Bitcoin Halving Cycle Math: Four Data Points, One Big Bet

Every four years, Bitcoin’s block reward gets cut in half, and every four years a familiar chart resurfaces: a sharp rally, a euphoric peak, a long grinding drawdown, then the whole sequence again. Four halvings have happened so far, in November 2012, July 2016, May 2020, and April 2024, and each one has been followed by some version of that shape. As of July 20, 2026, Bitcoin trades around $64,600, with a market capitalization near $1.29 trillion and a circulating supply of roughly 20.06 million coins, about 95.5% of its eventual 21 million cap, according to CoinGecko.

The question this piece asks is not whether the cycle exists. Four rallies and four drawdowns clearly happened; nobody disputes the historical record. The question is narrower and less comfortable: does four repetitions of anything amount to evidence of a real, predictable pattern, or is it simply what a handful of data points from a young, fast-growing asset would look like whether or not the halving had any causal role at all? That is a statistics question as much as a crypto question, and it deserves to be treated like one.

It matters beyond academic curiosity. Every price target built on halving timing, and there have been many, from six-figure floors to seven-figure moonshots, ultimately rests on the assumption that four past repetitions are enough to extrapolate a fifth. If that assumption is shakier than it looks on a chart, so is every prediction built on top of it, including the ones getting the most attention as Halving Five approaches in 2028.

The Supply Math Nobody Disputes

Bitcoin’s issuance schedule is not a theory. It is code. Every 210,000 blocks, roughly four years given Bitcoin’s 10 minute target block time, the subsidy paid to miners cuts in half, a rule enforced directly in Bitcoin Core’s consensus logic inside the GetBlockSubsidy function in validation.cpp. The schedule runs from 50 BTC per block at genesis in 2009 down toward zero at block 6,930,000, expected around the year 2140, after 33 total halvings. The reward has already gone from 50 to 25 to 12.5 to 6.25 to today’s 3.125 BTC per block. Post halving annual issuance now runs under 1%, and the stock to flow ratio, the number of years of current production it would take to replace the existing supply, sits around 122, and is set to roughly double again after the next halving.

As of this week, the network sits at block height around 958,861, with roughly 91,139 blocks left before it reaches block 1,050,000 and triggers Halving Five. CoinGecko’s live halving countdown currently estimates that will land around mid April 2028, a date that shifts slightly with every fluctuation in average block time. None of this is in dispute. What happens to price around that date is where the argument starts.

The logic behind the design is straightforward. A fixed, publicly known issuance schedule, immune to central bank discretion or corporate board votes, was meant to give Bitcoin a scarcity profile closer to gold than to a fiat currency, one that gets programmatically harder to inflate rather than easier. Whether that scarcity translates predictably into price is a separate question from whether the scarcity itself is real, and the two get conflated constantly in halving-cycle discourse.

The Pattern Everyone Points To

Lay the four completed halvings side by side and the shape is hard to miss, even though the size of the move has shrunk every single time.

The direction is consistent: up sharply, then down sharply, four times running. The magnitude is not. Each cycle’s peak multiple has been smaller than the one before it, roughly 95x, then 30x, then 8x, then around 2x. That decay is itself the most interesting part of the table, and it is real regardless of how the causality argument below turns out.

Part of that decay has an intuitive, almost mechanical explanation that has nothing to do with statistics: a fixed dollar amount of new buying pressure moves a small market cap far more than it moves a large one. Bitcoin’s market cap was a rounding error next to global risk assets in 2012 and is a trillion-dollar-plus asset today, so even identical dollar inflows would be expected to produce a smaller percentage move now than they did then. That base rate effect alone could explain much of the shrinking multiple without requiring any cyclical, halving-specific mechanism at all, which is exactly the kind of alternative explanation a rigorous test would need to rule out before crediting the halving specifically.

Grading the Current Cycle So Far

Halving Four landed on April 20, 2024, with Bitcoin around $64,000. Measured at the roughly halfway point of the cycle, in mid April 2026, Bitcoin was up about 15% from the halving, trading just under $75,000, according to CoinDesk, which cited Glassnode data showing the current cycle had already underperformed the equivalent stretch of every prior cycle, consistent with the diminishing returns visible in the table above.

