What Is the Bitcoin Four-Year Cycle?

The four-year cycle: halvings, issuance and the evidence. How Bitcoin’s issuance schedule and halvings work, what the dated record of past cycles shows, and why three or four cycles cannot settle whether the halving drives them.

Lesson 16 of 20Intermediateabout 30 minAll lessons

The short answer

The four-year cycle is the observed pattern of Bitcoin price peaks and deep declines roughly every four years, which many link to the halving, the cut in new coin issuance every 210,000 blocks. The record holds only three or four completed cycles, the published tests of a halving effect are mixed, and liquidity, leverage, attention and fund flows offer competing explanations.

You’ll learn

  • Calculate the block subsidy, daily and annual issuance, inflation rate and stock-to-flow ratio before and after each halving
  • Read the dated record of past cycle peaks, troughs, drawdowns and durations, and state how the data source changes the figures
  • Compare the halving thesis with rival explanations based on liquidity, leverage, attention and exchange-traded funds
  • Summarize what the published event studies of halvings find, and what their designs can and cannot show
  • Explain, with worked numbers, why three or four observations cannot establish a periodic cycle
Real story

Block 840,000, April 20, 2024

At 00:09 UTC on April 20, 2024, a miner produced Bitcoin block 840,000. From that block on, the new bitcoin created in each block fell from 6.25 to 3.125. This was the fourth halving, and the protocol had fixed its timing in 2009: the reward halves every 210,000 blocks. One thing about it was new. Five weeks earlier, on March 14, 2024, bitcoin had reached a record of $73,836, according to The Block. In every earlier cycle, the price had passed its previous record only after a halving. On March 5, 2024, it had already passed the November 2021 record, in a period when newly approved US spot bitcoin exchange-traded funds were drawing large inflows. So the most visible pattern in the asset’s history broke in the same cycle that was supposed to confirm it. This lesson sets out what the halving does mechanically, what the record shows, and what can and cannot be concluded from it.

Source: Block timestamp from <a href='https://mempool.space/block/840000'>mempool.space</a>; <a href='https://www.theblock.co/news/business/2024-03-15-spot-etf-inflows-slow-bitcoin-retreats-all-time-high-282767'>The Block (March 15, 2024)</a>; <a href='https://fortune.com/crypto/2024/03/05/bitcoin-breaks-record-69000-spot-etfs/'>Fortune (March 5, 2024)</a>

How new bitcoin is issued

Bitcoin has no issuer. New coins enter circulation through the first transaction of each block, called the coinbase transaction, which the miner of that block writes. The protocol limits how much that transaction may pay out: the block subsidy plus the fees paid by the transactions in the block. Nodes reject a block that pays itself more. The original white paper describes the incentive in general terms. It says new coins are added “until” a predetermined number of coins has entered circulation, after which fees alone pay miners (Nakamoto, 2008).

Definition

Block subsidy: The amount of newly created bitcoin a block may award to its miner, set by the protocol as a function of block height alone. It excludes transaction fees, which are paid by users and create no new coins.

Definition

Halving: The point, every 210,000 blocks, at which the block subsidy is cut in half. In Bitcoin Core’s code the halving interval is the parameter nSubsidyHalvingInterval = 210000.

The rule in Bitcoin Core is short. The function GetBlockSubsidy counts the halvings so far as the block height divided by 210,000, using whole-number division. It starts from 50 bitcoin, expressed in satoshis (one bitcoin is 100,000,000 satoshis), and shifts that number right by one binary place per halving. A right shift by one place halves a whole number and discards any remainder. After 64 halvings the function returns zero.

Subsidy(h) = floor( 5,000,000,000 / 2floor(h / 210,000) ) satoshis

h is the block height. 5,000,000,000 satoshis is 50 bitcoin. The floor operations reflect integer arithmetic in the code, so the subsidy is always a whole number of satoshis.

