How this site works

Where the numbers come from

Market data (prices, market capitalizations, trading volumes, supply figures) is synced from CoinGecko, supplemented by DeFiLlama and the Alternative.me Fear & Greed index, into our own database on a continuous schedule. Technical indicators (RSI, moving averages, volatility, support and resistance levels, correlations) are computed by us, deterministically, from that accumulated history.

A hard production rule follows from this: every figure displayed on this site is injected directly from the database. Numbers are never typed by an author and never produced by a language model. An automated validator rejects any draft page section containing a figure that cannot be traced to the underlying data record.

Editorial standards

• Guides, explainers, reviews, and written analysis are produced to a documented house style and reviewed against quality gates before publication; drafts that fail them are rejected.
• Pages that do not meet minimum data-quality thresholds for an asset are not published at all.
• Nothing on this site is investment advice, and pages never tell you to buy or sell anything; see the disclaimer.
• Where a page is sponsored, produced with a partner, or contains links that can earn us a commission, that is disclosed on the page. Market data, indicators, and model outputs are never adjusted for any commercial relationship.

How the price predictions are made

Price predictions are one part of the site, and the part most easily done badly, so they get the strictest process. Ours are produced by a deterministic statistical model. The current version, vb-momentum-1.1.0, is a volatility-banded momentum projection: an asset's recent return volatility sets the width of the forecast range, and its damped, capped longer-run momentum sets the drift of the central estimate. The same inputs always yield the same forecast. No human judgment, no editorial thumb on the scale, and no language model anywhere in the number-producing path.

Concretely: the central estimate projects the latest price forward at a damped annualised drift derived from the trailing six months of daily returns (clamped so a hot streak can never extrapolate without limit, and decaying with horizon, since recent momentum informs the months ahead far more than the years ahead), and the range around it spans roughly one standard deviation of log-returns scaled to the horizon, meaning reality is expected to land inside the range about two-thirds of the time, and the Accuracy Ledger exists to check whether it actually does. Stablecoins and pegged assets get no forecast, and neither do assets with too little price history or insufficient market size. When the method changes, the version number changes; old forecasts remain attributed, and scored, under the version that produced them.

The Accuracy Ledger

Every forecast is written to an append-only record at the moment of publication: asset, horizon, forecast range, model version, timestamp. Forecasts are never edited, regenerated, or backfilled. When a forecast's horizon matures, it is scored automatically against the actual market price, and the result, hit or miss, is published on the Accuracy Ledger. Our misses stay on the record alongside our hits.

Corrections

If you find an error (a stale figure, a broken source, a claim that does not hold up), email us via the contact page. Corrections are applied to the live page; the forecast record itself is never altered.