The cycle’s actual peak came earlier than that halfway snapshot suggests. Bitcoin hit its current all time high of $126,198 on October 6, 2025, in what traders nicknamed Uptober, then fell into the fastest correction of its cycle history, dropping roughly 36% to around $80,000 within six weeks. By early February 2026, NYDIG’s Global Head of Research, Greg Cipolaro, was already ranking the drawdown among Bitcoin’s worst, noting it had become the seventh largest and seventh longest drawdown in the asset’s history at that point, 122 days in with no bottom yet visible. The actual low came months later: $59,375 on June 5, 2026, a drawdown of about 53% from the October peak, before Standard Chartered’s Geoffrey Kendrick called the bottom, a read that looked reasonable in hindsight once Bitcoin recovered into its current mid $60,000s range.

The Other Way to Slice the Data

Halving day to peak is not the only way to measure the pattern. A second, independently constructed methodology, measuring peak to peak instead of trough to peak, tells a similar diminishing returns story from a different angle entirely.

That framing was central to a CoinDesk analysis published in July 2026 pushing back on the $300,000 to $500,000 targets some analysts, including veteran trader Peter Brandt, have floated for a hypothetical 2029 peak. Its argument: Bitcoin’s growing institutional and derivatives infrastructure is making the asset less volatile and more Wall Street like, which cuts against the odds of repeating even the smallest of its historical multiples, let alone something closer to the 30x to 95x moves of its first two cycles.

Two independently built tables, using different starting points and different math, land on the same qualitative conclusion. That consistency is real and worth taking seriously. It is also exactly the shape a purely statistical skeptic would expect from any asset that grew from a few dollars to six figures over fifteen years, cyclical mechanism or not, which is where the more uncomfortable questions start.

The Core Statistical Problem: Four Data Points

Strip away the charts and the four-year cycle is, statistically speaking, a claim resting on four observations. A November 2025 Protos analysis laid the case out bluntly: with only four completed cycles since 2009, there is no way to reliably distinguish a genuine, repeating structural pattern from what four random draws off a volatile, generally rising asset would produce anyway. Confidence intervals built on a sample of four are, by any conventional statistical standard, enormous.

Even analysts broadly sympathetic to cycle based thinking concede this. Writing in July 2026, one quantitative trader backtesting Bitcoin’s cycle timing signals against the 2018 and 2022 drawdowns put it plainly in a Quant Journey newsletter: “two data points isn’t a statistically robust sample,” a caveat attached to a model that was nonetheless used to project a deeper bottom for Bitcoin, around $56,000, forming in early 2027. The honesty is refreshing, and it also illustrates the underlying tension: the same people building trading models around the cycle are, in the same breath, conceding they do not have nearly enough history to validate them.

The Multiple Testing Trap

A second, related problem is what statisticians call multiple testing, or data mining: test enough different starting points, ending points, chart scales, and band widths against the same price series, and you will eventually find a combination that looks like a clean, repeating cycle purely by chance. Protos makes this point directly, arguing that backtesting a large number of different time periods against Bitcoin’s price history will almost always surface something that looks statistically significant, a reflection of how many candidate “cycles” are available to be found if you keep looking, not evidence that any single one of them is real.

Stock to flow, PlanB’s once dominant halving based valuation model, is a useful worked example of a related flaw: curve fitting dressed up as a causal model. A Bitcoin Magazine critique pointed out that the model’s core regression is close to tautological, since it effectively tests Bitcoin’s cumulative supply against its own market value, which is itself supply multiplied by price, and that the model’s statistical significance collapses toward zero once the analysis is adjusted for autocorrelation, a standard correction for time series data where each day’s value is heavily influenced by the day before it. A model can look extremely tight on a chart and still fail nearly every formal statistical test underneath it.

This is not a problem unique to Bitcoin. Traditional finance has its own long history of once-celebrated calendar effects and seasonal patterns that looked robust in backtests and then weakened or vanished once traders started positioning around them and once genuinely new, out of sample data arrived. The lesson generalizes: a pattern discovered by scanning historical data for the best fitting story is a different, weaker kind of evidence than a pattern predicted in advance and then confirmed. Bitcoin’s cycle narrative, discovered and refined entirely after the fact across four data points, sits closer to the first category than the second, however much it may still turn out to be directionally right.

