DeFiStar.io

Real-time Stablecoin Yields Tracker

Home Predictive Dashboard APY Momentum Reliability TVL Momentum Forecast Signal Confluence Predictions Guide Forex Intelligence

⚡ SYSTEM DIAGNOSTIC REPORT ⚡

Generated: 23 Jan 2026, 15:00 UTC

DeFiStar Prediction Engine V1 | Complete System Analysis

📊 SYSTEM OVERVIEW
Total Predictions
328
Verified
231
Overall Accuracy
17.3%
Data Freshness
26 mins ago
Tracked Pools
65
History Records
91,987
🔧 ENGINE PERFORMANCE BREAKDOWN
Engine Total Verified Pending Correct Incorrect Neutral Accuracy Avg MAE Status
APY Momentum 46 22 24 11 11 0 50% 5.603% MODERATE
Reliability Index 15 13 2 3 10 0 23.1% N/A NEEDS TUNING
TVL Momentum 55 41 14 11 30 0 26.8% 21.002% NEEDS TUNING
Utilisation Forecast 212 155 57 15 9 131 9.7% 6.162% NEEDS TUNING
↔ Swipe left/right to see full table on mobile
🧠 GARCH-LSTM NEURAL NETWORK ENGINE
ℹ️ DEEP LEARNING PREDICTION SYSTEM
This is a separate ML-based system using GARCH (volatility forecasting) + LSTM (deep learning) to predict APY, TVL, and Utilisation. Unlike the rule-based engines above, this system learns patterns from historical data and continuously adapts through incremental training.

Key Differences:
  • Prediction Window: Fixed 7-day forecast for all metrics
  • Approach: Neural network (vs rule-based EMA/signals)
  • Learning: Models train daily with incremental learning (2 epochs/day)
  • Metrics: APY, TVL, Utilisation only (Reliability NOT predicted)
  • Storage: Separate table defi_ml_predictions
Total Predictions
7,567
Verified
6,079
Accuracy
42.1%
Trained Models
161
Avg Model Size
148.3 KB
Next Verification
23 Jan 00:07

Per-Metric Performance

Metric Total Verified Pending Correct Incorrect Accuracy Avg Confidence Avg Volatility Avg MAE Status
apy 2530 2035 495 790 1245 38.8% 80.4% 4.68 0.6168 ❌ NEEDS TUNING
tvl 2515 2020 495 918 1102 45.4% 86.5% 64.7 40,258,640.2505 ⚠️ MODERATE
utilization 2522 2024 498 854 1170 42.2% 88.5% 4.66 4.3348 ⚠️ MODERATE
↔ Swipe left/right to see full table on mobile

Volatility & Confidence Distribution

Volatility Range Confidence Range Predictions Verified Correct Accuracy
High (10-15) Low (40-60%) 85 85 58 68.2%
High (10-15) Very Low (<40%) 34 34 13 38.2%
Low (<5) High (80%+) 5016 5016 2047 40.8%
Low (<5) Medium (60-80%) 103 103 29 28.2%
Medium (5-10) Low (40-60%) 150 150 61 40.7%
Medium (5-10) Medium (60-80%) 262 262 108 41.2%
Very High (15+) Very Low (<40%) 429 429 246 57.3%
↔ Swipe left/right to see full table on mobile
📊 VOLATILITY INSIGHT
Pattern Analysis: Check if high volatility predictions have lower accuracy. If so, consider filtering predictions when volatility > 15.

Model Training Coverage

Metric Unique Pools Expected Models Actual Models Coverage Status
apy 56 56 53 94.6% ✅ COMPLETE
tvl 56 56 54 96.4% ✅ COMPLETE
utilization 56 56 54 96.4% ✅ COMPLETE
↔ Swipe left/right to see full table on mobile

Recently Trained Models (Last 10)

Model File Size Last Modified Age
Venus_USDT_BNB Chain_utilization.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Venus_USDT_BNB Chain_tvl.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Venus_USDT_BNB Chain_apy.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Venus_USDC_BNB Chain_utilization.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Venus_USDC_BNB Chain_tvl.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Venus_USDC_BNB Chain_apy.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Spark_USDT_Gnosis_utilization.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Spark_USDT_Gnosis_tvl.keras 148.3 KB 02 Jan 2026 23:42 495.3 hours ago
Spark_USDT_Gnosis_apy.keras 148.3 KB 02 Jan 2026 23:41 495.3 hours ago
Spark_USDT_Ethereum_utilization.keras 148.3 KB 02 Jan 2026 23:41 495.3 hours ago
↔ Swipe left/right to see full table on mobile
⚠️ STALE MODELS DETECTED
Some models haven't been updated in over 7 days. Ensure defi_lstm_engine.py is running daily via cron.
Check: crontab -l | grep lstm

