Methodology
How DEKIV turns free public data into an honest, one-minute read.
Honest freshness
The pulse animates only on genuinely live data. Prices and detected facts are polled continuously; our interpretation (the “read”) is updated roughly every two hours and always shown with its real timestamp. We never label data fresher than it is — prices live ~1–2 min · facts ~1 min · our read ~2h.
Probabilities, never advice
DEKIV never tells you to buy or sell. Every forward statement is a probability tied to a concrete price level or range — e.g. “60% holds above $95k or rises.” The percentage is confidence in a scenario, not a prediction.
No jargon
We never show bare “bullish / bearish / neutral.” Direction is always plain language (“more likely up / sideways / down”) anchored to a level. This is enforced as a hard rule inside the analysis itself and validated before anything is shown.
How the read is made
A tiered AI model produces the flagship market read and the per-coin reads, fed only with current prices and detected facts. Output is capped and checked; if it fails our rules, nothing is published and the previous dated read stands.
Data sources
Prices and volume from public exchange / CoinGecko data, market sentiment from the Fear & Greed index, stablecoin supply from DefiLlama, delayed macro proxies, and crypto news headlines. All free, public sources — a modest budget does not buy an institutional feed, and we never pretend otherwise.
Grading
Every read is recorded and graded 30 days forward against the real price. Those outcomes build the calibration and accuracy numbers on the track record — no backfilling.