Summing the schedule gives the supply cap. Each of the 33 eras that pays anything contributes 210,000 blocks times that era’s subsidy. Because the shift discards fractions of a satoshi, the sum is slightly below 21 million: 20,999,999.9769 bitcoin. The subsidy reaches zero at the 33rd halving, at block 6,930,000. The real circulating supply is lower still, because some coins are lost and a few early blocks claimed less than the permitted amount. The cap is an upper bound, not a count of usable coins.

Time enters only through blocks. The protocol targets one block per 600 seconds (nPowTargetSpacing = 10 * 60) and resets the mining difficulty every 2,016 blocks, a target span of two weeks. The white paper describes the idea as a moving average targeting an average number of blocks per hour. At the target, a day has 24 × 6 = 144 blocks. When hash power grows faster than difficulty resets, blocks arrive early, which is why halvings have come before the four-year mark.

EraStart date (UTC)Days in eraAverage blocks per day
Block 0 to 210,000January 3, 20091,425147.4
210,000 to 420,000November 28, 20121,319159.2
420,000 to 630,000July 9, 20161,402149.8
630,000 to 840,000May 11, 20201,440145.8

Days between the dates of the boundary blocks, from block timestamps on mempool.space. The first row is measured from the genesis block. Every era ran faster than the 144-block target, most of all during the rapid growth in hash power from 2012 to 2016.

Issuance before and after each halving

Two ratios describe issuance. The annual issuance rate is new coins per year divided by the existing stock, which is a supply inflation rate. Stock-to-flow is its reciprocal: the stock divided by one year of new supply. Both are known in advance from the schedule, up to the small variation in block times.

Definition

Stock-to-flow (S2F): The existing stock of an asset divided by its annual new production. A ratio of 60 means current production would take 60 years to equal the existing stock. The ratio is commonly applied to commodities such as gold; an analyst writing as PlanB applied it to bitcoin in 2019, crediting the idea to Saifedean Ammous.

Annual issuance rate = F / S     S2F = S / F     F ≈ subsidy × 144 × 365

S is the stock (coins in existence), F is the flow (new coins per year). The 144 × 365 term assumes the ten-minute target; actual block rates ran a few percent higher.

Halving (block)Date (UTC)SubsidyStock at the halvingAnnual flow before → afterIssuance rate before → afterS2F after
1st (210,000)Nov. 28, 201250 → 2510,500,0002,628,000 → 1,314,00025.0% → 12.5%8.0
2nd (420,000)July 9, 201625 → 12.515,750,0001,314,000 → 657,0008.3% → 4.2%24.0
3rd (630,000)May 11, 202012.5 → 6.2518,375,000657,000 → 328,5003.6% → 1.8%55.9
4th (840,000)Apr. 20, 20246.25 → 3.12519,687,500328,500 → 164,2501.67% → 0.83%119.9

Stock is the scheduled total at that height; flows use 144 blocks per day. Dates are block timestamps. Coin Metrics’ measured supply at the end of each of those days was within a few hundred coins of the scheduled figure, the gap being the blocks mined later that day.

Example: the 2024 halving in coins and dollars

  1. Before: 6.25 × 144 = 900 bitcoin per day at the target rate, or 900 × 365 = 328,500 per year.
  2. After: 3.125 × 144 = 450 per day, or 164,250 per year.
  3. Stock at block 840,000 = 210,000 × (50 + 25 + 12.5 + 6.25) = 19,687,500 bitcoin.
  4. Issuance rate after = 164,250 / 19,687,500 = 0.834% a year. S2F = 19,687,500 / 164,250 = 119.9.
  5. Check against measured data: Coin Metrics records 931.25 new bitcoin on April 19, 2024, which is 149 blocks × 6.25, and 437.5 on April 21, which is 140 blocks × 3.125.
  6. Measured first-year issuance after the halving (April 21, 2024 to April 20, 2025) was 166,166 bitcoin, about 1.2% above the 164,250 target because blocks arrived slightly faster than every ten minutes.
  7. In dollars, at Coin Metrics’ reference price of $64,908 on April 20, 2024: daily new supply fell from about 900 × 64,908 = $58.4 million to about $29.2 million.
The halving removed roughly $29 million a day of new coins at that price. Whether that is large depends on what it is compared with, which is the question the rest of this lesson takes up.

