Bitcoin Price Prediction Methods: Which Ones Actually Work, and Which Are Just Hype

Stock-to-Flow, the Power Law, Metcalfe scaling, on-chain valuation and the four-year cycle — five of the most popular bitcoin price prediction methods, what each one really measures, how each has performed, and whether any of them belong in a real investment process.

Published 31 July 2026 · Research & education · ~12 min read

Disclaimer This article is educational content, not investment advice, and nothing in it is a recommendation to buy or sell any asset. We are not licensed financial advisors. Bitcoin is a volatile asset and you can lose your entire position. Every model described below has failed at some point, and past accuracy of any model tells you nothing reliable about its future accuracy. All figures are cited with their source date and were accurate as published; markets move, so verify current data before acting on anything. Do your own research and consider speaking to a regulated advisor about your own situation.

What's in this article

Why anyone models bitcoin at all
Stock-to-Flow (S2F) — the model that broke
The Power Law — the one that got peer-reviewed
Metcalfe's Law and network-value models
On-chain valuation: MVRV, realized price, NUPL
The four-year halving cycle
Honourable mentions: liquidity, flows, technicals
Scorecard: which methods survive contact with reality
So can you actually use these?
FAQ

1. Why anyone models bitcoin at all

Bitcoin has no cash flows, no earnings, no coupon and no board of directors. Every valuation tool built for equities and bonds — DCF, P/E, yield spreads — has nothing to grip. That vacuum is exactly why bitcoin has produced more homemade valuation models than any asset in modern financial history, and why so many of them are indistinguishable from astrology with a log chart.

The context matters for judging them. Bitcoin set an all-time high of $126,198 on 6 October 2025 and traded near $63,875 on the morning of 31 July 2026 — roughly half the peak, after a first half of 2026 that started above $93,000 and bottomed near $57,000 in June. Almost every popular model looked brilliant in October 2025 and looks strained today. That is the single most useful thing to know about this entire category.

DataAll-time high and 31 July 2026 price from reported daily market data; the January open and June low are approximate secondary figures. Every model below looked convincing at the left-hand bar.

Below are the five methods that dominate the conversation, in rough order of how much noise they generate — followed by an honest scorecard.

2. Stock-to-Flow (S2F): the model that broke in public

Verdict: historically important, practically dead

What it claims

Published in 2019 by the pseudonymous analyst PlanB, Stock-to-Flow borrows a commodity metric: divide existing supply (stock) by annual new production (flow). Gold has a high ratio, so gold is scarce, so gold is valuable. Bitcoin's issuance halves roughly every four years, so its ratio rises on a schedule — and PlanB fitted a regression from that ratio straight to market value.

Market value ≈ a × (Stock ÷ Flow)^b → price predicted from the issuance schedule alone

What actually happened

From roughly 2015 to late 2021 the fit was uncanny, which is what made S2F go viral. Then it detached. PlanB's "worst case" call of $98,000 for the end of November 2021 missed — bitcoin closed that month around $57,000 — and the December $100,000 target missed too, with the year closing near $47,000. Through 2022 the model implied six figures while bitcoin bottomed below $16,000. By August 2024 the gap between the model's line and spot had widened to roughly $130,000. The cross-asset variant had called for $288,000 by 2024.

DataModel targets as published by PlanB; actual prices as reported at each date. The 2022 and August 2024 figures are approximate as cited by contemporaneous coverage.

Why it broke

It has no demand term. Supply is fixed and known by everyone in advance; a variable every participant can read off a calendar cannot, by itself, explain price.
Statistical fragility. Critics have pointed to autocorrelation and spurious regression between two non-stationary, time-trending series — a fit that looks strong for reasons that have nothing to do with causation.
Unfalsifiable in practice. After the 2021 miss the model was defended as a "range" rather than a point forecast. A model that can absorb any outcome has stopped being a model.

Use it for Understanding why scarcity narratives are seductive. Not for position sizing, not for targets, not for timing.

3. The Power Law: the one that got peer-reviewed

Verdict: the most defensible of the group — but only on long horizons

What it claims

Physicist Giovanni Santostasi, who first sketched the idea in a 2014 Reddit post, argues that bitcoin's price grows as a power of time since the genesis block — the same scale-invariant pattern seen in city growth, earthquake magnitudes and animal metabolism. In his formulation price is proportional to time raised to roughly 5.8–5.9, usually rounded to six.

