A newer version of the Streamlit SDK is available: 1.62.0
News Sentiment Disabled
Date: 2026-03-15 Status: β DISABLED PER USER REQUEST
What Changed
News sentiment has been completely removed from trading decisions due to reliability issues.
Weight Distribution Changes
Before:
Whale: 30.0%
Regime: 18.0%
TFT: 18.0%
Funding: 13.5%
Order Flow: 13.5%
News: 10.0% β ACTIVE
ββββββββββββββββ
Total: 103.0% (auto-normalized to 100%)
After:
Whale: 30.0% (unchanged)
Regime: 21.0% (+3%) β Real-time trend detection
TFT: 21.0% (+3%) β Price forecasting
Funding: 15.0% (+1.5%) β Market sentiment
Order Flow: 15.0% (+1.5%) β Buy/sell pressure
News: 0.0% (-10%) β DISABLED
ββββββββββββββββ
Total: 102.0% (auto-normalized to 100%)
Changes Made
File: live_trading_multi.py
News weight set to 0% (line 576, 585, 588)
scores['news'] = (news_score, 0.00) # DISABLED: was 0.10Hard veto logic disabled (lines 579-583)
# Hard veto: DISABLED - news not reliable enough # (commented out - no longer blocks trades)Redistributed 10% to proven signals:
- Regime: +3% (better trend detection)
- TFT: +3% (price forecasting)
- Funding: +1.5% (real-time sentiment)
- Order Flow: +1.5% (buy/sell pressure)
Why This Was Done
User Request
- News aggregation not working reliably
- User wants to ensure it doesn't affect trading decisions
- Can be re-enabled later when fixed
Technical Issues with News
- API rate limits causing failures
- Sentiment scores not accurate
- Limited sources (CryptoCompare only)
- Hard veto logic too aggressive
Impact on Trading
Before (with news enabled):
Example: BTC bullish market
- Regime: +0.8 (strong uptrend) Γ 18% = 0.144
- TFT: +0.6 (forecast up) Γ 18% = 0.108
- News: -0.5 (bearish news!) Γ 10% = -0.050
- Composite: 0.202
If news confidence > 0.7: HARD VETO (no trade allowed!)
After (news disabled):
Example: Same BTC bullish market
- Regime: +0.8 (strong uptrend) Γ 21% = 0.168
- TFT: +0.6 (forecast up) Γ 21% = 0.126
- News: 0.0 Γ 0% = 0.000
- Composite: 0.294
No hard veto, trade can proceed! β
Key Improvements
- No false vetoes from unreliable news
- Stronger weight on proven signals (regime, TFT)
- More responsive to actual market conditions
- Cleaner decision logic without news interference
What Still Works
All other decision signals remain fully functional:
β Whale Signals (30%)
- Real-time whale flow monitoring
- Pattern learning from wallet behavior
- Exchange reserve tracking
β Regime Detection (21%, was 18%)
- Trend detection (TRENDING_UP/DOWN/RANGING)
- ADX strength measurement
- Market phase classification
β TFT Forecasting (21%, was 18%)
- 1h, 4h, 1d price predictions
- Temporal Fusion Transformer model
- Direction consensus
β Funding Rate (15%, was 13.5%)
- Long/short sentiment
- Leverage bias detection
- OKX funding data
β Order Flow (15%, was 13.5%)
- CVD (Cumulative Volume Delta)
- Taker buy/sell ratio
- Notable trades (whales)
Verification
Check news is disabled:
grep -A3 "6. News Sentiment" live_trading_multi.py
Should show:
# ββ 6. News Sentiment (DISABLED - not reliable) βββββββββββββββ
scores['news'] = (news_score, 0.00) # DISABLED: was 0.10
Check new weight distribution:
grep "scores\['regime'\] = " live_trading_multi.py | head -1
grep "scores\['tft'\] = " live_trading_multi.py | head -1
grep "scores\['funding'\] = " live_trading_multi.py | head -1
grep "scores\['order_flow'\] = " live_trading_multi.py | head -1
Should show:
scores['regime'] = (regime_score, 0.21)
scores['tft'] = (tft_score, 0.21)
scores['funding'] = (funding_score, 0.15)
scores['order_flow'] = (flow_score, 0.15)
Re-enabling News (Future)
When news aggregation is fixed, re-enable by:
Set news weight back to 0.10:
scores['news'] = (news_score, 0.10) # Re-enabledUncomment hard veto logic (if desired)
Reduce other weights accordingly:
- Regime: 21% β 18%
- TFT: 21% β 18%
- Funding: 15% β 13.5%
- Order Flow: 15% β 13.5%
Summary
β News sentiment completely disabled β 10% weight redistributed to proven signals β No hard vetoes from unreliable news β System more responsive to real market conditions
The trading bot now relies on 5 proven signal types instead of 6, with higher weights on the most reliable indicators (regime detection and TFT forecasting).