Survivorship Bias and the Forecasts That Quietly Disappeared

The cycle narrative also benefits from a selection effect: forecasters who got a call right stay visible, and the ones who did not tend to fade from the conversation rather than face a public reckoning. Stock to flow’s own track record is the clearest example available. PlanB’s widely cited “worst case” floor for the 2021 cycle was around $98,000; Bitcoin actually peaked near $57,000 to $58,000 that cycle. The commonly cited 2024 to 2028 band for the model, often quoted around $500,000, has run far above Bitcoin’s actual $59,000 to $71,000 trading range through the first half of 2026. Yet stock to flow style thinking persists in crypto discourse, in part because the model’s early, more accurate years received far more attention than its later, badly missed ones.

A companion piece on this site, Bitcoin Halving Cycle Math: Grading the Price Predictions, went through this exercise in more detail, scoring named forecasters against their actual calls rather than their reputations. It is worth doing precisely because reputational memory in crypto is short and forgiving, and a model’s marketing tends to outlive its accuracy.

A Market That Keeps Changing Underneath the Model

A fourth problem, non-stationarity, is arguably the hardest one for cycle theory to answer. Treating four halvings as four trials of the same experiment assumes the underlying market is roughly the same each time, give or take noise. It plainly is not. The 2012 halving happened in a market with no regulated derivatives, no ETFs, and a few hundred million dollars of total value; the 2024 halving happened in a market with regulated spot ETFs holding about 6.5% of the entire circulating supply, deep CME futures and options markets, and multibillion dollar public mining companies. US crypto enforcement policy alone has shifted substantially since the first two halvings, changing who is legally able to hold Bitcoin and how.

The data backs this up in at least one place. Bitcoin’s realized capitalization, a measure of the aggregate cost basis of every coin in circulation, crossed $1 trillion for the first time in mid-2025 and held above that level even through the sharp correction that followed the October 2025 peak, something that did not happen in prior bear markets, when realized cap typically fell alongside price. That is a genuine structural difference, not a story about chart shapes. Skeptics of the cycle-is-dead reading counter that realized cap staying elevated could just as easily reflect long-term holders refusing to sell at a loss rather than genuine new structural demand, a distinction that will not be fully testable until the next serious drawdown either breaks that floor or does not.

It is also worth noting what a halving driven cycle is not: a universal feature of crypto assets, even ones that share Bitcoin’s code. Ethereum, for comparison, has run a completely different issuance model since its 2022 move to proof of stake, burning a portion of transaction fees under EIP-1559 rather than cutting a mining subsidy on a fixed schedule; liquid staking protocols built on top of that model, compared in detail in Lido vs Rocket Pool vs Frax, have no halving shaped event to point to at all. A companion piece on this site, Testing the Pattern Beyond BTC, pushed this further by comparing Bitcoin against other halving coins, Litecoin, Bitcoin Cash, and Zcash, alongside Monero’s permanent tail emission and Dogecoin’s never halved, fixed per block issuance, and found the pattern held unevenly even among assets that share Bitcoin’s own halving mechanism, let alone ones that do not.

Correlation, Causality, and What the Research Actually Finds

None of this settles whether the halving itself causes anything. Academic attempts to test causality directly have produced mixed, sample dependent results. One peer reviewed study, published in the International Journal of Financial Studies, tested for Granger causality between Bitcoin’s price and its hashrate, a proxy for miner behavior directly affected by halvings, across two different sample windows. In the earlier window, 2016 to 2017, the study found no evidence of Granger causality or cointegration in either direction. In the later window, 2017 to 2020, it found unidirectional causality running from price to hashrate, not the other way around, meaning price moves appeared to drive miner behavior more than miner economics drove price. That is close to the opposite of the simple story where a halving driven supply shock mechanically pushes price higher.

This does not mean the halving is irrelevant to price. It means the causal chain is almost certainly more tangled than “less new supply, therefore higher price,” and that the four year rhythm in the charts could be picking up a mix of halving effects, broader adoption cycles, macro liquidity conditions, and leverage cycles that happen to have historically moved on similar timescales. Separating those out with four overlapping examples is, honestly, closer to guesswork than to hypothesis testing, however well informed the guesswork is.

