Trading Signals
How signals are generated, signal types, confidence scoring, and tracking performance.
What Are Trading Signals?
Thrive's trading signals are AI-generated directional views on individual assets, produced by machine learning models that analyze multi-dimensional market data in real time. Each signal includes a direction (long or short), a confidence score (0-100), a categorized signal type, the contributing data sources, and an AI-written rationale explaining why the signal was generated.
Signals are not trade recommendations. They are quantitative outputs from models trained on historical derivatives, on-chain, and price data. The confidence score reflects the model's assessed probability that the directional view will be correct over the signal's defined time horizon. You should always validate signals against your own analysis and risk framework before acting on them.
Not financial advice
Signal Types
Every signal is classified into one of four categories based on the market condition the model has identified. Understanding the signal type helps you contextualize the setup and apply the appropriate trading strategy.
Momentum
Momentum signals are generated when the model detects strong directional persistence backed by rising volume, increasing open interest, and aligned funding rates. These signals favor trend continuation. They work best in trending markets and tend to underperform in range-bound conditions. Momentum signals typically have a 4-hour to 24-hour time horizon.
| Best Market Condition | Trending markets with clear directional bias. |
| Typical Horizon | 4 hours to 24 hours. |
| Key Inputs | Price momentum, volume profile, OI changes, funding alignment. |
| Risk Profile | Moderate. Momentum can reverse sharply at overextended levels. |
Reversal
Reversal signals fire when the model identifies conditions consistent with a directional change: statistical extremes in z-scores, funding rate capitulation, divergence between price and derivatives data, or on-chain accumulation at support levels. Reversal signals are inherently contrarian and carry higher risk but also higher reward when correct.
| Best Market Condition | Extended trends approaching statistical extremes. |
| Typical Horizon | 24 hours to 7 days. |
| Key Inputs | Z-scores, divergence detection, funding rate extremes, exchange netflows. |
| Risk Profile | Higher. Reversals can be early. Position sizing should be conservative. |
Breakout
Breakout signals are generated when the model detects compression in volatility (narrowing Bollinger Bands, declining ATR) combined with building open interest and volume. The signal anticipates an imminent expansion in range. The direction is determined by the balance of derivatives positioning and order flow bias.
| Best Market Condition | Consolidation ranges with declining volatility and building OI. |
| Typical Horizon | 1 hour to 12 hours from signal generation. |
| Key Inputs | Volatility compression, OI accumulation, volume patterns, order flow imbalance. |
| Risk Profile | Moderate to high. False breakouts are common. Use tight invalidation levels. |
Divergence
Divergence signals are triggered when the model detects a meaningful disagreement between price and one or more derivatives metrics (funding rate, open interest, volume). These signals overlap with the standalone Divergence Detection module but include the additional context of AI confidence scoring and multi-factor analysis.
| Best Market Condition | Any market where price and derivatives data are in conflict. |
| Typical Horizon | 12 hours to 3 days. |
| Key Inputs | Price-funding divergence, price-OI divergence, volume divergence. |
| Risk Profile | Moderate. Divergences are statistically robust but timing can be imprecise. |
Confidence Scoring
Every signal carries a confidence score from 0 to 100 that represents the model's assessed probability of the directional view being correct. The score is not a percentage prediction of price movement magnitude; it is a probability estimate of directional accuracy.
| 80-100 (High) | Strong multi-factor alignment. Multiple data sources confirm the signal. Historically the most reliable tier. |
| 60-79 (Medium) | Good directional alignment with some conflicting data points. The majority of signals fall in this range. |
| 40-59 (Low-Medium) | Mixed signals across data sources. Use as a watchlist trigger rather than an action trigger. |
| Below 40 (Low) | Weak alignment. The model detects a potential setup but lacks confirming data. Exercise caution. |
Filter by confidence
Signal Data Sources
Each signal lists the data sources that contributed to its generation. This transparency lets you validate the signal by checking the underlying data yourself. Signals that draw from more data sources tend to be more robust.
Multi-timeframe price structure, support/resistance levels, and trend analysis.
Cross-exchange funding rate data with z-score context.
OI levels, changes, and the OI-price relationship matrix.
Volume patterns, anomalies, and relative volume compared to historical averages.
Exchange netflows, smart money positioning, and large transaction data.
Aggregated social sentiment from Twitter, Telegram, and Discord.
Liquidation clusters and estimated liquidation levels that could act as magnets.
Signal Lifecycle
Signals move through a defined lifecycle from generation to resolution. Understanding this lifecycle helps you know when to act and when to wait.
Generation
The model processes the latest data and generates a new signal with a direction, confidence score, type classification, and rationale. The signal appears on the Signals page and triggers notifications for users who have alert rules configured for that asset.
Active monitoring
While active, the signal's confidence score is updated as new data arrives. If the underlying conditions strengthen, confidence increases. If conditions deteriorate, confidence decreases. You can see the confidence trajectory on the signal detail view.
Resolution
A signal resolves when it either (a) reaches its time horizon, (b) hits a predefined invalidation level, or (c) the model explicitly closes it due to changed conditions. Resolved signals are tagged as "Win" or "Loss" based on whether price moved in the predicted direction.
Performance tracking
All resolved signals are recorded in the historical performance database. You can view aggregate performance stats, filter by signal type, confidence tier, and asset, and track how signal accuracy evolves over time.
Historical Performance
The Signals page includes a performance dashboard that tracks the historical accuracy of AI signals across multiple dimensions. This data is fully transparent and updated in real time as signals resolve.
Key Performance Metrics
Percentage of signals where price moved in the predicted direction by the time horizon.
The mean confidence score across all generated signals.
Win rate broken down by confidence brackets (80+, 60-79, 40-59, below 40).
Win rate for momentum, reversal, breakout, and divergence signals individually.
Win rate per asset, helping you identify which assets the model performs best on.
The mean price change in the predicted direction across all resolved signals.
A trailing window metric that shows recent model performance, more relevant than all-time stats.
Full transparency
Credit Costs
Signal-related actions consume credits from your monthly allocation. The cost structure is designed so that casual signal browsing is affordable, while deeper AI analysis costs more to reflect the computational resources involved.
| View a signal | 2 credits. Includes direction, confidence, type, and contributing sources. |
| AI signal interpretation | 10 credits. Expands the AI rationale with a detailed narrative explanation of the setup. |
| Signal performance query | 5 credits. Runs a filtered query against the historical performance database. |
| AI deep research on a signal | 200 credits. Generates a comprehensive research report covering the signal from every angle. |
Optimize credit usage
Using Signals Effectively
Signals are most powerful when used as one input in a broader decision-making framework, not as standalone triggers. Here is a practical workflow for incorporating signals into your trading.
Scan for high-confidence signals
Open the Signals page and filter for confidence above 70. Sort by most recent. This gives you a short list of the model's highest-conviction current views.
Check the signal type
Is this a momentum continuation, a reversal, a breakout, or a divergence setup? Each type requires a different entry strategy and risk management approach.
Validate on the asset page
Navigate to the asset's detail page. Check the z-scores, funding rate history, OI trends, and on-chain data. Do you see independent confirmation of the signal's thesis?
Size your position accordingly
High-confidence signals with multi-source confirmation warrant larger position sizes within your risk framework. Lower confidence or single-source signals should be sized conservatively or skipped entirely.