The halving’s direct effect is on miners. Their subsidy revenue in bitcoin halves overnight, while their costs, mostly power and hardware paid in dollars, do not. Unless price or fees rise, the least efficient operations become unprofitable and may switch off, and miners as a group have half as many new coins to fund their costs. Glassnode’s documentation of the Puell Multiple, a miner-revenue ratio taught in the next lesson, notes that a halving mechanically cuts that ratio in half. The miner channel is real. Its size relative to all trading is the open question.

The stock-to-flow model and its test

PlanB’s March 22, 2019 article, “Modeling Bitcoin’s Value with Scarcity,” regressed the logarithm of bitcoin’s market value on the logarithm of its stock-to-flow ratio and reported an R2 of 95%. The article gave stock-to-flow values of 62 for gold, 22 for silver and 25 for bitcoin at the time. It translated the model into a market value of about $1 trillion after the May 2020 halving, about $55,000 per coin. That figure is described here as a past model output, not as a claim about value.

The model has a structural problem that a student can see without data. Stock-to-flow for bitcoin is a deterministic function of block height, which is close to a deterministic function of time. A regression of log price on log S2F is therefore close to a regression of log price on a step-shaped time trend. Any series that grows on average over time fits such a trend well in sample. A high R2 is evidence of a common trend, which is weak evidence of a causal link. A second problem is that the explanatory variable was known years in advance. If the schedule set the price, the known future halvings would already be reflected in prices before they happened.

The model also made a dated, checkable claim. Decrypt reported on June 21, 2022 that the model had projected a range of $100,000 to $110,000 for 2022, while bitcoin fell below $20,000. In the same report, Vitalik Buterin called the theory that halvings cause price rises “unfalsifiable” and said such models were harmful. PlanB replied that the model had had a good run from March 2019 to March 2022. A model that fits three years and then misses by a factor of five is a standard example of the gap between in-sample fit and out-of-sample performance.

The dated record of the cycles

Any claim about cycles rests on a short list of dates. The table below uses Coin Metrics’ daily reference price, a single price per day built from several exchanges. Intraday highs and lows on individual exchanges are higher at tops and lower at bottoms, so other sources date and price the same turning points slightly differently. For example, The Block gives $73,836 for March 14, 2024, while Coin Metrics’ daily reference price peaked at $73,082 on March 13. Neither is wrong. They measure different things.

CyclePeak (daily reference)Trough after the peakDrawdownPeak to troughPeak multiple vs prior peak
2011June 8, 2011: $29.03Nov. 18, 2011: $2.1192.7%163 daysn/a
2013–15Dec. 4, 2013: $1,134.93Jan. 14, 2015: $175.6484.5%406 days39.1×
2017–18Dec. 16, 2017: $19,640.51Dec. 15, 2018: $3,185.0783.8%364 days17.3×
2021–22Nov. 8, 2021: $67,541.76Nov. 9, 2022: $15,758.2976.7%366 days3.44×
2025–26Oct. 6, 2025: $124,824.45June 30, 2026: $58,525.08 (lowest so far)53.1% so far267 days so far1.85×

Source: Coin Metrics community data (PriceUSD), retrieved November 21, 2026, computed for this lesson. The trough is the lowest daily reference price after each peak. For 2025–26 it is the lowest through November 20, 2026, when the reference price was $58,851, and it is not a confirmed cycle low.

Three regularities appear, and each needs a caveat. First, drawdowns have become smaller: 93%, 85%, 84%, 77%, and 53% so far. Second, the peak multiple has shrunk sharply, from 39 times to under 2 times. Third, timing has been regular. The three completed trough-to-peak spans ran 1,067, 1,059 and 1,062 days. The halving-to-peak spans ran 371, 525, 546 and 534 days. The 2013 cycle complicates the count, because it had two peaks: one near $231 on April 9, 2013 and the higher one in December. Which one is called the cycle top changes the 2013 spacing.

