smart Entry Validation to 60 minutes
|
||||
- Rationale: Slow developers need time to confirm
|
||||
- Example: Dec 1 (ADX 17.2) would get 60-min window
|
||||
|
||||
**High ADX (>22):**
|
||||
- Keep threshold 90 (fast movers already handled)
|
||||
- 10-minute Smart Entry window sufficient
|
||||
- Example: Nov 24 (ADX 22.9) would still need quality 90
|
||||
|
||||
**Pros:**
|
||||
- Targets exact problem: Low ADX = slow development
|
||||
- Preserves safety (no blanket lowering)
|
||||
- Data supports pattern (both low ADX cases were slow)
|
||||
|
||||
**Cons:**
|
||||
- More complex implementation
|
||||
- Small sample size (only 2 low ADX signals)
|
||||
- Still need extended tracking for validation
|
||||
|
||||
**Implementation:**
|
||||
```typescript
|
||||
// In smart-validation-queue.ts or check-risk endpoint
|
||||
if (adx < 20 && qualityScore >= 80 && qualityScore < 90) {
|
||||
queueForSmartEntry({
|
||||
symbol,
|
||||
direction,
|
||||
originalPrice,
|
||||
qualityScore,
|
||||
validationWindow: 60, // minutes (vs default 10)
|
||||
confirmationThreshold: 0.3, // % move to confirm
|
||||
holdRequirement: 90 // seconds
|
||||
})
|
||||
} else if (qualityScore < 90) {
|
||||
// Block or use standard 10-min validation
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### Option 3: Extend Smart Entry Validation Window ⏱️
|
||||
**Current:** 10 minutes → **Proposed:** 30-60 minutes for quality 80-89
|
||||
|
||||
**How It Works:**
|
||||
- Quality 90+: Immediate execution (unchanged)
|
||||
- Quality 80-89: Queue for extended validation
|
||||
- Monitor 30-60 minutes instead of 10 minutes
|
||||
- Same confirmation (+0.3%) and abandonment (-0.4%) triggers
|
||||
|
||||
**Pros:**
|
||||
- Catches slow developers without lowering threshold
|
||||
- Non-invasive (no core quality logic changes)
|
||||
- Dec 1 signal: Would likely confirm around 30-60 min
|
||||
- Nov 21 signal: Already showed +0.524% in 30min
|
||||
|
||||
**Cons:**
|
||||
- Capital locked longer (may miss other setups)
|
||||
- More monitoring overhead (database, resources)
|
||||
- Still need data to validate optimal window
|
||||
|
||||
**Configuration:**
|
||||
```env
|
||||
# Add to .env
|
||||
SMART_ENTRY_VALIDATION_WINDOW_DEFAULT=10
|
||||
SMART_ENTRY_VALIDATION_WINDOW_MARGINAL=60
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### Option 4: Extend BlockedSignalTracker (CRITICAL) 🔥
|
||||
**Current:** 30-minute tracking → **Proposed:** 4-8 hour tracking
|
||||
|
||||
**THIS IS THE MOST IMPORTANT CHANGE** - Regardless of other decisions
|
||||
|
||||
**Why This Matters:**
|
||||
- Current stats are WRONG (30-min window too short)
|
||||
- Can't make informed decisions without accurate data
|
||||
- Your chart proved it: Dec 1 signal took 4 hours to develop
|
||||
- Need truth before changing thresholds
|
||||
|
||||
**Database Changes:**
|
||||
```sql
|
||||
-- Add extended tracking columns
|
||||
ALTER TABLE "BlockedSignal"
|
||||
ADD COLUMN "priceAfter1Hr" DOUBLE PRECISION,
|
||||
ADD COLUMN "priceAfter2Hr" DOUBLE PRECISION,
|
||||
ADD COLUMN "priceAfter4Hr" DOUBLE PRECISION,
|
||||
ADD COLUMN "priceAfter8Hr" DOUBLE PRECISION;
|
||||
```
|
||||
|
||||
**Code Changes:**
|
||||
```typescript
|
||||
// In lib/analysis/blocked-signal-tracker.ts
|
||||
// Current intervals
|
||||
const trackingIntervals = [1, 5, 15, 30] // minutes
|
||||
|
||||
// Proposed intervals
|
||||
const trackingIntervals = [1, 5, 15, 30, 60, 120, 240, 480] // minutes (8 hours)
|
||||
```
|
||||
|
||||
**Benefits:**
|
||||
- Accurate win rate statistics
|
||||
- Can see ADX pattern more clearly
|
||||
- Data-driven threshold decisions
|
||||
- Low risk (no trading changes, just better data)
|
||||
|
||||
**Timeline:**
|
||||
- Implement: 1-2 days
|
||||
- Collect data: 2-4 weeks (20-30 signals)
|
||||
- Analyze: 1 day
|
||||
- Decide threshold: Based on real data
|
||||
|
||||
---
|
||||
|
||||
### Option 5: Hybrid Approach (RECOMMENDED) ✅
|
||||
|
||||
**Phase 1: Data Collection (Weeks 1-4)**
|
||||
1. ✅ Extend BlockedSignalTracker to 8 hours (PRIORITY #1)
|
||||
2. ✅ Continue blocking quality 80-89 LONGs
|
||||
