Why We Forecast Crypto in Years, Not Weeks

Ask where Bitcoin will be next week and nobody can answer with evidence. Ask where it is likely to be in three years and there is something to work with: a cycle we have now seen four times, and a price history long enough to test a model against. That difference is why every forecast on StakeBible starts 90 days out and runs, quarter by quarter, to about ten years.

Here is what changes between a short-term and a long-term crypto forecast, and what the long one is for.

Can you predict crypto short term?

Not with anything we can measure. Day to day, crypto returns are close to random: yesterday's move tells you very little about today's. Monthly patterns appear, but they rarely survive the next cycle.

We tested one of the most repeated: that September is a bad month for Bitcoin. On average it is. It was negative in 9 of the 15 years in our data, and the effect survives a control for each year's place in the cycle. But when we fitted a month-by-month pattern on the earlier cycles and checked it against the latest one, it got the direction right in 6 months out of 12. A coin toss. A pattern that does not carry into the next cycle is a fit, not a signal, so it stays out of the model.

Inside a single year, the noise is large. Between 2012 and 2025, Bitcoin's yearly high was never less than about 1.6 times its low, and in the median year it was about three times. Every accumulation phase so far has had drops of 30–40% along the way. Nobody has shown how to tell which week they arrive.

What the model actually reads

Our Bitcoin model reads a cycle, not a chart pattern. The first part is a support line that follows a power law of time since Bitcoin's first block: it keeps rising, but more slowly each year. The second is the cycle between two halvings, whose swing above that line has shrunk each time: the peak stood roughly 20 times above support in 2013, 13 times in 2017, 6 times in 2021 and about 2.5 times in 2025. The cycle lows are the steadiest part. In each of the last three cycles, the low landed at almost the same point of the cycle, within a few percent of its length.

The other coins hang off that clock. Ethereum is Bitcoin times the ETH/BTC ratio, which tends to pull back towards its long-run average. For a few large coins, a steady ratio to Bitcoin forecast their own past better than the altcoin drift did, so they keep today's ratio. For a few others, today's ratio lowered at a declared pace, the average at which the largest coins of past cycles lost ground to Bitcoin, did better than the altcoin drift, so that is their forecast. Every other coin is Bitcoin times its ratio to Bitcoin today, pushed by a drift measured across the whole altcoin universe.

All of these are statistical models run on our own price history. No price, return or band on StakeBible is written by a language model. Language models do research: they score the projects they cover from 0 to 1000, and a separate search can read the estimated date of the next halving (without it, the model uses the theoretical four-year cycle). The score can keep a coin out of the forecasts and can cap how confident we are. It never moves a price.

A four-year cycle says a lot about which quarter the price is likely to be in, and almost nothing about which day. So the grid is quarterly and the shortest point is 90 days. Even there, the forecast starts from today's price and lets its gap to the cycle's shape fade over months: at 90 days, today's price still weighs heavily; two years out, the cycle is nearly everything. The Bitcoin price prediction page lays each year out in quarters and marks where the year's expected low and high most likely fall. That is as fine as the model's clock gets.

Longer horizons, wider bands

A long-term forecast is not one number. Each point comes with a band, a low and a high, built for a stated level: 85% for horizons up to three years, then falling in a straight line to 65% at ten years. For the same level, a band never gets narrower as the horizon gets longer. We cannot be surer about 2033 than about 2028.

Those levels are targets we chose, not results we measured. For Bitcoin, the band is then sized from how the model did on data it had not seen: past cycles left out one at a time, and forecasts re-run at past dates with only what was known then. We take the wider of the two. No other coin's band can be narrower than Bitcoin's at the same point.

Four Bitcoin cycles are four observations, not five thousand days, because the days inside a cycle move together. Forecasts re-run at past dates give a real test at short horizons, but the further out you look, the fewer independent windows there are, and beyond about three years there are almost none. There, the band leans on how different past cycles were from one another, and the 65% at ten years is a claim no test has confirmed yet. That is why the level steps down: at ten years we claim less, and a narrow ten-year range would be a promise with nothing behind it.

The confidence we publish starts from the project's research score and can only go down from there. It never rises above the band's level, never rises above Bitcoin's confidence at the same point, and once enough forecasts have been graded, the accuracy tracker can lower it if the bands miss more often than they claim. As of September 2026, no forecast from the current models has reached its horizon yet. Until the tracker has graded enough of them, we publish no accuracy figure; when one appears, it will come from the tracker. The full rules are in our methodology.

Why most long-term altcoin forecasts sit below today's price

Look at a far year in All Coins: most altcoins are forecast below their current price. That is neither a glitch nor a verdict on each project. It is the base rate.

We measured every coin in our database priced at both Bitcoin peaks, 2021 and 2025. Only 3.5% of those 368 altcoins gained ground on Bitcoin, and the median lost 95% of its value measured in bitcoin. The sample only holds coins that still trade, so the real figure is worse. The full measurement is in Why Most Altcoins Lose to Bitcoin.

Our altcoin model carries this pattern as its universe-wide drift. Over a few weeks it barely shows. Over years it adds up, and it is why the forecast points down. A short-term forecast would hide it; a long-term one has to show it.

How to use a long-term forecast

• Match the horizon to your plan. If you need the money in six months, a ten-year band is not about you.
• Read the band, not only the central figure. Plan as if the low end can happen. At an 85% level, about one point in seven is expected to land outside the band, above or below.
• Compare coins against Bitcoin, not only against the dollar. A coin that doubles while Bitcoin triples has lost ground.
• Leave the timing to a schedule. Timing is the part nobody can forecast. Dollar-cost averaging removes it by buying a fixed amount on a fixed schedule, which is why it fits a long horizon.

What it cannot do

• Time the next month. A 30–40% drop inside a phase is normal and sits below the model's resolution.
• Draw the road between two points. We publish a level for each quarter, not the path between them.
• See events with no trace in the data. Bitcoin has never traded through a credit crisis like 2008, so there is nothing to calibrate that on.
• Pick the altcoin exceptions. The base rate says most lose to Bitcoin; it cannot say which ones will not.
• Tell you what to buy. A forecast is model output with its uncertainty attached, not investment advice.

A forecast in years is not a bolder claim than one in weeks. It is a smaller one: it says only what a repeating cycle and a measured base rate can support, and shows how wide the range is.

The forecasts themselves, with their bands, are on our crypto predictions page.