For traders and analysts
Compare horizons, review the three engines and place each forecast next to the price history.
LIVE MARKET FORECASTS · 200+ MARKETS · 8 HORIZONS
A forecast is only useful if you can check it later.
That is how Avaltix works. It follows more than 200 markets, records each forecast before its horizon closes, and shows what happened next. Explore it in the dashboard or use the same data in your product or agent.
Recorded before outcomeMethodology publishedNo guaranteed signals
LIVE FORECAST EXAMPLE60 min delayed
BTC/USD
IA1 · SCORE49.61
IA2 · Δ%-0.021%
IA3 · P(L)4.8%
recorded 2026-08-24 06:45 UTCOpen JSON ↗
78markets followed today
427k+forecasts recorded
254k+prices evaluated
3forecasting engines
See for yourself
Open a real forecast. No account required.
The public demo uses delayed data and the same format as the API. If it fits your product, a free key takes less than a minute.
Terminal · no key
curl 'https://avaltix.io/api/agent/v1/demo.php?symbol=BTC/USD'
API
One API. One documented format.
Use REST or MCP to request forecasts, compare horizons and measure their distance from observed prices. Limits, pricing and attribution terms are published upfront.
How we measure
Distance matters more than a label.
We compare the forecast price with the observed price at the expected time and inside two-hour and three-hour windows on either side. Everything is expressed as percentage distance and compared with a simple reference: leaving the initial price unchanged.
| Horizon | Sample | Error at target time | Best error ±2 h | Best error ±3 h | Versus unchanged price |
|---|---|---|---|---|---|
| 1min | 24,820 | 0.111% | 0.006% | 0.005% | 0.045 pp worse |
| 10min | 26,028 | 0.180% | 0.017% | 0.015% | 0.024 pp worse |
| 20min | 14,203 | 0.233% | 0.016% | 0.014% | 0.022 pp worse |
| 25min | 11,315 | 0.242% | 0.019% | 0.013% | 0.023 pp worse |
| 35min | 7,738 | 0.264% | 0.031% | 0.015% | 0.018 pp worse |
| 40min | 6,673 | 0.284% | 0.037% | 0.019% | 0.017 pp worse |
The evaluation measures percentage distance; it does not label each forecast as right or wrong. The earlier side of a window never starts before the forecast was created. The unchanged-price reference is published beside the model to prevent misleading percentages.
One forecast · two ways to use it
Read it yourself or put it to work.
For products and agents
Add Avaltix through a documented REST API or MCP. Every response includes the timestamp, horizon and context needed to interpret it.
Before you trust a forecast, see how close it came.
Check the percentage error, the timing margin and the comparison with the initial price. No sales call required.