Price ≈ A × (days since genesis block)^n, n ≈ 5.8–5.9

Santostasi's derivation stacks two relationships: address growth follows a power law in time (roughly cubed), and price responds to address growth roughly as a square — a Metcalfe-style relationship. Multiply and you get time to the sixth power. On a log-log chart the result is a straight corridor with a support floor and an overheat band, which is also what the popular Rainbow Chart is doing in prettier colours.

SchematicIllustrative shape only — not plotted from price data. The corridor concept is what the Rainbow Chart renders in colour bands.

Track record

Better than S2F, and now formally reviewed: a study of the model was published online in Elsevier's Nonlinear Science on 29 June 2026, analysing 5,696 daily prices from July 2010 to February 2026 and reporting that the curve explains around 96% of historical price variation, with the gap between model and measurement staying under 1.6% over the studied period. That is real academic validation of a long-term fit — the first for a bitcoin price model.

The catch is timing. Santostasi's widely repeated projection of a roughly $210,000 cycle peak in January 2026 did not happen. The corridor's shape held; the schedule did not. His longer projections — around $1 million in about eight years, $10 million in roughly twenty — sit far outside any horizon over which the model has been tested.

Structural criticisms

Fitting a curve to log-log data over a period of explosive early growth almost guarantees a good R². High explained variance is not the same as predictive power.
The model's own parameters have been re-estimated as new data arrived, which softens the "it predicted this" claim.
It says nothing about drawdowns. A 50% fall inside the corridor is fully "on model" and financially devastating.
Because it is a function of time only, it cannot respond to anything — an ETF launch, a regulatory shock, a quantum-computing scare — until after the price has already moved.

Use it for A long-horizon sanity band: is price historically stretched or historically compressed? Not for dates, not for peaks, and never as a floor you leverage against.

4. Metcalfe's Law and network-value models

Verdict: the best theory of the group, the worst data

What it claims

Metcalfe's Law says a network's value scales with the square of its users. Applied to bitcoin, you proxy "users" with active addresses, non-zero-balance addresses or transacting entities, and compare that to market cap. This is the only model in the list with a genuine economic mechanism underneath it: bitcoin is a monetary network, and monetary networks really do get more useful as more people can be paid in them.

Network value ≈ k × (active users)² → compare to market cap for over/under-valuation

It is also the theoretical engine inside the Power Law, which is why the two tend to agree — and why agreement between them is less independent confirmation than it looks.

ConceptAdoption keeps growing while the on-chain series that measures it undercounts by a widening margin — so any Metcalfe fit calibrated before 2024 is reading a broken instrument.

Where it struggles

Addresses aren't people. One person can hold a thousand addresses; an ETF can hold hundreds of thousands of beneficial owners behind a handful. Since 2024, custodial and ETF ownership has grown fast while on-chain address counts do not reflect it at all.
Layer 2 and exchange netting move real economic activity off the base chain, so the input series systematically undercounts adoption.
The exponent is fitted, not derived. Analysts variously use n=1.5, 2, or a free parameter — and the "prediction" changes enormously depending on which they pick.

Use it for Direction of travel over years, and as a reality check on adoption narratives. Not for price levels — the input data is too polluted in the institutional era.

5. On-chain valuation: MVRV, realized price, NUPL

Verdict: the most useful family here — because it isn't a prediction

What it claims

On-chain valuation does something different: instead of forecasting a price, it measures where the market currently sits relative to what holders actually paid. Realized cap values every coin at the price it last moved, giving an aggregate cost basis. MVRV is market cap divided by realized cap. The MVRV Z-score standardises that gap by the historical volatility of market cap, so extremes are comparable across cycles.

MVRV Z = (market cap − realized cap) ÷ σ(market cap)

Historically, deep negative or near-zero Z-scores have clustered around the 2014, 2018 and 2022 bear-market lows, and very high readings around tops. In June 2026, with bitcoin's Z-score approaching zero after the sell-off, analysts noted the metric was re-entering the zone that preceded prior recoveries. By 13 July 2026 the Z-score read 0.37 with raw MVRV at 1.19 — market value sitting only slightly above aggregate cost basis.

DataZ-score 0.37 and raw MVRV 1.19 dated 13 July 2026. Zone labels are qualitative — providers draw the bands differently and compute on a lag, so check your own source before acting.

Related metrics in the same family: NUPL (net unrealised profit/loss) for aggregate holder psychology, SOPR for whether coins are being spent at profit or loss, short-term-holder realized price as a dynamic support/resistance level, and Thermocap and the Puell Multiple for miner-side stress.

