Loading dashboard...
Machine learning prediction — estimating price movement probability from on-chain patterns
The AI Score is a gradient-boosted machine learning model trained on historical on-chain data and price outcomes. It learns which combinations of holder behavior, exchange flows, and market dynamics have historically preceded significant price moves.
Unlike our rule-based scores (Squeeze Score, SSS), the AI model discovers patterns we didn't explicitly program. It might find that a specific combination of holder growth rate + exchange outflow speed + volume shape is predictive — patterns too complex for manual rules.
The model outputs a probability score (0–100) and a signal confidence %. Higher = more likely to see significant movement.
Important: This is experimental
The AI Score is our newest signal and is still being refined. It should be used as a supplementary signal alongside the Squeeze Score and SSS — not as a standalone trading indicator. The model is continuously retrained as we collect more outcome data.
AI predictions work best when they confirm what the rule-based scores already show. A token with a high Squeeze Score + high SSS + high AI Score = maximum conviction.
Concentration across top wallets (top-10, top-20, top-50, top-100 %), holder count, and distribution shape.
Current exchange supply %, 7-day flow direction, CEX vs DEX breakdown, and rate of change.
Multi-timeframe price changes (1h, 6h, 24h), volume patterns, buy/sell ratios, and volume acceleration.
Our own squeeze score, SSS score, pattern score, and signal phase as model inputs — layering ML on top of our rule-based systems.
Token age, holder growth trajectory, liquidity depth, and market cap — the model weighs signals differently for new vs established tokens.
Model has high confidence in price movement. Multiple on-chain features align with historical patterns that preceded significant moves.
Some positive features detected but the picture is mixed. Worth watching, especially if combined with a high squeeze or SSS score.
Model doesn't see strong evidence of imminent movement. This doesn't mean the token won't move — it means on-chain data alone isn't compelling.
AI 57 — Model prediction score 0–100. Higher = more on-chain features align with historical winners.
52% signal — Model's confidence that this token will see significant price movement. Above 50% = moderate confidence.
📈 — Signal is trending up compared to previous scan. The model is becoming more bullish.
Confirmation signal — Use alongside Squeeze Score and SSS. When all three agree, conviction is highest.
Discovery tool — The model sometimes flags tokens that rule-based scores miss, surfacing non-obvious patterns.
NOT a standalone indicator — Don't trade on AI score alone. It's one lens among several.
Channel: Telegram (customer channel)
Frequency: Every scan cycle (hourly)
Tokens shown: Top 8 by AI prediction score
Dashboard: AI scores visible on token detail pages at squeeze.icm-analytics.com