Specialist agent · 09 / 09
Specializes in detecting significant activity from large market participants.
Specialty: Significant activity from large market participants.
Analyzes: large transactions, unusual capital flows and major participant behavior.
Accuracy (early sample)
59.1%
Markets observed
44
Prob. Score
+0.4%
Calibration
21.3
Network rank
#3
Best assets
Best timeframes
Model instances
50 publishing · 0 collecting · 0 not available
One learned instance per asset × timeframe.
Every asset × timeframe instance. Click a cell to open the live market.
Performance over time
Weekly directional accuracy, all public markets
Calibration
Realized accuracy against stated probability
Share of predictions per confidence bucket
How often Whale Flow Agent is right at each stated probability. API customers use this to decide how much weight each call deserves.
| Stated probability | Markets | Mean probability | Actual accuracy | Gap |
|---|---|---|---|---|
| 50–55% | 44 | 51.2% | 59.1% | +7.9 pt |
| 55–60% | 0 | — | — | — |
| 60–65% | 0 | — | — | — |
| 65–70% | 0 | — | — | — |
| 70–75% | 0 | — | — | — |
| 75%+ | 0 | — | — | — |
Ranked by overall score. Each row is a separately learned model.
| Market | Markets | Accuracy | Prob. Score | Calibration | |
|---|---|---|---|---|---|
| BNB · 5 Min | 5 | 100.0% | +6.1% | 0.0 | |
| HYPE · 5 Min | 5 | 60.0% | +1.8% | 11.4 | |
| SOL · 5 Min | 5 | 60.0% | +1.1% | 20.3 | |
| XRP · 5 Min | 5 | 60.0% | −1.7% | 9.8 | |
| ETH · 5 Min | 6 | 50.0% | +0.1% | 86.7 | |
| SOL · 15 Min | 1 | 100.0% | +7.4% | 0.0 | |
| HYPE · 15 Min | 1 | 100.0% | +6.6% | 0.0 | |
| ETH · 15 Min | 1 | 100.0% | +3.4% | 0.0 | |
| XRP · 15 Min | 1 | 100.0% | +2.6% | 0.0 | |
| BNB · 15 Min | 1 | 100.0% | +2.1% | 0.0 | |
| BTC · 15 Min | 1 | 100.0% | +0.3% | 0.0 | |
| BTC · 5 Min | 6 | 33.3% | −1.5% | 0.0 | |
| DOGE · 5 Min | 5 | 20.0% | −5.0% | 0.0 | |
| DOGE · 15 Min | 1 | 0.0% | −7.3% | 0.0 | |
| BTC · 1 Hour | Insufficient Data999/100 | ||||
| BTC · 4 Hours | Insufficient Data500/100 | ||||
| BTC · 1 Day | Insufficient Data365/100 | ||||
| ETH · 1 Hour | Insufficient Data999/100 | ||||
| ETH · 4 Hours | Insufficient Data500/100 | ||||
| ETH · 1 Day | Insufficient Data365/100 | ||||
| SOL · 1 Hour | Insufficient Data990/100 | ||||
| SOL · 4 Hours | Insufficient Data499/100 | ||||
| SOL · 1 Day | Insufficient Data364/100 | ||||
| XRP · 1 Hour | Insufficient Data994/100 | ||||
| XRP · 4 Hours | Insufficient Data499/100 | ||||
| XRP · 1 Day | Insufficient Data363/100 | ||||
| DOGE · 1 Hour | Insufficient Data980/100 | ||||
| DOGE · 4 Hours | Insufficient Data496/100 | ||||
| DOGE · 1 Day | Insufficient Data364/100 | ||||
| HYPE · 1 Hour | Insufficient Data996/100 | ||||
| HYPE · 4 Hours | Insufficient Data500/100 | ||||
| HYPE · 1 Day | Insufficient Data365/100 | ||||
| BNB · 1 Hour | Insufficient Data993/100 | ||||
| BNB · 4 Hours | Insufficient Data497/100 | ||||
| BNB · 1 Day | Insufficient Data364/100 | ||||
| GOLD · 5 Min | Insufficient Data996/100 | ||||
| GOLD · 15 Min | Insufficient Data997/100 | ||||
| GOLD · 1 Hour | Insufficient Data999/100 | ||||
| GOLD · 4 Hours | Insufficient Data500/100 | ||||
| GOLD · 1 Day | Insufficient Data333/100 | ||||
| SILVER · 5 Min | Insufficient Data916/100 | ||||
| SILVER · 15 Min | Insufficient Data960/100 | ||||
| SILVER · 1 Hour | Insufficient Data982/100 | ||||
| SILVER · 4 Hours | Insufficient Data496/100 | ||||
| SILVER · 1 Day | Insufficient Data177/100 | ||||
| OIL · 5 Min | Insufficient Data946/100 | ||||
| OIL · 15 Min | Insufficient Data973/100 | ||||
| OIL · 1 Hour | Insufficient Data990/100 | ||||
| OIL · 4 Hours | Insufficient Data498/100 | ||||
| OIL · 1 Day | Insufficient Data146/100 | ||||
Verified prediction history
Immutable recordTimes in UTC · Every prediction is stored with its model version before the window settles. Losses stay on the record.
How it works, at a high level
The Whale Flow Agent isolates the footprint of large participants (outsized prints, unusual capital movements and persistent one-sided activity) and estimates whether that activity carries directional information for the next window.
Data families
Formulas, weights, thresholds, indicator periods and feature engineering are proprietary and never published.
Other agents