Key takeaways
- Forecast accuracy is almost never tracked, so confident predictors face no cost for being wrong.
- Price already reflects widely known information, which is why publicly available reasoning rarely gives an edge.
- Precise targets and dates signal confidence, not knowledge. Vagueness is often more honest.
- Read constraints instead: issuance, miner economics, liquidity depth and leverage describe the system rather than guessing its next move.
Search for where any major cryptocurrency is heading and you will find no shortage of answers. Specific figures, specific months, delivered with conviction. What you will almost never find alongside them is a record of how the same source’s previous forecasts turned out.
That absence is the whole story. This piece explains why forecasting an asset price is genuinely difficult, why the prediction industry persists regardless, and what is worth reading instead.
Why prediction is hard in principle
Price already contains the public argument
A market price is the outcome of many participants acting on what they know. When information becomes widely available, it tends to be reflected in price quickly. That is why reasoning built entirely from public information rarely produces an advantage: everyone reading it can also act on it, and their acting is what moves the price.
Crypto markets are less efficient than large established markets in some respects, which is why some participants do find edges. But those edges tend to come from speed, infrastructure, or information that is not in a published article. If a compelling argument is in a free article, its implications are usually not still available.
The dominant drivers are unobservable
Large price moves frequently originate in flows nobody outside the participants can see. Portfolio rebalancing, forced liquidation of leveraged positions, treasury operations, changes in access to banking, or a single large holder deciding to reduce exposure. None of these are visible in advance, and no chart contains them.
Reflexivity
Financial markets are not passive systems being measured. Forecasts influence behaviour, behaviour influences price, and price influences the next round of forecasts. A widely shared expectation can partly produce its own outcome for a while, and then reverse violently when positioning becomes crowded. This makes the target of prediction unstable in a way that physical systems are not.
Regime changes
Relationships that held for a period stop holding. Correlations with other asset classes appear and vanish. A model fitted to one regime will look excellent until conditions change, at which point it fails. Because regime changes are recognisable only afterwards, backtested performance systematically overstates what a method will deliver going forward.
Why bad forecasting survives
The forecasting economy has an accountability gap. Predictions are cheap to produce, emotionally satisfying to consume, and almost never scored. A prediction that fails is quietly replaced by a new one, and nobody maintains the ledger.
Selection effects reinforce this. Given enough forecasters making enough varied calls, some will be right about a significant move purely by chance. Those few will be prominently promoted afterwards, creating the appearance that accurate prediction is achievable and that its practitioners can be identified in advance.
Distribution matters too. Bold, specific claims travel further than careful ones. An article that says a range of outcomes is plausible and the uncertainty is wide is a worse product by every engagement metric, even when it is a more accurate description of reality.
And some predictions are not attempts at accuracy at all. Anyone holding a position benefits from others sharing their view. That does not make every bullish or bearish commentator dishonest, but it does mean the incentive to publish is not always the incentive to be correct.
Reading forecasts more usefully
If you are going to read them, a few checks help. Is there a stated time horizon? A prediction without one cannot be wrong. Is there a range and a probability, or a single number? Single numbers signal confidence rather than knowledge. Are the conditions for being wrong stated? Genuine analysis specifies what would falsify it. Is there a public record of previous calls, including the failures? And does the author disclose a position?
Most forecasting content fails several of these. That is not a reason to be cynical about everyone, but it is a reason to weight accordingly.
What to read instead
The alternative to prediction is not passivity. It is studying the constraints and mechanics of the system, which are observable in a way that future prices are not.
Issuance and supply mechanics
How much new supply arrives, on what schedule, and under what rules. These are defined by protocol rather than by opinion, which makes them among the few genuinely knowable facts in the space.
Miner economics
Mining is where new supply meets real costs. Operators pay for electricity and hardware in ordinary currency and receive coin, so a portion of issuance tends to reach the market as an operating necessity. When difficulty rises without a matching rise in revenue, margins compress and higher-cost capacity comes under strain. This is a structural pressure worth understanding, and it is visible in advance rather than inferred afterwards. Current network figures are on our mining dashboard, and the profitability calculator shows how sharply outcomes swing with power price.
Liquidity and leverage
How deep the order books are and how much leverage is outstanding tells you how violently the market can move for a given amount of flow. It does not tell you direction. It does tell you about the potential magnitude of moves, which is arguably more actionable for risk purposes than a direction guess.
Rules and access
Changes to what institutions may hold, how venues must operate, and how participants access the market change the set of possible behaviours. These are slower moving and better documented than sentiment.
Protocol changes
Alterations to consensus, issuance or fee mechanics change the system itself. They are announced in advance and debated publicly, which makes them among the more analysable events available. Background on the underlying mechanisms is in blockchain technology and learn.
The honest position
Nobody at this site knows where any price is going, and we do not publish targets. What can be done responsibly is describe the network’s condition, explain the mechanics that shape supply and cost, and be clear about the boundary between measurement and inference. That boundary is set out in our methodology.
Uncertainty is not a failure of analysis. It is an accurate description of the situation, and content that hides it is selling comfort rather than information. Nothing here is financial advice, and no outcome described anywhere on this site should be treated as assured. For per-asset data see coins and markets.
Frequently asked questions
If predictions are so unreliable, why are there so many of them?
Because demand for them is enormous and the cost of being wrong is close to zero. Uncertainty is uncomfortable, and a confident number relieves that discomfort more effectively than an honest description of a probability distribution. Meanwhile almost nobody keeps score. Predictions are published, forgotten, and replaced by fresh ones, so a forecaster's reputation depends on how compelling they sound rather than on any measured track record. Add attention-driven distribution, where bold claims travel further than careful ones, and the incentives point firmly toward more prediction and more confidence.
Is technical analysis useless then?
Not useless, but frequently over-claimed. Some of what it studies is real: liquidity clusters near obvious levels, and many participants watch the same charts, so behaviour around those levels can become partly self-fulfilling in the short term. What it cannot do is deliver reliable forecasts from price history alone, and its flexibility is a problem. Enough indicators and timeframes exist that a supporting pattern can be found for nearly any conclusion after the fact. It is more defensible as a framework for managing risk and defining exits than as a predictive engine.
Does on-chain or mining data predict price better?
It does not predict price, and anyone presenting it that way is overselling. What it does is describe the system's state, which narrative coverage generally omits. Difficulty and hashrate indicate how much capacity competes for issuance and how compressed margins have become. Issuance schedules define the rate of new supply. Miner economics explain a persistent source of structural selling. These are useful for understanding conditions and constraints. Converting them into a target price reintroduces exactly the false precision that makes forecasts unreliable in the first place.
What would a genuinely honest forecast look like?
It would give a range rather than a point, attach explicit probabilities, state a time horizon, name the conditions under which it would be considered wrong, and be published somewhere it can be scored later against a public record of the forecaster's previous calls. It would also acknowledge that most of the variance comes from factors nobody can observe in advance. Such forecasts exist but they are rare, partly because they are far less satisfying to read than a specific number, and partly because they expose the forecaster to being measured.
Crypto writer at Cryptocurrency Miners.
