How the index works
Three rules: publish before the outcome, grade in public, and only show what has evidence behind it.
Why "Brier"
The Brier score is the standard way to grade probability forecasts: it punishes being confident and wrong. The name is a promise. Every forecast on this site gets a grade, and the grades are public.
What goes into the index
- Trend gate. GREEN when BTC is up over the last 28 days, RED otherwise. It passed an out-of-sample test on historical data.
- Expected range. Percentiles of the BTC close at 1, 7 and 30 days from a volatility model (EWMA). The median is today's price by design. It beat a simpler baseline, but only slightly, and the live record is still short.
A new reading joins the index only after its test passes. Ideas still being tested can appear in articles, labelled as such, but never inside the index.
The Brier score
Each day the index also publishes chances that Bitcoin closes above four price levels in 7 and 30 days. When the day arrives, each chance is graded: the Brier score is the average of (chance − outcome)², where the outcome is 1 if the price closed above and 0 if not. 0 is perfect; always saying 50% scores 0.25.
A low score alone can flatter. So the front page shows the skill score, which compares ours with a reader who only knows history: how often Bitcoin has ended above that level over the same number of days, counting only the days before each forecast. 0 means no better than history, 1 means perfect, below 0 means worse. The 1-day chances are left out of it: they are almost always near 0% or 100%. Chances are never shown as 0% or 100%; anything beyond 1% or 99% is shown as "<1%" or ">99%".
What is public and what is not
Public: what kind of model it is, every forecast it makes, the exact rule each one is graded by, and every grade, misses included. Not public: the model's settings and our research on alternatives. Anyone can copy a formula; nobody can copy a record that was published before the outcome and graded after it.
When a forecast counts
Prices are the Binance BTC/USDT daily close at 00:00 UTC. A forecast dated a given day uses that day's 00:00 UTC close as its starting price, is recorded the same day, and is graded on the 00:00 UTC close of its expiry date. Since October 22, 2026 each day's forecasts are also stamped into the Bitcoin blockchain when they are made, so you can check they came first: see the ledger.
What it does not do
It does not tell you whether BTC will go up or down. We tested seven return-prediction signals and none held up. Expected returns in crypto are not measurable with the data available; risk is.
Money
The site is paid for by donations. Everything on it, including the full forecast record, is free. If it ever carries a sponsor, the sponsor is labeled "Sponsored", never appears inside the index or the scorecard, and is never an exchange, prediction market or token that the index forecasts or grades. There are no affiliate links: nobody pays us when you trade.
Privacy
We count visits with Cloudflare Web Analytics: no cookies, no fingerprinting, no tracking across sites, and we never see or store your IP address. What you enter in "Your call" and "Practice" stays in your own browser and is never sent to us. Details: privacy and cookies.
Corrections
Errors are corrected with a dated note. Nothing is deleted, and forecasts are never edited after publication.
About the author
Andres Gaya writes about crypto markets, on-chain investigations and forecasting. The author may hold crypto assets, including BTC.