🔧 Tuning Opportunities

Priority Issue Recommended Action File/Location
HIGH APY accuracy very low (38.8%) Review apy LSTM architecture. Consider: (1) Increasing data window from 200 to 720 points, (2) Adding more LSTM neurons (50→100), (3) Using different activation functions, (4) Collecting more training data over time defi_lstm_engine.py
MEDIUM TVL accuracy below target (45.4%) Tune tvl hyperparameters. Try: (1) Increasing training epochs from 10→20 for new models, (2) Increasing incremental epochs from 2→5, (3) Adding dropout layers (0.2) to prevent overfitting defi_lstm_engine.py lines 155-157
MEDIUM UTILIZATION accuracy below target (42.2%) Tune utilization hyperparameters. Try: (1) Increasing training epochs from 10→20 for new models, (2) Increasing incremental epochs from 2→5, (3) Adding dropout layers (0.2) to prevent overfitting defi_lstm_engine.py lines 155-157
HIGH Data window may be too short (200 points) For 7-day forecasts, 200 data points may be insufficient. Increase to 720 points (30 days at hourly intervals) for better pattern recognition. This is line 100: limit=200 → limit=720 defi_lstm_engine.py line 100
MEDIUM Fixed 7-day window for all metrics APY and TVL change faster than Utilisation. Consider adaptive windows: {'apy': 3, 'tvl': 3, 'utilization': 7} for better accuracy on fast-moving metrics. defi_lstm_engine.py line 28
↔ Swipe left/right to see full table on mobile

Recent Verified Predictions (Last 15)

Verified Pool Metric Direction Start→Predicted Actual Error Confidence Volatility Outcome
23 Jan 12:00 mwUSDC
Morpho 		apy FLAT 4.42 → 4.44 3.92 0.5179 93.6% 1.28 INCORRECT
23 Jan 12:00 USDT
Venus 		utilization FLAT 77.47 → 78.21 81.49 3.2817 74.8% 5.04 INCORRECT
23 Jan 12:00 USDT
Compound V3 		utilization FLAT 81.11 → 81.31 82.46 1.1469 93.6% 1.29 CORRECT
23 Jan 12:00 mwUSDC
Morpho 		tvl DOWN 34,904,616.25 → 34,118,312.76 31,740,954.99 2,377,357.7676 60.6% 7.89 CORRECT
23 Jan 12:00 steakUSDC
Morpho 		apy UP 3.72 → 3.75 3.74 0.0123 55% 8.99 INCORRECT
23 Jan 12:00 steakETH
Morpho 		apy UP 1.91 → 1.93 2.05 0.1198 56.2% 8.77 CORRECT
23 Jan 12:00 steakUSDC
Morpho 		tvl FLAT 439,401,269.86 → 440,226,578.16 426,065,485.38 14,161,092.7778 96.9% 0.61 INCORRECT
23 Jan 12:00 USDT
Compound V3 		tvl DOWN 4,731,842.03 → 4,525,220.39 4,443,912.45 81,307.9393 95.8% 0.84 CORRECT
23 Jan 12:00 USDT
Compound V3 		tvl FLAT 1,826,099.27 → 1,831,925.02 1,805,685.41 26,239.6146 93.6% 1.29 CORRECT
23 Jan 12:00 USDT
Compound V3 		apy FLAT 2.94 → 2.95 3.15 0.2045 94.4% 1.12 INCORRECT
23 Jan 12:00 mwUSDC
Morpho 		utilization FLAT 81.23 → 81.17 79.83 1.3416 0% 96.84 CORRECT
23 Jan 12:00 steakETH
Morpho 		tvl FLAT 98,881,632.69 → 99,157,534.39 99,561,256.22 403,721.8313 91.9% 1.62 CORRECT
23 Jan 12:00 steakUSDC
Morpho 		utilization FLAT 86.51 → 86.30 89.56 3.2583 0% 96.87 INCORRECT
23 Jan 12:00 USDT
Compound V3 		utilization FLAT 73.43 → 73.37 78.85 5.4754 95.6% 0.88 INCORRECT
23 Jan 12:00 USDT
Compound V3 		apy FLAT 3.24 → 3.26 3.30 0.0441 89% 2.21 INCORRECT
↔ Swipe left/right to see full table on mobile

