Crypto
The Crypto page brings together crypto price predictions from Polymarket and Polywhaler’s Bitcoin forecast. Browse live probabilities, USD price targets and resolution windows, then explore the Bitcoin model’s forecast and the evidence behind it.
Crypto price markets#
The Markets tab is public. Browse price predictions for Bitcoin, Ethereum, Solana and other assets, covering five-minute, fifteen-minute, hourly, four-hour, daily, weekly, monthly, yearly and price-target markets when available.
- Filters: choose an asset, market duration and prediction type. Search by asset or price level and sort by 24-hour volume or soonest expiry.
- Prediction types: Up / down compares the price at the start and end of a window; Above / below asks about a terminal level; Price range asks where the price will end; Hit price asks whether a target will be touched before the deadline.
- Probability chart: select an event and a price level or outcome. The chart shows that outcome’s implied probability over time, with chart-history controls at the bottom-left. A market’s duration and the chart’s history window are separate controls.
- Spot price: the asset’s current USD quote and 24-hour change, sourced from CoinGecko. This is separate from prediction shares, whose prices range from 0 to 100 cents.
- Market details: inspect outcome probabilities, USD target levels, resolution dates, trading volume and liquidity. Open the existing event or chart page for more detail.
- Freshness: market data refreshes automatically. Current intraday markets roll over when their window ends. Delayed data is labeled; missing spot quotes do not block market browsing.
Bitcoin forecast#
Guests can read the one-month forecast. Sign in to explore every forecast horizon; the full model distribution, supporting markets and insights require Pro.
- Predicted price: the model's median forecast for the selected horizon, with the expected move from spot next to it. The horizon tabs show the real date each forecast targets.
- Confidence: a 0 to 100 score. It rises when many independent, liquid markets agree with each other and falls when evidence is thin or contradictory.
- Bear / Base / Bull: the 10th percentile, median, and 90th percentile of the fitted distribution. A four-in-five chance the price lands between Bear and Bull.
- Forecast chart: Bitcoin's price history flowing into the forecast path, wrapped in 50, 80, and 95 percent probability bands that widen with time.
- Probability distribution (Pro): the implied density of future prices for the selected horizon, plus the chance Bitcoin ends above key round levels.
- Supporting markets (Pro): every market feeding the model, with its platform, odds, strike, liquidity, volume, and the exact weight the model gave it. Sort and filter freely.
- Model insights (Pro): what changed in the last 24 hours, which markets drive the forecast, whether the markets lean bullish or bearish, and where Polymarket and Kalshi disagree.
How the forecast is built#
Prediction markets already price Bitcoin levels every day: markets like "Bitcoin above $120,000 on August 15?" trade at a probability, and that probability says how likely the market thinks that price is by that date. The forecast reads every one of those prices together and works out the range of prices they imply, horizon by horizon.
What goes in#
- Every open Bitcoin price market on Polymarket and Kalshi, read again every hour: above-or-below markets for a date, price-range markets, year-end markets and "will Bitcoin hit X" markets. Markets with no order book and no recent trades are left out.
- Each market is weighted by how much it can be trusted. Deep, actively traded markets with a tight spread and recent trades move the forecast; a thin or stale one barely nudges it.
- Prices are corrected for the known habit of prediction markets to overprice long shots, and prices sitting right next to 0% or 100% are set aside, because at that point a quote only says "very unlikely", not how unlikely.
- "Will Bitcoin hit X" markets ask whether the price touches a level at any point, not where it ends, so they are converted to an ending-price view and count for less.
- The same level quoted on both platforms shares credit instead of being counted twice, and prices that contradict each other are marked down and flagged. As markets resolve, a platform whose prices have proved more accurate earns a little more weight.
What comes out#
- For each horizon (one week, one month, three months, six months and one year), a single smooth price curve is fitted through the weighted markets. Its middle is the predicted price, and its spread gives the Bear and Bull cases and the bands on the chart.
- A horizon with no markets expiring near its date is read off the horizons around it, marked Interpolated horizon, and given a lower confidence.
- A confidence score from 0 to 100 for each horizon, and one for the forecast overall. The next section explains how to read it.
Limits#
- The forecast reports what the markets collectively imply. That is real information with real money behind it, but it is not a guarantee, and it is not financial advice.
- The curve has one standard shape, so it smooths over fat tails that markets sometimes price. A level priced far above what that shape allows still widens the range a little.
- The conversion of "hit" markets assumes Bitcoin has no strong drift up or down over the horizon.
- A thin market can move a horizon with few other markets until liquidity arrives, which is part of why confidence drops when evidence is thin.
The full specification is in the Developer portal for Quant subscribers.
Reading the confidence score#
Confidence is not conviction about direction; it is a measure of evidence quality. A confidence of 80 with a flat forecast means the markets agree Bitcoin is going roughly nowhere. A confidence of 40 with a bullish forecast means the upside view rests on thin evidence. The score moves when markets are added, expire, or start disagreeing.
Model versions#
Every forecast carries the version of the model that produced it, so a change in the model is visible in the data.
- btc-forecast/1.1.0: fit and weighting corrections. Near-certain prices are left out of the fit instead of being pinned near the edge. Prices near 0% or 100%, whose exact level is mostly noise, count for less than prices near even odds. Markets expiring on different days within a horizon are lined up to a common date before fitting. Spreads are measured the same way whichever outcome is called Yes, and each platform's track record is judged more fairly at every price level. A new lean insight, the chance Bitcoin ends above today's price, replaces the skew insight, and the 24-hour change now compares with the run from a day earlier at the same horizon. On the markets used to check the change, the three-month forecast came out much narrower and much closer to spot. When backtesting across the change, keep runs from each version apart.
- btc-forecast/1.0.0: initial release.