Validate Predictions
Review all pending predictions and check them against current market data.
Workflow
Step 0: Auto-Find Expired Predictions
Delegate using the Task tool with subagent_type: general-purpose and model: opus:
"You are the learning-agent. Read agents/learning-agent.md for your analysis framework. First, use get_crypto_prices() from crypto-data MCP to get current prices for major coins (bitcoin, ethereum, solana, etc.). Then call find_expired_predictions(current_prices='{\"BTC/USDT\": ..., \"ETH/USDT\": ...}') from crypto-learning-db to discover predictions whose timeframe has passed. For each expired prediction, reason about how close it was and validate with an NL evaluation using validate_prediction(). Do NOT use the Edit tool."
Step 1: Check Remaining Pending Predictions
Delegate using the Task tool with subagent_type: general-purpose and model: opus:
"You are the learning-agent. Call query_predictions(status='pending') from crypto-learning-db for predictions still within their timeframe. For each prediction:
- Use get_exchange_prices(symbol=...) from crypto-exchange MCP to check current price
- Compare current price against the prediction's target_value
- Report current progress toward or away from target
Do NOT use the Edit tool."
Step 2: Present Results
Show a summary table:
## Prediction Validation Report
### Resolved This Check
| ID | Agent | Prediction | Target | Actual | Result |
|----|-------|-----------|--------|--------|--------|
### Still Pending
| ID | Agent | Prediction | Target | Current | Progress | Expires |
|----|-------|-----------|--------|---------|----------|---------|
### Overall Accuracy
- Total predictions: X
- Correct: X (X%)
- Incorrect: X (X%)
- Pending: X
### Track Record by Setup Type
| Setup Type | Total | Correct | Accuracy | Trend |
|-----------|-------|---------|----------|-------|