→ View complete GARCH-LSTM audit trail (all verified predictions)

📅 UPCOMING VERIFICATIONS
Engine Pool Prediction Made Due Status
Utilisation bbETH (Morpho) Decay 03 Jan 00:00 10 Jan 00:00 OVERDUE
Utilisation bbETH (Morpho) Decay 10 Jan 00:00 17 Jan 00:00 OVERDUE
Reliability mwUSDC (Morpho) Maintain robust 16 Jan 20:00 23 Jan 20:00 PENDING
Utilisation bbETH (Morpho) Decay 17 Jan 00:00 24 Jan 00:00 PENDING
APY USDC (Compound V3) Cooling down 21 Jan 21:15 26 Jan 21:15 PENDING
Reliability FDUSD (Aave V3) Maintain robust 21 Jan 20:00 28 Jan 20:00 PENDING
TVL USDC (Aave V3) Capital inflow 21 Jan 20:00 28 Jan 20:00 PENDING
TVL USDC (Aave V3) Capital inflow 21 Jan 20:00 28 Jan 20:00 PENDING
TVL USDC (Aave V3) Capital inflow 21 Jan 20:00 28 Jan 20:00 PENDING
TVL USDC (Compound V3) Capital flight 21 Jan 21:00 28 Jan 21:00 PENDING
APY EURC (Aave V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY USDC (Aave V3) Cooling down 21 Jan 21:15 28 Jan 21:15 PENDING
APY DAI (Aave V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY USDC (Compound V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY USDC (Compound V3) Cooling down 21 Jan 21:15 28 Jan 21:15 PENDING
APY LUSD (Aave V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY USDT (Compound V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY USDS (Aave V3) Cooling down 21 Jan 21:15 28 Jan 21:15 PENDING
APY crvUSD (Aave V3) Heating up 21 Jan 21:15 28 Jan 21:15 PENDING
APY DAI (Aave V3) Cooling down 21 Jan 21:15 28 Jan 21:15 PENDING
↔ Swipe left/right to see full table on mobile
⚠️ OVERDUE VERIFICATIONS
2 prediction(s) are past their verification date. Run: php verify-all.php --verify --clear
✅ RECENT VERIFICATIONS (Last 10)
Engine Pool Prediction Outcome MAE Verified Details
Utilisation USDS (Sky Protocol) Stable NEUTRAL 0.000% 23 Jan 12:00
Utilisation USDS (Sky Protocol) Stable NEUTRAL 0.000% 23 Jan 12:00
Utilisation DAI (Maker DSR) Decay NEUTRAL 0.000% 23 Jan 00:00
Utilisation DAI (Maker DSR) Decay NEUTRAL 0.000% 23 Jan 00:00
Utilisation LUSD (Aave V3) Stable NEUTRAL 0.300% 23 Jan 00:00
Utilisation USDtb (Aave V3) Decay NEUTRAL 2.570% 22 Jan 18:00
Utilisation EURC (Aave V3) Stable NEUTRAL 4.560% 22 Jan 06:00
Utilisation RLUSD (Aave V3) Decay NEUTRAL 0.040% 22 Jan 06:00
Utilisation USDC (Compound V3) Stable NEUTRAL 17.540% 22 Jan 06:00
Utilisation USDC (Aave V3) Decay NEUTRAL 0.170% 22 Jan 06:00
↔ Swipe left/right to see full table on mobile
💾 DATABASE TABLE STATUS
Table Description Row Count Size Status
pools_history Historical pool data (hourly snapshots) 83,096 24.55 MB EXISTS
yield_predictions APY Momentum predictions 45 0.08 MB EXISTS
reliability_predictions Reliability Index predictions 15 0.14 MB EXISTS
tvl_predictions TVL Momentum predictions 53 0.16 MB EXISTS
utilization_predictions Utilisation Forecast predictions 208 0.09 MB EXISTS
signal_confluence_audit Signal Confluence records 13 0.06 MB EXISTS
↔ Swipe left/right to see full table on mobile
📈 PREDICTION TYPE BREAKDOWN

