For several years the Stock-to-Flow model was the most cited Bitcoin valuation framework in existence. It produced a clean chart, a simple story, and price targets specific enough to argue about. It has since diverged badly from actual price, and the reasons it failed are more instructive than the model itself ever was.
New to the model itself? Start with our Stock-to-Flow explainer, which covers how the ratio is calculated. This piece focuses on why its predictions failed and what that teaches.
What the model claimed
Stock-to-Flow borrows a concept from commodity markets. Stock is the total existing supply of an asset, flow is the amount produced each year, and the ratio between them is a measure of scarcity. Gold has a high ratio because the above-ground stock is enormous relative to annual mining output. A perishable commodity has a low ratio because most of what exists was produced recently.
Bitcoin’s supply schedule is fixed and public, so its ratio can be calculated exactly and projected forward. Every halving cuts issuance in half, which doubles the ratio overnight. The model, published in 2019, fitted a power-law relationship between that ratio and Bitcoin’s market value across its history, and the fit looked remarkably good. Extended forward, it implied Bitcoin should reach six figures after the 2020 halving and considerably higher after 2024.
Why the fit was less impressive than it looked
The core statistical problem is that both variables trend upward over time. Stock-to-Flow rises monotonically by construction, because issuance only ever falls. Bitcoin’s price also rose over the sample period. Regressing one rising series against another rising series produces a strong-looking relationship that may contain no causal information at all, a well-known trap called spurious regression. Several published critiques showed that substituting an arbitrary counter that simply increments over time produced a similar quality of fit.
The second problem is that the model has no demand term. Scarcity is a supply property. Price is set where supply meets demand. A model that predicts price purely from a supply schedule that was known to every market participant in advance is implicitly claiming markets fail to price in publicly available information over multi-year horizons, which is a strong claim to make without argument.
The third is sample size. Bitcoin has experienced four halvings. Fitting a model with strong parameters to four events and then treating its forward projections as reliable is a stretch no matter how good the in-sample fit looks.
What actually happened
Price departed from the model’s band and stayed away. The 2021 cycle peak fell well short of the trajectory, and the years since have diverged further rather than reverting. The model’s own author acknowledged the deviation. At this point the honest summary is that Stock-to-Flow described Bitcoin’s early history reasonably well and has not usefully predicted anything since it became widely known.
What the failure teaches about onchain metrics generally
The useful lesson is about what kind of metric earns trust. Stock-to-Flow projects forward from a known schedule. The metrics that have held up better measure what holders are actually doing right now, which is information the market has not fully digested by definition.
MVRV Z-Score compares market value to what holders collectively paid, so it responds to real behaviour rather than a calendar. Exchange reserves measure how much supply is genuinely positioned to sell. The Puell Multiple tracks miner revenue against its own history, which captures the economic pressure on the one group of participants who must sell to cover costs. None of these predicts a price. All of them describe a condition, which is a more modest and more defensible thing for a metric to do.
That distinction is worth carrying into every model you encounter. Ask whether it measures behaviour or extrapolates a schedule, and ask what would have to happen for it to be wrong. A model with no failure condition is not a model.
Does the halving still matter?
Yes, but through a narrower channel than Stock-to-Flow suggested. Each halving cuts miner revenue in half overnight, which forces the least efficient operations offline and changes the volume of coins miners must sell to cover electricity and debt. That is a real supply effect with a real economic mechanism, and it shows up in miner-focused metrics. What it is not is a mechanical price multiplier that can be read off a chart four years in advance.
Common questions
Is Stock-to-Flow completely worthless? As a description of Bitcoin’s scarcity relative to commodities it is a reasonable framing. As a price forecasting tool it has failed out of sample, which is the only test that matters for a forecasting tool.
Why do people still reference it? It is simple, visual, and it flatters an outcome many holders want. Models with those properties survive longer than their track records justify.
What should I use instead? No single metric replaces it, which is rather the point. A small panel of behavioural metrics read together gives a better picture than any one framework promising a number. The cycle-top indicators piece walks through a reasonable panel, and the free tools guide covers where to find them.
Leave a Reply