The academic literature on Bitcoin’s broader statistical structure tells a similar story of unresolved tension. A 2026 working paper on Bitcoin’s long-run power-law price trend, carrying the deliberately paradoxical title “Weak Structure, Strong Forecasts,” captures it well: the researchers found the statistical structure underneath Bitcoin’s price curve looks fragile under formal testing, even as simple power-law fits have still tracked the long-run trend more closely than most comparable models manage for other assets. Weak foundations and strong looking results are not mutually exclusive, which is an uncomfortable thing for any single number price target to rest on.

The Case for the Cycle, Made by the People Who Actually Trade It

None of this means the cycle camp is arguing in bad faith, or that every practitioner using it is ignoring the statistics. Fidelity’s Director of Global Macro, Jurrien Timmer, has been one of the more prominent voices defending the pattern’s continued relevance. Writing in December 2025, Timmer noted that “if we visually line up all the bull markets, we can see that the October high of $125,000 after 145 weeks of rallying fits pretty well with what one might expect,” and predicted 2026 could be a “year off” for Bitcoin, effectively betting the cycle’s rhythm, if not its exact multiples, would hold.

Not everyone agrees, including within the same industry. Bitwise CIO Matt Hougan argued in December 2025 that the cycle was set to break in 2026, writing that “the forces that previously drove four-year cycles, the bitcoin halving, interest rate cycles, and crypto’s leverage-fueled booms and busts, are significantly weaker than they’ve been in past cycles,” and pointing to reduced leverage after late 2025 liquidations, expected rate cuts, and ETF driven institutional access as reasons the old rhythm might not repeat. Timmer and Hougan are looking at the same four data points and drawing close to opposite conclusions, which on its own says something about how much genuine certainty the pattern actually supports.

That disagreement is worth sitting with rather than resolving artificially. Both analysts have access to the same four halvings, the same price charts, and largely the same on-chain data. What separates their conclusions is not a dispute over facts, it is a dispute over how much weight a short, noisy history deserves relative to structural arguments about what has changed. That is precisely the kind of disagreement four data points should be expected to produce, and precisely why treating either side’s confidence as proof of anything is a mistake.

What a Rigorous Test Would Actually Require

If four data points cannot settle the question, what would? Statisticians would generally want to see a few things cycle theory currently lacks.

• A specific, published prediction fixed before the outcome is known, not adjusted after the fact once price action starts to diverge from it.
• An out of sample test, where a model built on cycles one through three is used to predict cycle four’s actual shape without being allowed to see it first, then graded on the result whether it is flattering or not.
• A clear, pre-stated definition of what result would count as the theory failing, rather than being reinterpreted as confirmation.

That last point is where the non-falsifiability critique bites hardest. When Bitcoin’s 2026 drawdown ran longer and shallower than prior cycles, the response from cycle proponents was largely to describe the cycle as compressed or maturing rather than to treat the deviation as evidence against the underlying claim. A theory that can absorb any outcome by redefining itself after the fact is difficult to falsify, and a claim that cannot be falsified is difficult to call scientific, however useful it might still be as a rough, probabilistic heuristic rather than a law.

None of this is purely academic pedantry. Portfolio decisions, leverage levels, and entire content calendars in crypto media get built around specific dates derived from halving timing. A pattern treated as near certain by an audience that mostly sees the confirming cases is a pattern that will eventually disappoint someone relying on it at the worst possible moment, which is roughly what has already played out, repeatedly, with stock to flow.

Six Objections, Side by Side

Pulled together, the statistical case against treating the four year cycle as settled science comes down to six distinct objections, summarized here for reference.

No single objection here is fatal on its own, and cycle proponents have reasonable responses to most of them individually. Taken together, though, they explain why professional statisticians tend to describe the four-year cycle as, at best, an interesting prior rather than a validated model, and why that distinction matters far more to someone sizing a real position than it does to someone simply admiring a chart.