Where it struggles

It's a thermometer, not a forecast. It tells you the market is cheap relative to cost basis. It cannot tell you it won't get cheaper, or for how long.
Sample size of three. "This zone marked the bottom in 2014, 2018 and 2022" is three observations. That is a story, not a statistical edge.
Custodial distortion. Coins moving into and out of ETF custody create on-chain events that don't mean what pre-2024 movements meant, which shifts realized cap in ways the historical bands never had to price in.
Data providers compute these differently and on lags, so two dashboards can disagree on the same day.

Use it for Risk posture — dialling exposure up in historically cheap zones and down in historically euphoric ones, over months. Not for entries, exits or targets.

6. The four-year halving cycle

Verdict: a real pattern being diluted in real time

What it claims

The simplest framework of all: halving → supply shock → mania → crash → long accumulation, repeating roughly every four years. It described 2012, 2016 and 2020 well enough to become folklore. The mechanics are not in dispute — the fourth halving occurred at block 840,000 on 20 April 2024, cutting the block subsidy from 6.25 to 3.125 BTC, and the next is expected around April 2028 at block 1,050,000, taking it to 1.5625 BTC.

ProtocolSubsidy figures and block heights from the Bitcoin issuance schedule; the 2028 date is an estimate because block timing varies. Note how each cut matters less in absolute terms than the one before it.

Why 2026 turned it into an argument

2025 closed in the red — the first time the post-halving year broke pattern — and the debate split the industry down the middle. Bitwise CIO Matt Hougan published a December 2025 memo titled "The Four-Year Cycle Is Dead," arguing that halving supply shocks, rate cycles and leverage booms have all weakened as institutional adoption matured, and predicting new highs in 2026. Grayscale took a similar line the same month. Michael Saylor has argued price is now driven by capital flows and institutional adoption rather than programmed supply cuts.

On the other side, analyst Benjamin Cowen has argued the cycle behaved exactly as it always does, topping in the fourth quarter, and that the "it's different" arguments are mental gymnastics. Fidelity's Jurrien Timmer characterised 2026 as a normal "off year" consistent with the historical pattern.

The honest reading is that the halving still matters but no longer explains the market by itself. The subsidy cut is arithmetic; ETF flows, corporate treasuries, derivatives positioning and global liquidity are now larger inputs than the change in daily issuance. If the pattern is breaking, it is breaking by dilution rather than disappearance — and the evidence to settle it will not arrive for years.

Use it for Understanding market psychology and expectations positioning. Not as a calendar you trade against — the 2025–26 stretch is the counter-example.

7. Honourable mentions

Three more frameworks come up constantly and deserve a line each:

Global liquidity / M2 overlay. Plotting bitcoin against global money supply with a lead-lag of roughly 8–12 weeks. Economically sensible — bitcoin is a long-duration risk asset and trades like one — but the lag is fitted after the fact and gets quietly adjusted whenever it stops working.
ETF and treasury flow models. Estimating price impact from net creations, redemptions and corporate purchases. This is the most genuinely new input since 2024 and arguably the most relevant, but flows are reported with a lag and are themselves price-reactive, which makes causality circular.
Technical analysis. Moving averages, RSI, Fibonacci levels, chart patterns. Useful as a shared language for where other traders have placed stops and orders; close to worthless as a forecast of a specific price on a specific date.

Our editorial placement of each method. The upper-right quadrant is nearly empty on purpose — no bitcoin valuation model currently combines a defensible mechanism with reliable, actionable output.

8. Scorecard

Method	What it really measures	Best horizon	Main failure mode	Verdict
Stock-to-Flow	Issuance schedule, nothing else	None	No demand variable; broke publicly in 2021–22	Hype
Power Law	Long-run growth corridor vs. time	Years	Accurate on shape, wrong on dates	Usable with care
Metcalfe / network value	Adoption vs. valuation	Years	Address data no longer proxies users	Usable with care
On-chain (MVRV etc.)	Price vs. aggregate holder cost basis	Months	Describes the present, not the future	Most useful
Four-year cycle	Historical rhythm + market expectations	Quarters to years	Institutional flows now dominate issuance	Contested
Technical analysis	Where other traders' orders sit	Days to weeks	Self-referential; no predictive edge	Context only

9. So can you actually use these — or is it all hype?

Both, and the distinction is sharper than it sounds. The failure isn't in the models; it's in what people ask of them.

What these methods genuinely provide: a rough sense of whether bitcoin is historically expensive or historically cheap, a framework for thinking about adoption and scarcity, and — in the on-chain family — an actual measurement of what the market has paid rather than a guess about what it will pay. That is worth something. It is the difference between "I'm buying because it's down" and "I'm adding because market value is close to aggregate cost basis, which has historically been a low-risk zone."

