Live market data

How our price predictions work

Short version: nobody can reliably predict crypto prices, and this site does not claim to. What we can do is measure how much each coin has really been moving, project that forward honestly as a range, and check how often that kind of range has been right. Every number on the prediction pages comes from the same model described here, and it is recalculated about every five minutes.

Predictions on this site are statistical estimates produced by a simple model from past prices. They are not investment advice, a recommendation to buy or sell, or a promise of future returns. Crypto assets are highly volatile and you can lose all the money you put in.

1. Data

We use daily candles and 24-hour tickers from the Binance USDT-M Futures market (the perpetual contracts such as SOLUSDT). Up to about three years of daily closes are used, and today’s candle ends at the live price. Futures prices follow spot prices closely but can differ slightly.

2. The model

We treat the logarithm of the price as a random walk with a small drift:

  • Volatility (σ). The average of two estimates of the daily size of moves: an exponentially weighted one (decay 0.94, which reacts quickly) and a plain 90-day one.
  • Trend (μ). The average of the 90-day and 365-day slopes of the log price, multiplied by 0.25 (we keep only a quarter of it, because trends in crypto are very hard to estimate) and limited to ±60% a year.
  • Median forecast for d days ahead = price × exp(μ × d).
  • Likely range = median × exp(±σ × √d). This is a one-standard-deviation band: in a well-calibrated model about 68% of outcomes land inside it.
  • Beyond one year the trend fades out (it can never contribute more than about two years’ worth), but volatility keeps growing with √d. That is why the 2028–2030 ranges are enormous: it is the model’s way of saying it does not know.

3. Price targets

The probability of reaching a target price is the chance, under the same model, that the price on the target date is at or above that level. It is a statement about the end date, not about touching the level earlier.

4. Technical reading

The Bullish / Neutral / Bearish label counts four simple facts: price above the 20-day, 50-day and 200-day averages, and the 14-day RSI above 55 (or below 45). It describes the recent past and is not an input to the forecast.

5. Back-testing

For every coin we replay the model on its own history: every 7 days we pretend it is that day, fit the model using only older data, forecast 30 days ahead, and compare with what really happened. We publish how often the real price fell inside the range and how often the median pointed the right direction. If the model is no better than a coin flip at direction, the page says so.

6. What the model cannot do

  • It cannot see news, regulation, hacks, listings or changes in mood.
  • Crypto returns have fatter tails than this model assumes, so extreme moves happen more often than the range suggests.
  • Newer coins have little history, which makes every estimate shakier.

Use the forecasts as a way to think about how big a move is plausible, not as a signal to buy or sell. For alerts when a real move happens, see the Breakout Alarm app.

Questions or corrections? Contact us.