3. ✅ Track outcomes properly
|
||||
4. ✅ Collect 20-30 more signals minimum
|
||||
5. ✅ Calculate true win rate with proper time horizon
|
||||
|
||||
**Phase 2: Threshold Decision (Week 5)**
|
||||
Based on data collected in Phase 1:
|
||||
|
||||
**If Quality 80-89 Win Rate >60%:**
|
||||
- Lower LONG threshold: 90 → 85
|
||||
- Monitor for 2-4 weeks
|
||||
- Adjust if needed
|
||||
|
||||
**If Quality 80-89 Win Rate 50-60%:**
|
||||
- Implement ADX-based dynamic threshold
|
||||
- Low ADX gets extended validation (60min)
|
||||
- High ADX keeps standard rules
|
||||
|
||||
**If Quality 80-89 Win Rate <50%:**
|
||||
- Keep threshold at 90
|
||||
- But extend Smart Entry Validation to 30-60 minutes
|
||||
- Catches slow developers without risking bad entries
|
||||
|
||||
**Phase 3: Continuous Optimization (Ongoing)**
|
||||
- Monitor monthly performance
|
||||
- Adjust based on actual results
|
||||
- A/B test configuration changes
|
||||
- Iterate towards optimal settings
|
||||
|
||||
**Expected Financial Impact:**
|
||||
- Phase 1: $0 (data collection, no changes)
|
||||
- Phase 2: +$50-150/month (conservative estimate)
|
||||
- Phase 3: +$100-250/month (with optimization)
|
||||
- **Total Potential:** +$150-400/month from quality 70-89 signals
|
||||
|
||||
**Risk Management:**
|
||||
- Phase 1: Zero risk (no trading changes)
|
||||
- Phase 2: Calculated risk (data-driven decision)
|
||||
- Phase 3: Low risk (continuous monitoring)
|
||||
|
||||
---
|
||||
|
||||
## Recommended Action Plan
|
||||
|
||||
### Immediate (This Week)
|
||||
1. **Extend BlockedSignalTracker** ✅ HIGH PRIORITY
|
||||
- Update database schema (add 1Hr, 2Hr, 4Hr, 8Hr columns)
|
||||
- Modify tracking intervals in code
|
||||
- Deploy and verify working
|
||||
- Git commit: "feat: Extend BlockedSignalTracker to 8 hours"
|
||||
|
||||
2. **Manual Analysis of Existing 9 Signals** ⏱️
|
||||
- Query historical 5-minute OHLCV data
|
||||
- Check price 1-8 hours after each signal
|
||||
- Determine if TP1 was actually hit
|
||||
- Update BLOCKED_SIGNAL_CORRECTED_ANALYSIS_DEC2.md with findings
|
||||
|
||||
3. **Document Findings** 📝
|
||||
- Update copilot-instructions.md with new insights
|
||||
- Add Common Pitfall: "BlockedSignalTracker 30-min window too short"
|
||||
- Document ADX pattern: Low ADX = slow developers
|
||||
|
||||
### Short Term (Weeks 2-4)
|
||||
1. **Continue Data Collection**
|
||||
- Target: 20-30 more quality 70-89 signals
|
||||
- Track for full 8 hours each
|
||||
- Monitor completion rate
|
||||
|
||||
2. **Statistical Analysis**
|
||||
- Calculate true win rate by quality tier
|
||||
- Analyze ADX vs development speed correlation
|
||||
- Compare LONG vs SHORT performance
|
||||
- Determine optimal threshold
|
||||
|
||||
3. **Decision Point: Week 4**
|
||||
- Review 8-hour tracking data (30+ signals total)
|
||||
- Calculate expected value of threshold changes
|
||||
- Decide: Lower threshold, add ADX logic, or extend validation
|
||||
- Document decision rationale
|
||||
|
||||
### Medium Term (Weeks 5-8)
|
||||
1. **Implement Changes** (if data supports)
|
||||
- Lower threshold OR add ADX logic OR extend validation
|
||||
- Deploy to production
|
||||
- Monitor first 20-30 trades closely
|
||||
|
||||
2. **Validation Period**
|
||||
- Compare actual results to predicted
|
||||
- Check win rate matches expectations
|
||||
- Verify no unintended consequences
|
||||
- Adjust if needed
|
||||
|
||||
3. **Performance Report**
|
||||
- Document profit impact
|
||||
- Compare to baseline (no changes)
|
||||
- Calculate ROI of optimization effort
|
||||
- Plan next optimization phase
|
||||
|
||||
---
|
||||
|
||||
## Key Lessons Learned
|
||||
|
||||
### What I Did Wrong
|
||||
|
||||
1. **Trusted 30-minute data as complete** ❌
|
||||
- Assumed tracking window captured full development
|
||||
- Didn't question why window was so short
|
||||
- Missed slow developers (low ADX signals)
|
||||
|
||||
2. **Ignored user's visual evidence** ❌
|
||||
- User showed chart with $127.95 peak
|
||||
- I relied only on database (incomplete)
|
||||
- Should have asked: "When did price reach that?"