The 2011 cycle came before the first halving, and its peak and collapse happened on a market with a handful of exchanges. Some analysts include it and some do not. That choice alone changes the sample from five cycles to four. The 2025–26 cycle is unfinished. So the record contains at most four completed cycles, and at most three that each began after a halving.

Example: computing drawdown and multiples

  1. Drawdown = 1 − trough ÷ peak. For 2021–22: 1 − 15,758.29 / 67,541.76 = 1 − 0.2333 = 76.7%.
  2. Recovery needed from that trough to regain the peak = peak ÷ trough − 1 = 67,541.76 / 15,758.29 − 1 = 329%. A 76.7% fall needs more than a fourfold rise to undo.
  3. Peak multiple, 2021 to 2025 = 124,824.45 / 67,541.76 = 1.85.
  4. Drawdown from the October 2025 peak to the November 20, 2026 reference price = 1 − 58,851.02 / 124,824.45 = 52.9%.
Drawdowns and recoveries are not symmetric. Percent falls compress large ratio moves, so compare cycles in ratios or logarithms, not in percentage points.

The halving thesis and its rivals

The halving thesis, in its most common form, runs as follows. A halving cuts new supply. If demand holds, less new supply must be absorbed, so price rises. The rise draws attention and leverage, which overshoots, and the collapse that follows ends near the next halving. The thesis makes one causal claim (lower issuance raises price) and one dynamic claim (the market overshoots on a roughly four-year rhythm). These are separate claims, and evidence for one does not support the other.

The main objection is that the schedule is public. A trader who expects a supply cut to raise prices has every reason to act before it. In a market where information is priced quickly, a known future supply change should be reflected gradually, not as a jump after the event. A defender can reply that markets are not that efficient, that many participants are unaware of the schedule, or that the effect works through miner behavior that only changes when revenue actually falls. Those replies are possible, but each is a claim needing its own evidence.

Four rival or overlapping explanations have support of their own.

Monetary conditions and liquidity. An International Monetary Fund working paper by Che, Copestake, Furceri and Terracciano (2023) finds that one common factor explains about 80% of the variation in crypto prices, and that US monetary tightening lowers that factor through a risk-taking channel, as it does for equities. An IMF blog by Adrian, Iyer and Qureshi (2022) reports that the correlation of daily bitcoin and S&P 500 moves was 0.01 in 2017–19 and 0.36 in 2020–21. If crypto moves with global risk appetite, cycles in that appetite matter, and US monetary cycles do not run on Bitcoin’s block schedule.

Leverage. Derivatives allow positions larger than the capital behind them. Forced closures in a falling market can extend a decline, as the derivatives lessons in this course show for October 10–11, 2025. Leverage builds up in rising markets and is removed in crashes, which by itself can produce boom and bust behavior with no fixed period.

Attention. Kristoufek (2013), in Scientific Reports, found that bitcoin prices correlated strongly with Google search volume (0.8786) and Wikipedia views (0.8271), on log scales, with effects running in both directions. Rising interest raised prices mostly when prices were already above trend. Attention can amplify a move that starts for other reasons, and it partly follows price rather than leading it.

New investor channels. The US Securities and Exchange Commission approved the listing of spot bitcoin exchange-traded products on January 10, 2024. Its chair’s statement said the agency “did not approve or endorse bitcoin.” These funds gave a new route to exposure, and their flows are reported daily. On March 14, 2024, for example, BitMEX Research data cited by The Block showed net inflows of $132.5 million, after outflows of $257.1 million from Grayscale’s GBTC. A new demand channel that opens on a regulator’s calendar is independent of the halving schedule.

The halvings, US election years and several monetary-policy turns have all occurred on roughly four-year spacings in this short history. When candidate causes line up in time, the data cannot separate them. That is a confounding problem, not a finding in favor of any one cause.

What the published tests find

Formal tests are few and recent, and they disagree.

