Machine Learning models ranked by directional accuracy. Updated daily. No model escapes scrutiny.
STATIC
—
Models Tracked
—
Best 30d Accuracy
—
Best 90d Accuracy
Asset×
All categories use 30D / 90D / 1 year, based on completed outcomes' evaluation dates, not when calls were issued. Each forecast is still evaluated at its seven-day horizon. Accuracy excludes neutral moves within +/-3%; small samples are limited evidence, not proof of trustworthiness.
🏅 Badge Legend — how badges are earned & lost
Badge
Min 30‑day Accuracy
Min Resolved Predictions
Scope
Loading…
Badges are awarded per crypto (e.g. a model can hold a Precision badge for BTC
but only a Reliable badge for ETH) and an 'ALL' badge reflecting cross‑asset
performance. Badges are automatically revoked if accuracy drops below the tier
threshold or the prediction count falls below the minimum.
#
Model Name
Type
Acc 30d
Acc 90d
Acc 1yr
Predictions (1yr)
Avg return %
Return/risk
Confidence i
Reliability
Loading AI model rankings…
Scoring v2: price moves from -3% through +3%, inclusive, are neutral for both BUY and SELL. Acc = wins / (wins + losses); HOLD and neutrals are excluded. W/L/N shows wins, losses and neutrals in the selected window. Decisive coverage = (wins + losses) / all evaluated directional calls. Avg return includes neutral calls (BUY long, SELL short), not compounded portfolio profit. Return/risk = mean return / sample standard deviation, not annualized. Costs are excluded. Reliability = Wilson 95% lower bound on decisive accuracy, not a probability of future success. Original verdicts are retained; reclassification does not repair publication timing or claim-extraction errors.
Rankings update daily based on real prediction outcomes. Past accuracy does not guarantee future performance.
Signals are educational only and do not constitute financial advice.
#
Source / Analyst
Signal
Acc 30d
Acc 90d
Acc 1yr
Predictions (lifetime)
Avg return %
Return/risk
Conviction % i
Reliability
Cost
Loading source rankings
Accuracy is computed by comparing each source's publicly-stated directional predictions against 7-day price outcomes.
Prediction totals are lifetime tracked records, including pending and inconclusive outcomes; explicit synthetic fixtures are excluded. Legacy records without linked evidence remain marked unverified. Scored counts and accuracy use the selected rolling window and can decrease as older calls age out.
Scoring v2 applies equally to models and sources: -3% through +3%, inclusive, is neutral. Accuracy = wins / (wins + losses); W/L/N and decisive coverage expose the excluded neutral sample. Avg return includes neutrals (BUY long, SELL short); it is a gross mean, not portfolio profit. Costs are excluded. Reliability is the Wilson 95% lower bound on decisive accuracy, not a guarantee. Original verdicts are retained; missing price evidence is not treated as neutral. Posts older than seven days are marked stale.
Historical predictions are tracked and scored by ArtinFox AI. Not financial advice. Data may differ from self-reported accuracy.
#
Source
Signal
Acc 30d
Acc 90d
Acc 1yr
Predictions (lifetime)
Avg return %
Return/risk
Conviction % i
Reliability
Cost
Loading web sources rankings
Web Sources tracks web-based prediction platforms and crypto news outlets that publish directional price forecasts.
Scoring v2: -3% through +3%, inclusive, is neutral. Accuracy = wins / (wins + losses). W/L/N and decisive coverage include the neutral sample. Avg return includes neutrals (BUY long, SELL short); it is a gross mean, not portfolio profit. Costs are excluded. Reliability is the Wilson 95% lower bound on decisive accuracy, not a future win probability. Original verdicts are retained. Stale posts are not current signals. Not financial advice.
#
Source / Analyst
Signal
Acc 30d
Acc 90d
Acc 1yr
Predictions (lifetime)
Avg return %
Return/risk
Conviction % i
Reliability
Cost
Loading all sources
All influencer, research, and web prediction sources ranked together by directional accuracy.
Scoring v2: -3% through +3%, inclusive, is neutral. Accuracy = wins / (wins + losses). W/L/N and decisive coverage include the neutral sample. Avg return includes neutrals (BUY long, SELL short); it is a gross mean, not portfolio profit. Costs are excluded. Reliability is the Wilson 95% lower bound on decisive accuracy, not a future win probability. Original verdicts are retained. Stale posts are not current signals. Not financial advice.
Model Comparison
Head-to-Head Accuracy Timeline
vs
Select Pro tier to compare models. Choose two models above and click Compare.