Forecast Evolution — How Predictions Change Over Time
The /forecast-evolution endpoint reveals how the forecast for a symbol has changed over time — direction flips, confidence drift, upside_prob evolution, forecast range revision, and audited accuracy of each revision. Each revision includes market_key (e.g. btc-10m, btc-1h, or null for background scheduler predictions) and the timeframes used. This reveals the trajectory of the model's opinion, not just the latest reading.
Endpoint
GET /api/feeds/kronos/forecast-evolution/:symbolKey?hours=24Parameters
| Param | Type | Required | Description |
|---|---|---|---|
symbolKey | path | Yes | Asset symbol, e.g. btc_usdt, eth_usdt |
hours | query | No | Lookback in hours (default: 24, max: 168) |
market_key | query | No | Filter by specific market, e.g. btc-10m, btc-1h |
limit | query | No | Cap number of revisions (default: 144, max: 1000) |
summary | query | No | 1 for summary-only (no revisions array) |
Pricing
$0.005 per request (payment after delivery — settled only if the response is successfully sent).
Example request
curl -H "X-PAYMENT: <x402-payment-header>" \
"https://kronos.seshat.markets/api/feeds/kronos/forecast-evolution/btc_usdt?hours=24"Example response (truncated)
{
"symbol": "btc_usdt",
"hours": 24,
"summary": {
"total_revisions": 18,
"direction_flips": 2,
"confidence_drift": -0.04,
"upside_prob_start": 0.68,
"upside_prob_end": 0.61
},
"revisions": [
{
"revision_id": 1,
"predicted_at": "2025-06-15T00:00:00Z",
"market_key": "btc-1h",
"direction": "up",
"upside_prob": 0.68,
"confidence": 0.72,
"conformal_lower": 67000,
"conformal_upper": 68500,
"audited": true,
"correct": true,
"timeframes": ["1h", "4h"]
},
{
"revision_id": 2,
"predicted_at": "2025-06-15T06:00:00Z",
"market_key": "btc-1h",
"direction": "up",
"upside_prob": 0.64,
"confidence": 0.69,
"conformal_lower": 66800,
"conformal_upper": 68300,
"audited": false,
"timeframes": ["1h", "4h"]
},
{
"revision_id": 3,
"predicted_at": "2025-06-15T12:00:00Z",
"market_key": null,
"direction": "down",
"upside_prob": 0.42,
"confidence": 0.55,
"conformal_lower": 66500,
"conformal_upper": 68000,
"audited": false,
"timeframes": ["1h", "4h", "1d"]
}
]
}Use cases
Flip Detection
When the model flips direction (up→down), it signals a regime change. Track flip frequency — rapid flips mean high uncertainty; a stable direction over 24h means high conviction.
Confidence Trend Analysis
If upside_prob drifts from 0.68 to 0.42 over 12 hours, the model is losing conviction. Act on the trend, not just the latest value — a drifting signal is weaker than a stable one.
Revision Quality Control
Each revision has an audited correct flag. If early revisions are correct but later ones flip wrong, the model over-adjusted. Use this to filter out volatile periods.
Market-Specific Filtering
Filter by ?market_key=btc-1h to see only predictions from the 1h market scheduler. Isolate signal sources — background scheduler predictions vs market-triggered ones may have different accuracy profiles.
Frequently asked questions
How do I track how a crypto forecast changes over time?
GET /api/feeds/kronos/forecast-evolution/btc_usdt?hours=24. It shows direction flips, confidence drift, upside_prob evolution, and audited accuracy of each revision. You see the trajectory of the model's opinion, not just the latest reading.What does it mean when a crypto prediction model flips direction?
How much does the forecast evolution API cost?
$0.005 per request (payment after delivery). Pay with USDC on Solana or Base via x402. Use ?summary=1 for a compact summary without the full revisions array.Can I see confidence drift in crypto predictions?
confidence_drift field shows how confidence has changed over the lookback window. If upside_prob drifts from 0.68 to 0.42 over 12 hours, the model is losing conviction — a drifting signal is weaker than a stable one.