How to read this dashboard
A plain-English guide — no finance background needed. Every section below matches something you see on the main page.
The one thing to understand first
Nobody can actually predict Bitcoin's price. If they could, they'd quietly get rich instead of running a website. So this is not a crystal ball, and you should never bet money based on it. What it really is: a computer program that makes an honest best guess, explains its reasoning, shows how unsure it is, and — most importantly — keeps score of how right it's been over time. The value here is the transparency and the track record, not a promise.
What the site does, in one sentence
Every hour it reads the news and the mood of the crypto market, then publishes a best guess for where Bitcoin's price might be in 1 week, 1 month, and 1 year — always shown as a range with odds, never a single magic number.
The one-week call & the three horizons
The lead forecast leads with a band, not a number — that's deliberate. The shaded orange bar is the most-likely range; the thin tick inside it is just the single most-likely point. Below, the same thing for 1 week / 1 month / 1 year:
- most-likely range (e.g. $61,872 – $69,258)
- Where the price will probably end up. A wider band means more uncertainty, and the band always widens for longer time-frames — the further out you look, the blurrier the future.
- central guess (small, below the range)
- The single most-likely price — shown small and quiet on purpose, because leading with one number would imply a false precision the model doesn't have. It is not a target or a promise.
- P(up) — e.g. 49%
- "Probability up": the estimated chance the price is higher than today by that date. Around 50% is labelled a coin-flip — the honest answer for Bitcoin most of the time. Muted green for up, muted red for down (never bright — a colour is never a buy signal here).
- resolves in …
- How long until this forecast comes due and gets graded against what actually happened.
“What the model is watching” — the signals
These cells are the real-world things the program reads before guessing — the "world events" in the name. A small stale tag appears if a feed hasn't updated recently, so a frozen signal never quietly looks live.
- Fear & Greed (e.g. 11/100)
- A crypto mood gauge from 0 ("Extreme Fear") to 100 ("Extreme Greed"). A very low number means the market is fearful — which, historically, has sometimes been a buying opportunity.
- Perp funding (e.g. +0.0081%)
- A small fee traders using borrowed money pay each few hours. Positive means more people are betting up — a "crowded" market that can snap back.
- Open interest & implied volatility
- How much money is riding on leveraged bets, and how much price-swing the options market is pricing in. Both gauge how jumpy things are.
- News tone
- Whether global news coverage of Bitcoin right now is, on balance, positive or negative.
“Forecast vs. reality” — the chart
The orange line + shaded band is our forecast over time; the blue line is what actually happened. The dotted line is the "random walk" — the naive "no change" benchmark we grade ourselves against, drawn right in the chart so you can see how close to a coin-flip short-horizon Bitcoin really is.
“Why it moved” — the AI nudge
The forecast starts as plain math (a GJR-GARCH volatility model). Then Anthropic's Claude reads the signals and writes a short explanation of why it nudged the guess and widened or tightened the range. It is only allowed small, hard-capped adjustments — it can't swing the forecast wildly, which keeps it honest. A quiet "Claude overlay" chip means the AI was involved this hour; "baseline only" means you're seeing the plain math.
“The public record” — the honesty centerpiece
The most important part: it grades the site's own past guesses, so you can judge whether it's any good instead of just trusting it.
- “Random walk”
- The simplest possible "prediction": assume the price stays exactly where it is. It sounds dumb but is famously hard to beat — so every forecast is graded against it. If we can't beat "no change," we say so.
- The accuracy ledger (N · Brier · MAPE · Coverage)
- N = how many forecasts have come due; Brier = the up/down score (lower better); MAPE = the average % the price guess was off; Coverage = how often reality landed inside the band vs how often it should have. The little bar shows skill vs the random walk — centered = a tie.
- The reliability diagram (“the receipts”)
- When we say 60%, does it happen ~60% of the time? Each dot plots what we said against what actually occurred; dots on the diagonal line mean honest probabilities.
- The PIT histogram
- A check that the ranges are the right width. Flat bars = well-calibrated; lopsided bars mean the bands are systematically too wide or too narrow.
- “Beats the random walk?”
- A proper statistical test (Diebold-Mariano) that defaults to "no significant difference" and only claims skill once the evidence clears it. No overclaiming.
- The model vs. the crowd
- We line up our odds against a real-money prediction market (Polymarket) on the same question, and — once those markets resolve — keep score of who's actually been closer.
What "good" looks like here: for short horizons, roughly tying the random walk is the realistic, credible result — and we publish it openly rather than hiding it. A site that claims to crush the market every week is lying.
Should I trade on this?
No. This is an educational and portfolio project, not financial advice and not from a licensed adviser. The forecasts are model-implied probabilities and ranges, not guarantees, and Bitcoin can lose value fast. If you're making real financial decisions, talk to a licensed professional. Read the full disclaimer & method →