What none of them provide: a price on a date. Every single dated call examined above missed — S2F's $100,000 for December 2021, the Power Law's $210,000 for January 2026, the four-year cycle's expectation of a green 2025. The models that survive are the ones that stopped making dated calls.

The hype problem is structural, not accidental. Model-based targets are shared precisely because they are dramatic, and the sharing is loudest at cycle extremes — when a big number confirms what people already want to believe. A model quoted at you on social media is almost always being used as a marketing asset, not an analytical one. The tell is simple: does the person quoting it also quote the model's misses?

FrameworkOur editorial triage, not a published methodology. The test that does most of the work is the first one.

A defensible way to use them

Use models for posture, not for prediction. "Historically cheap zone → I'm comfortable accumulating" is a legitimate use. "Model says $210k in January → I'll leverage" is how people get liquidated.
Never let a model set a floor you borrow against. A "support level" from a fitted curve is not support. It is a line on a chart that price is free to cut straight through.
Weight mechanism over fit. Ask what would make the model wrong. If nothing could, it isn't telling you anything.
Cross-check independent families. Power Law and Metcalfe share an engine, so agreement between them proves little. On-chain valuation and flow data are genuinely separate inputs.
Assume 50%+ drawdowns are normal. Bitcoin is currently roughly half its October 2025 high, and that is well inside its historical range of behaviour. Size positions so that a repeat doesn't force a decision you'd regret.
Prefer boring execution. When the cycle debate is unresolved and every model disagrees, dollar-cost averaging is the strategy that doesn't require any of them to be right.

The short answer to "are these methods usable or just hype": the descriptive ones are usable, the predictive ones are hype, and most of what circulates online is the predictive kind.

10. FAQ

Which bitcoin price prediction method is the most accurate?

Over long horizons the Power Law has the strongest empirical fit, and a study of it was published in Elsevier's Nonlinear Science in June 2026, reporting that it explains roughly 96% of historical price variation over 5,696 daily observations. But accuracy over a fitted historical window is not the same as predictive accuracy — its dated call of about $210,000 for January 2026 did not materialise. For decisions you can actually act on, on-chain valuation metrics are more useful, because they measure the present rather than forecasting the future.

Is the Stock-to-Flow model dead?

As a forecasting tool, effectively yes. It detached from price in late 2021, implied six figures through the 2022 bear market when bitcoin traded below $16,000, and by August 2024 sat roughly $130,000 above spot. Its defenders reframed it as a wide range rather than a point forecast, which removed most of its remaining usefulness. It remains historically interesting as the model that shaped a generation of bitcoin narratives.

Is bitcoin's four-year cycle over?

Unresolved, and honest analysts on both sides say so. 2025 closed red, breaking the post-halving pattern, and firms including Bitwise and Grayscale argued in December 2025 that institutional flows have replaced the halving as the dominant driver. Others, including Benjamin Cowen and Fidelity's Jurrien Timmer, read the same data as the cycle behaving normally. The next halving is expected around April 2028; the evidence to settle the argument will not arrive before then.

What is the MVRV Z-score and why do analysts watch it?

It compares bitcoin's market capitalisation to its realized capitalisation — the aggregate price at which all coins last moved — and standardises the difference by the historical volatility of market cap. Readings near or below zero have coincided with the major bear-market lows of 2014, 2018 and 2022; very high readings have coincided with tops. It measures how stretched the market is relative to holder cost basis. It does not predict when that changes.

Should I invest based on any of these models?

No model in this article should be the reason you buy or sell, and this article is not investment advice. Every one of them has been wrong at some point, and several have been catastrophically wrong at exactly the moments they were most popular. Use them to frame risk, not to justify conviction — and consider a regulated advisor for decisions that matter to your finances.

Sources and dates: price and market-cap figures for 31 July 2026 and the 6 October 2025 all-time high; Elsevier Nonlinear Science power-law study published online 29 June 2026; MVRV Z-score reading of 0.37 dated 13 July 2026; halving block heights and subsidy figures from the Bitcoin protocol schedule. Figures were accurate as of the dates shown and are not updated in real time.

Disclaimer: Educational content only. Not financial, investment, tax or legal advice, and not a recommendation to transact in any asset. Cryptocurrency is highly volatile and you may lose your entire capital. We hold no responsibility for decisions taken on the basis of this article. Always do your own research.

Ready to put numbers on it? Open the calculator →