Virtonen (2025), an arXiv preprint that has not been through peer review, uses the synthetic-control method. The method builds a comparison series, a weighted combination of other assets that tracked bitcoin before the event, and treats the gap after the event as the effect. For the April 2024 halving, the paper finds a positive effect three months after the event, equal to about one fifth of bitcoin’s percentage price change over a 17-month window from April 2, 2023 to July 21, 2024. For 2020, it fails to obtain a statistically significant and robust estimate. The method depends on finding assets that resemble bitcoin and were not themselves affected by the halving or by the same news, a strong assumption for crypto assets that move together.

Veloso, Gatsios, Magnani and Lima (2025), in the Journal of Risk and Financial Management, run an event study of cumulative abnormal returns and volatility around the 2012, 2016, 2020 and 2024 halvings, focusing on the last two. They report that the 2024 halving coincided with elevated abnormal returns and volatility, but that halving-period returns and volatility declined from 2012 to 2024, which they relate to rising regulation and institutional participation. An event study measures what happened near a date. Unless the comparison model is right, it cannot separate the halving from anything else that happened near the same date.

Chen (2025), a theoretical arXiv preprint, derives a model in which a halving predicts a contraction in both hash rate and price, the opposite of the popular thesis, and proposes a test without reporting results. It is a prediction from a model, not evidence.

Read together, the studies show that the sign and size of any halving effect are contested, that the best-identified estimate covers one event, and that none can test the claim of a recurring four-year rhythm, which needs many cycles to test.

Why three or four observations limit any conclusion

A cycle claim has two parts that need testing: that the pattern repeats, and that it repeats for a stated reason. Both are hard with three or four observations.

Example: a confidence interval from three drawdowns

  1. Take the three completed post-halving drawdowns: 84.5%, 83.8% and 76.7%. Their mean is 81.67%.
  2. Sample standard deviation: deviations are 2.83, 2.13 and −4.97. Squares sum to 37.24. Divide by n − 1 = 2 to get 18.62, so s = 4.32.
  3. Standard error = s / √3 = 4.32 / 1.732 = 2.49.
  4. With 2 degrees of freedom, the 95% t value is 4.303. Half-width = 4.303 × 2.49 = 10.7 points.
  5. 95% interval: 71.0% to 92.4%.
The interval is about 21 points wide. The 2025–26 drawdown so far, 53.1%, lies far outside it, which illustrates a second problem: the series is not drawn from one fixed distribution. Each cycle happened in a larger, different market, so averaging them assumes a stability the data reject.

Example: how often a timing pattern appears by chance

  1. Suppose, as a simple null model, that each cycle peak falls at a random point in the 48 months after a halving.
  2. A window chosen in advance, 12 to 18 months after the halving, covers 6 of 48 months, so it has probability 1/8 for one peak.
  3. Three independent peaks all landing in it: (1/8)3 = 1/512, about 0.2%.
  4. But the window was chosen after the peaks were seen. Suppose analysts looked at 20 possible patterns (other windows, other anchor dates, troughs instead of peaks, election years). The chance that at least one fits all three by luck is 1 − (1 − 1/512)20 = 3.8%.
  5. With 100 candidate patterns, the figure is about 18%.
A pattern found after looking at the data is weaker evidence than its raw probability suggests. The problem is the same multiple-testing problem covered in the evidence module of this course.

There are further problems. The observations are not independent, because a cycle that is widely believed in can coordinate behavior: if enough traders expect a peak about 18 months after a halving and act on it, the timing can recur for reasons unrelated to supply. The sample has survivorship, since bitcoin is the asset studied because it survived. And the cycle is often redrawn after the fact, for example by moving the anchor from halving to trough when one stops fitting. A falsifiable version would state in advance which dated outcome would count against it.

Open question

Is the four-year cycle still a useful description?