APY Momentum Types

Type Total Verified Correct Accuracy
Heating Up 26 15 6 40%
Cooling Down 20 7 5 71.4%

Utilisation Forecast Types

Type Total Verified Correct Accuracy
Extreme Surge 32 24 15 62.5%
Rising 46 34 0 0%
Stable 107 79 0 0%
Decay 27 18 0 0%
🏆 POOL PERFORMANCE ANALYSIS

🏆 Best Performing Pools

Need 3+ verified predictions per pool for rankings

⚠️ Worst Performing Pools

Need 3+ verified predictions per pool for rankings
💡 SYSTEM RECOMMENDATIONS
Priority Issue Recommended Action Command
HIGH 2 overdue verifications Run verification script to process pending predictions php verify-all.php --verify --clear
HIGH Low overall accuracy (17.3%) Review prediction thresholds and algorithm parameters. Consider tuning EMA periods or confidence thresholds. Analyze worst performing pools for patterns
↔ Swipe left/right to see full table on mobile
📋 COPY-PASTE SUMMARY FOR SHARING
ℹ️ HOW TO USE
Click "Show Text Report" below, then copy the entire text block and paste it into your message to get expert help with system tuning and optimization. 
=============================================================================
DEFISTAR PREDICTION ENGINE - SYSTEM DIAGNOSTIC REPORT
Generated: 23 Jan 2026, 15:00 UTC
=============================================================================

SYSTEM OVERVIEW:
- Total Predictions: 328
- Verified: 231
- Overall Accuracy: 17.3%
- Data Freshness: 26 minutes old
- Tracked Pools: 65
- History Records: 91,987

ENGINE PERFORMANCE:
APY Momentum:
  - Total: 46 | Verified: 22 | Pending: 24
  - Correct: 11 | Incorrect: 11 | Neutral: 0
  - Accuracy: 50% | Avg MAE: 5.603%
  - Next Due: 26 Jan 2026 21:15
Reliability Index:
  - Total: 15 | Verified: 13 | Pending: 2
  - Correct: 3 | Incorrect: 10 | Neutral: 0
  - Accuracy: 23.1% | Avg MAE: N/A
  - Next Due: 23 Jan 2026 20:00
TVL Momentum:
  - Total: 55 | Verified: 41 | Pending: 14
  - Correct: 11 | Incorrect: 30 | Neutral: 0
  - Accuracy: 26.8% | Avg MAE: 21.002%
  - Next Due: 28 Jan 2026 20:00
Utilisation Forecast:
  - Total: 212 | Verified: 155 | Pending: 57
  - Correct: 15 | Incorrect: 9 | Neutral: 131
  - Accuracy: 9.7% | Avg MAE: 6.162%
  - Next Due: 10 Jan 2026 00:00

OVERDUE VERIFICATIONS: 2
ACTION REQUIRED: Run php verify-all.php --verify --clear

RECOMMENDATIONS:
[HIGH] 2 overdue verifications
  Action: Run verification script to process pending predictions
  Command: php verify-all.php --verify --clear
[HIGH] Low overall accuracy (17.3%)
  Action: Review prediction thresholds and algorithm parameters. Consider tuning EMA periods or confidence thresholds.
  Command: Analyze worst performing pools for patterns

=============================================================================
✕
INFO
SIGNAL CONFLUENCE
Multi-engine consensus trading signals.
GO

Notice: This website provides informational analytics and data services only. We are not authorised or regulated by the Financial Conduct Authority (FCA). We do not offer, facilitate, or provide any financial services or products, including cryptocurrencies, digital assets, or derived products. Content on this site does not constitute financial advice.
DeFiStar.io is an independent data utility. We do not accept listing fees, we do not have an affiliate relationship with protocols, and we do not sell financial products. Our rankings are 100% algorithmic based on on-chain liquidity and smart contract data.
By accessing or using this website, you agree to be bound by our Terms of Service and acknowledge our Risk Disclosures.

Learn more: About Our Mission | Ranking Methodology |
DeFiStar.io, Clyde Offices, 2nd Floor, 48 West George Street, Glasgow, G2 1BP, Scotland, United Kingdom
© 2025–2026 defistar.io