What a random walk means
A random walk is a path built from steps whose direction and size cannot be predicted from earlier steps. Picture someone flipping a coin each second and stepping one pace forward on heads and one pace back on tails. After any number of flips, the best guess for the next step is zero, because nothing in the past tells you which side will land.
Applied to prices, the random walk hypothesis says that the change in price from one period to the next is unpredictable from past prices. The best guess for tomorrow is then today’s price, plus or minus an amount that cannot be anticipated. This is a claim about predictability, not a claim that prices are meaningless or that nothing causes them to move.
Where the idea comes from
The earliest mathematical treatment of price movements as random is usually credited to Louis Bachelier, who wrote a doctoral thesis on the topic in 1900. In the 1960s and 1970s, economists including Eugene Fama developed the efficient market hypothesis, which argues that prices already reflect available information, so new changes arise mainly from news that cannot be foreseen. If that is right, price changes should look much like a random walk.
The two ideas are related but not identical. A random walk describes the statistical pattern. Market efficiency is an explanation of why that pattern might occur. Efficiency can also hold in a weaker or stronger form, depending on what counts as available information.
A coin-flip illustration
Here is a small example, using a made-up price that starts at 100. Suppose each day it rises by 1 or falls by 1, decided by a fair coin. After ten days the price must be an even number of steps away from the start, somewhere between 90 and 110. A run of six heads in a row would take it to 106, and on a chart that would look like a strong uptrend.
Nothing caused that trend except luck. Runs of this kind are common in random sequences, which is part of why people find patterns in charts that have no predictive value. It also explains why a trader looking at a chart and a simulated coin-flip series can struggle to tell them apart. The same caution applies when reading chart patterns or trend momentum into a short history.
Arguments that crypto prices are close to a random walk
Several observations fit the hypothesis.
- Day-to-day changes in major coins often show little correlation with the previous day’s change, which is what a random walk implies.
- Many apparent trading rules stop working once costs and out-of-sample testing are added.
- Markets are open to many competing participants, including automated traders, who can act quickly when they see a pattern, so any easy pattern tends to vanish.
- New information, such as a regulatory announcement, arrives unpredictably, and prices jump when it does.
Some academic studies on the largest coins have found behaviour consistent with weak-form efficiency, particularly in recent years and for large, liquid coins. The conclusions vary with the period, the coin and the statistical test used.
Arguments against
Other research finds departures from a pure random walk, especially in the early years of Bitcoin and for smaller coins.
- Smaller and less liquid markets have wider spreads and fewer arbitrageurs, so mispricings may last longer.
- Returns often show momentum over some horizons or reversals over others, and the size of price swings clusters in time rather than being constant.
- Fat tails are common: extreme moves occur more often than a simple bell-curve random walk would suggest.
- Behavioural explanations point out that traders are influenced by attention, herd behaviour and emotion, as described in the pieces on FUD and FOMO and meme coins.
Findings of this kind do not necessarily mean profits are available. A statistical pattern can exist and still be too small to survive transaction costs, as covered in articles on exchange pricing and the bid-ask spread.
Refinements worth knowing
A real market need not be a perfect random walk to be hard to forecast. The direction of changes may be nearly unpredictable while the size of changes is partly predictable. This is why volatility models can be useful even if directional forecasts are not. Some researchers also describe markets as becoming more or less efficient over time, as participants and tools change. A pattern that worked in one period can disappear when enough people trade on it.
Why this matters to readers
If prices are close to a random walk, then claims of reliable prediction should be treated with scepticism. A long run of correct calls by one person is not strong evidence of skill, since among many forecasters, some will have good streaks by chance. The point is developed in the guide to evaluating predictions.
None of this settles the matter. The evidence is mixed, the data history is short, and results depend on method. What can be said fairly is that cryptocurrency prices are volatile, that simple rules rarely deliver a dependable edge, and that past performance does not predict future results. People can lose all they invest, and nothing here is financial advice.