Halving Five and the Next Real-Time Test

Whatever the right statistical read is today, the theory is about to get a fifth, real time test it cannot dodge. Halving Five is expected at block 1,050,000, which the network is on pace to reach around mid April 2028 based on CoinGecko’s live countdown, roughly 91,000 blocks and about 21 months from today. The block subsidy will fall again, from 3.125 BTC to 1.5625 BTC, continuing the same hard coded schedule that has run without deviation since 2009.

What happens to price in the twelve to eighteen months on either side of that date will not, by itself, turn a four data point sample size problem into a five data point one that is meaningfully more conclusive. But a fifth consecutive smaller peak, following the roughly 95x, 30x, 8x, and 2x multiples already on the board, would at least be consistent with a genuine decay curve rather than random noise. A result that broke the pattern entirely, a bigger multiple than 2024’s cycle produced, or no meaningful rally at all, would be far harder to wave away as compression or delay.

A few specific things are worth watching as that window approaches: whether the eventual peak arrives before or after the halving itself, since all four prior peaks came afterward, ranging from roughly twelve to eighteen months post-halving; whether the resulting multiple lands anywhere close to a continued decay curve versus 2024’s roughly 2x; and whether realized cap and ETF-style institutional holdings, the two clearest markers of the market’s changed structure, keep behaving differently than they did in every cycle before 2024. Any of those breaking sharply from pattern would be more informative than another year of debate over chart shapes.

What This Means If You’re Using the Cycle to Make a Bet

None of the statistical objections in this piece mean the four year cycle is useless, and none of the pattern’s real, repeated history means it is reliable enough to bet a portfolio on with confidence. The honest middle ground, and roughly where most of the serious quantitative work referenced here actually lands, is to treat the cycle as a probabilistic prior worth a modest amount of weight, not a calendar and not a deterministic rule. Four data points can shift a prior. They cannot, on their own, prove a law. Anyone using halving timing to size a position is probably better served asking why this cycle might differ from the last one, given how much of the surrounding market structure has changed, than asking which month on a chart historically marked the bottom.

Frequently Asked Questions

Is Bitcoin’s four-year cycle based on real data, or is it just a pattern people see in charts?

It is based on real, documented price behavior. Bitcoin has rallied hard and then drawn down significantly following each of its first four halvings. The dispute is not whether those four rallies happened, it is whether four repetitions are enough to call the pattern a reliable, predictive cycle rather than a coincidence of a young asset’s early growth curve. Whether four data points meet the bar for genuine predictive evidence is a statistics question, not a chart-reading one, and most professional statisticians would say four is not enough on its own.

How many Bitcoin halvings have happened so far?

Four halvings have occurred: November 2012, July 2016, May 2020, and April 2024. Each cut the block subsidy in half, most recently from 6.25 BTC to 3.125 BTC. A fifth halving is expected around April 2028, once the network reaches block 1,050,000. Each halving has coincided with the start of a new multi-year price cycle so far, though the size of the resulting rally has shrunk with every repetition.

Why do statisticians say four halvings isn’t enough to prove a cycle exists?

Four observations produce an extremely wide confidence interval, too wide to reliably separate a genuine repeating pattern from random variation. Critics also point to the multiple testing problem: if you test enough different time windows and chart styles against Bitcoin’s price history, you will eventually find one that looks cyclical purely by chance. Survivorship bias compounds the problem, since forecasters whose cycle-based calls happened to work stay prominent, while those whose calls missed tend to quietly disappear from the conversation.

When is the next Bitcoin halving, and what will the reward drop to?

The fifth halving is expected around block 1,050,000, which CoinGecko’s live countdown currently estimates for around mid April 2028. It will cut the block subsidy from 3.125 BTC to 1.5625 BTC. That date has already shifted by a few days across recent recalculations as average block times fluctuate, so treat any specific date as an estimate rather than a certainty.

Did PlanB’s stock-to-flow model actually predict Bitcoin’s price correctly?

No. Its widely cited “worst case” floor of about $98,000 for late 2021 came in against an actual price near $57,000 to $58,000, and its 2024 to 2028 band, commonly cited around $500,000, has run far above Bitcoin’s actual mid-2026 trading range in the $59,000 to $71,000 area. The model’s early, more accurate years drew far more attention than its later, larger misses, a pattern common to popular price models generally.

Written by the HOGE Wire markets desk.