What Happened to the Stock-to-Flow Model?
For a few years it was the most shared chart in Bitcoin. A single clean line running through a decade of price history, with one simple idea behind it. Scarcity drives value, and Bitcoin's scarcity is known years in advance, so the price is knowable too.
Then it stopped working, and the chart quietly disappeared from the timelines that used to post it every week. This is what actually happened, and more usefully, how to spot the next one.
What the model said
Stock-to-flow compares how much Bitcoin already exists to how much is mined each year. A high ratio means new supply is small next to the existing pile. The model turns that one ratio into a price with a short formula, price equals 0.4 times the ratio cubed.
It was published in March 2019, in an article called Modeling Bitcoin Value with Scarcity, and it was fitted to monthly Bitcoin data running from 2009 to early 2019.
Those two dates are the whole story, and almost nobody puts them next to each other. The model was released in 2019. It was fitted to data ending in 2019. Every year of that famous ten-year fit was drawn after the events it was fitting.
The first real test
A model only starts being tested the day it is published. Everything before that is homework it already had the answers to. So the honest question is what stock-to-flow did once the past ran out.
For this cycle the formula pointed at roughly $714,000. Bitcoin topped at $126,272 on 6 October 2025. That is a miss of 5.7 times, in the exact cycle the model was most famous for predicting.
Every cycle top, scored
One miss is an anecdote. Score all four cycle tops and the pattern is not what most people assume.
- November 2013. Actual $1,163, model $291. The model ran 0.3x the top. Inside the fitted window.
- December 2017. Actual $19,666, model $6,565. Again 0.3x. Inside the fitted window.
- November 2021. Actual $69,000, model $75,796. 1.1x, nearly exact. Live.
- October 2025. Actual $126,272, model $714,416. 5.7x. Live.
Look at the 2021 row for a moment. A ratio of 1.1, almost perfect, at precisely the moment the model was most widely believed and most widely shared. That was its single best call, and it was the last one before it broke. Peak accuracy and peak confidence arrived together, which is usually when a model is at its most dangerous.
Why nobody noticed sooner
Here is the part that makes this case worth studying. Split the model's life into the period it was fitted to and the period it had to survive on its own, and measure two different things: how well it tracks, and how far off it is.
- 2009 to March 2019, fitted. Fit to price 0.93. Bitcoin traded at 0.56x the model.
- March 2019 to end 2021, live. Fit 0.75. Bitcoin at 0.68x.
- 2022 onward, live. Fit 0.86. Bitcoin at 0.17x.
Read those two columns against each other. The fit barely moved, sitting at 0.86 even through the years the model was failing. But the level fell off a cliff, from 0.56x down to 0.17x. Bitcoin trades roughly 91% below the model today.
That is why defenders of a failing model can point at a chart and honestly say it still tracks. Correlation measures whether two lines move together, not whether either one is right. On a log scale, across fifteen years and six orders of magnitude, almost anything that generally rises will look like it tracks Bitcoin. The eye reads shape. The money cares about level.
The fix is simple and almost nobody does it. Stop looking at the two lines together. Plot the ratio between them instead, actual divided by predicted, and watch whether that stays near 1. A model that is working produces a flat, boring line. Stock-to-flow's ratio has been falling for four straight years.
Why it broke
Stock-to-flow measures supply and nothing else. Demand, liquidity, interest rates and who is actually buying are all invisible to it.
That is a fatal blind spot for a specific reason. Bitcoin's supply schedule is fully known years in advance and never surprises anybody. So the only thing that could ever surprise this model is the half of the market it refuses to look at.
And that half changed completely. In the early years Bitcoin was thin and retail-driven, and scarcity genuinely was the story. By this cycle, ETFs and institutions were pricing Bitcoin against global liquidity, rates and portfolio allocation math that has nothing to do with how many coins were mined last year. The model kept producing the same output as if the buyer had never changed.