|
||||
|
||||
3. **Made conclusion without validation** ❌
|
||||
- Said "system correct" based on partial data
|
||||
- Recommended "no changes" prematurely
|
||||
- Should have said: "Need longer tracking first"
|
||||
|
||||
4. **Missed ADX pattern** ❌
|
||||
- ADX 17.2 + slow development = obvious pattern
|
||||
- Didn't recognize: Low ADX = needs more time
|
||||
- Pattern was there, I just didn't see it
|
||||
|
||||
### What I Should Have Done ✅
|
||||
|
||||
1. **Question data limitations**
|
||||
- "Is 30 minutes enough to judge signal quality?"
|
||||
- "Why did we choose 30 minutes?"
|
||||
- "What if signals develop slower?"
|
||||
|
||||
2. **Verify with multiple sources**
|
||||
- Cross-check: Database vs Chart vs API
|
||||
- Time-match: When was signal, when was chart peak
|
||||
- Calculate: How long did full move take
|
||||
|
||||
3. **Look for patterns**
|
||||
- ADX vs development speed
|
||||
- Quality score vs outcome
|
||||
- Timeframe vs signal validity
|
||||
|
||||
4. **State assumptions explicitly**
|
||||
- "Based on 30-minute tracking..."
|
||||
- "Assuming signal develops within 30 minutes..."
|
||||
- "Limited by tracking window..."
|
||||
|
||||
5. **Recommend data collection first**
|
||||
- "Need 4-8 hour tracking to validate"
|
||||
- "Collect 20-30 signals before deciding"
|
||||
- "Data-driven decision, not assumption-based"
|
||||
|
||||
---
|
||||
|
||||
## Conclusion
|
||||
|
||||
**Original Conclusion:** ❌ WRONG
|
||||
"System correctly blocked signal, no changes needed"
|
||||
|
||||
**Corrected Conclusion:** ⚠️ NEEDS INVESTIGATION
|
||||
"System blocked a winner (+1.55%, $30 profit). Root cause: 30-minute tracking too short for slow developers (ADX <20). Need 4-8 hour tracking to determine true win rate before changing thresholds."
|
||||
|
||||
**Immediate Action:** Extend BlockedSignalTracker to 8 hours
|
||||
|
||||
**Next Decision Point:** After collecting 20-30 more signals with proper tracking
|
||||
|
||||
**Expected Outcome:** Data-driven threshold optimization worth $100-250/month
|
||||
|
||||
---
|
||||
|
||||
**Analysis Status:** CORRECTED
|
||||
**Confidence Level:** HIGH (backed by user's chart evidence)
|
||||
**Recommendation:** Implement Option 4 (extend tracking) immediately, then decide on Options 1-3 based on data
|
||||
|
||||
---
|
||||
|
||||
*Initial Analysis: Dec 2, 2025 08:35 UTC (WRONG)*
|
||||
*Correction: Dec 2, 2025 09:15 UTC (ACCURATE)*
|
||||
*Method: 30-min tracking + User chart verification + Extended analysis*
|
||||
*Analyst: AI Trading Assistant (lesson learned: trust user evidence!)*
|
||||
Reference in New Issue
Block a user