Why predictions need evaluating
Predictions about cryptocurrency prices are everywhere: social media threads, videos, newsletters, analyst notes and automated tools. They are cheap to make and rarely checked afterwards. A forecaster who is right gets attention, and one who is wrong is usually forgotten. That imbalance means the predictions you see are a skewed sample, and your impression of how good forecasting is can end up far too generous.
This guide gives general tools for judging a claim. It does not endorse any forecast and is not financial advice.
Is the prediction checkable?
The first test is whether the claim can be proven wrong. “Bitcoin will do well this cycle” cannot. A checkable prediction names the asset, a measurable outcome and a date: for example, a specific price level reached by a specific day. Vague wording leaves room to claim credit whatever happens. Watch for hedges such as “could,” “may” or “on its way to,” and for forecasts that quietly shift their dates after the first date passes.
A prediction without a stated confidence is also hard to judge. “Probably” means different things to different people. A forecaster who gives probabilities, such as a 30 per cent chance of an outcome, can be scored over many forecasts. One who offers certainty cannot be scored fairly, and certainty is itself a warning sign in a market this volatile.
Survivorship bias
Survivorship bias means looking only at the cases that survived a filter and forgetting the ones that did not. In forecasting it appears in several ways.
Suppose, as a made-up illustration, that 1,000 people each publicly predict whether the price will rise over the next month, purely by flipping a coin. After one month about 500 are right. After a second month about 250 have been right twice, and after a third, about 125. After a fourth month, about 62 or 63 have a perfect four-month record. Nobody here has any skill, yet dozens now have an apparent streak and a following. The audience sees the winners and not the 937 or so who dropped out.
The same effect shapes lists of “top analysts” and fund track records, since failed ones close or delete their posts. It affects tools too: a forecasting app that publishes only its successful calls looks better than it is.
Overconfidence and hindsight
People tend to be more certain than their accuracy justifies. Confident, simple stories are more persuasive than careful, uncertain ones, so overconfident forecasters often gain a larger audience. Hindsight bias adds to the problem. After a price move, the reasons seem obvious and it feels as though the move was predictable all along. Looking back, we forget how uncertain things looked at the time.
A helpful habit is to write down your own expectation before the outcome, with a rough probability, and review later. Most people find the exercise humbling.
Questions to ask of any forecast
- What is the track record, and is it complete? Is every past call shown, including the wrong ones, with dates?
- Was the record made in real time, or is it a chart fitted to the past? A model tuned on old data will naturally look good on that same data, a problem that the article on backtesting basics explains.
- How does it compare with doing nothing clever? If the forecaster cannot beat a simple baseline, the added complexity is not buying much.
- Does the forecaster explain what would change their mind?
- What is their incentive? Selling a course, subscription, token or signal service creates a reason to sound certain. Holding a position creates a reason to talk it up, a theme covered in the pieces on social media influence and pump and dump schemes.
- Are the costs included? A forecast that ignores fees and price slippage may not be usable in practice.
Correlation, causes and cherry-picking
Many forecasts rest on a chart that lines up two things, such as an indicator and price, or a past cycle and the present one. Be cautious about comparisons built from only a few cycles. With a short history and thousands of possible indicators, some will line up impressively by chance. Selecting the time window that makes the pattern look best, or dropping inconvenient periods, is cherry-picking.
Overlays that claim the present “rhymes” with an earlier period often leave out the differences: a different market structure, different participants and different macro conditions. It is also worth asking whether the indicator is a cause of the price, a result of it, or just a coincidence.
Telling analysis from hype
Careful analysis tends to look like this: it states assumptions, shows ranges and scenarios, admits what it cannot know, and separates evidence from opinion. Hype tends to rely on urgency, certainty, screenshots of wins, and calls to act now. Messages that promise guaranteed or very large returns deserve extra suspicion, and the guidance on crypto scams applies.
Using predictions sensibly
A forecast can be a prompt to think about risks, not an instruction. Because prices are volatile and no method is reliable, the more useful question is often how much you could lose if the forecast is wrong, which is the subject of risk management. Past performance does not predict future results, people can lose all they invest, and decisions should reflect your own circumstances and, where appropriate, professional advice.