Worth being fair here. Stock-to-flow is not a case of classic overfitting. It has very few moving parts, which normally counts in a model's favour. It failed because it was one-sided, not because it was stuffed with parameters.
The four checks that would have caught it
None of what follows needs hindsight. All four were answerable in 2019 using nothing but the model's own publication date and its own description of itself.
1. Was the track record earned or drawn? Find the publication date, then ignore everything before it. Any record inside the fitting window proves nothing, because the model was built knowing those answers. Stock-to-flow fails badly. Of the sixteen years people cite, only the stretch after March 2019 was ever a real test, and the market broke it inside three years.
2. What does it deliberately ignore? Every model simplifies. The question is whether the thing it drops can move the outcome on its own. List every force that moves the price, then cross off the ones the model actually looks at. Whatever is left is your blind spot. For stock-to-flow, that is the entire demand side.
3. What result would prove it wrong? A real model names the outcome that would kill it before that outcome arrives. If every result can be absorbed as early, late, or a temporary deviation, it is not a model, it is a belief with a chart attached. As price fell further below the line, the gap was explained rather than counted.
4. Does it work anywhere it was not fitted? Another asset, an earlier era, a different regime. A relationship that is real usually shows up somewhere else. The cross-asset argument for scarcity exists and is where much of the debate has focused, but the Bitcoin version never worked out of sample anywhere else.
Three clear failures and one weak pass, all visible before a single dollar was at risk.
What this does not say
Failing these checks means a claim is unproven, not false. Plenty of unproven ideas turn out to be right. The checks tell you how much to risk on one, not whether to believe it.
It also does not settle the stock-to-flow argument. Its author has responded to criticism over the years and the debate continues in public. What is not in dispute is the arithmetic. The formula pointed at one number and the market delivered another five times smaller.
And it does not mean you should ignore models. The alternative to a flawed model is not no model, it is an unexamined gut feeling, which fails all four checks by definition.
The part that generalises
Beliefs rarely die because they were wrong from the start. They die because the conditions that made them true quietly changed, and nobody rechecked. Stock-to-flow had ten years of evidence behind it, which is exactly what made it so hard to question.
If you take one habit from this, make it this one. Whenever a chart makes you feel certain, go and find its publication date. That single question does more work than any other, and it takes thirty seconds.
Education, not financial advice.
Common questions
What happened to the stock-to-flow model?
It stopped working after 2021. The model was published in March 2019 and fitted to data from 2009 to early 2019, so most of its famous track record predates it. For this cycle it pointed at roughly $714,000 while Bitcoin topped at $126,272 on 6 October 2025, a miss of 5.7 times. Bitcoin trades about 91% below the model today.
Is the stock-to-flow model dead?
Its predictions have been badly wrong since 2021, though its creator has responded to criticism and the debate continues. What is not in dispute is the arithmetic. Scored at every cycle top the model ran at 0.3x the 2013 and 2017 peaks, 1.1x the 2021 peak, and 5.7x the 2025 peak.
Why did the stock-to-flow model fail?
It measures supply and nothing else. Demand, liquidity, interest rates and who is buying are invisible to it. Bitcoin's supply schedule is known years ahead and never surprises anyone, so the only thing that could surprise the model is the demand side it does not look at. That side changed completely once ETFs and institutions arrived.
What did stock-to-flow predict for this cycle?
Using PlanB's published formula of 0.4 times the stock-to-flow ratio cubed, and Bitcoin's supply after the April 2024 halving, the model pointed at roughly $714,000. Bitcoin's actual intraday high for the cycle was $126,272.
How can I tell if a price model is backfit?
Find its publication date and ignore every result before it, since the model was built knowing those answers. Then ask what force it ignores, what outcome would prove it wrong, and whether the relationship holds on an asset or era it was not fitted to. Stock-to-flow fails three of those four, and all four were answerable in 2019.
Keep reading
We break down the market like this every day, free on Instagram and YouTube, and in depth inside the community.
Education, not financial advice. Trading involves real risk.