The case that it is. The October 2025 peak fell 534 days after the 2024 halving, close to the 525 and 546 days of the previous two cycles, and the trough-to-peak spans of the last three cycles were within eight days of each other. Benjamin Cowen, an analyst who publishes as Into The Cryptoverse, argued in March 2026 that bitcoin tops in the fourth quarter of the year after a halving, and that the October 2025 peak fitted that timing, as Cointelegraph Magazine reported. A defender can add that a pattern does not need a supply mechanism to be useful as a description; coordination around a shared belief can sustain it. The case that it is not. Grayscale said in December 2025 that it expected the end of the so-called four-year cycle, and the same Cointelegraph report lists several fund managers and analysts who reject strict cycle timing. Several features of 2024–26 departed from earlier cycles: the record came before the halving, the peak multiple was the smallest on record, and the drawdown so far is much shallower. Large new buyers, such as exchange-traded funds, act on flows that have no link to block height. What the evidence allows. The timing regularity is real as a description of three or four events. It is too short a record to separate a supply effect, a self-fulfilling belief, a monetary cycle that happened to line up, and chance. Both sides can point to the same data, which is itself the main finding.

Common mistakes

  • Treating the halving as a surprise. Its block height is fixed and public years in advance.
  • Calling stock-to-flow a measure of demand. It is a supply ratio computed from the schedule; any price model built on it must add a demand assumption.
  • Reading a high in-sample R² as proof of cause. A deterministic trend fits any series that rises over time.
  • Mixing data sources. An intraday high on one exchange and a daily reference price give different peaks and dates.
  • Counting the unfinished current cycle as a completed observation, or treating its lowest price so far as the cycle low.
  • Averaging drawdowns across cycles as if they came from one stable distribution.
  • Choosing the timing window after seeing the peaks, then reporting the chance of that window as if it had been fixed in advance.

Problem set

  1. Compute the block subsidy at heights 1,049,999 and 1,050,000, the daily issuance at 144 blocks before and after, and the scheduled stock at block 1,050,000.
    Show solution
    floor(1,049,999 / 210,000) = 4, so the subsidy is 50 / 24 = 3.125. floor(1,050,000 / 210,000) = 5, so it is 50 / 25 = 1.5625. Daily issuance falls from 3.125 × 144 = 450 to 1.5625 × 144 = 225 bitcoin. Stock = 19,687,500 + 210,000 × 3.125 = 19,687,500 + 656,250 = 20,343,750 bitcoin.
  2. Using the stock from the previous problem, compute the annual issuance rate and stock-to-flow ratio immediately after the fifth halving at the 144-block target. What does the result imply for any model that uses S2F as its only input?
    Show solution
    Annual flow = 225 × 365 = 82,125. Rate = 82,125 / 20,343,750 = 0.404%. S2F = 20,343,750 / 82,125 = 247.7. The ratio doubles at each halving and is known decades ahead. A model with S2F as its only input therefore makes a fixed, pre-set path of implied values; it has no way to respond to demand, which is exactly what out-of-sample tests check.
  3. The 630,000 to 840,000 era lasted 1,440 days. Compute the average block interval in minutes. How many days would it have lasted at exactly 144 blocks a day?
    Show solution
    1,440 days × 1,440 minutes = 2,073,600 minutes / 210,000 blocks = 9.87 minutes per block. At 144 blocks a day the era would take 210,000 / 144 = 1,458.3 days, about 18 days longer.
  4. A drawdown of 84% is followed by a drawdown of 77%. Compute the price ratio (trough ÷ peak) for each and say how much larger the second is in ratio terms.
    Show solution
    Ratios are 1 − 0.84 = 0.16 and 1 − 0.77 = 0.23. 0.23 / 0.16 = 1.44, so the second trough kept 44% more of its peak value, although the percentage-point gap is only 7. Ratios describe the change better.
  5. An analyst says: “The 2024 halving cut new supply by $29 million a day, and the price then rose to a record, so the halving caused the rise.” Give three reasons the inference is weak.
    Show solution
    (1) The record of March 2024 came before the halving, so the cut cannot explain the rise to that point. (2) Other demand changed at the same time, for example spot fund inflows, which in single days were reported in the hundreds of millions of dollars, so the halving is confounded with other causes. (3) The cut was known in advance, so a forward-looking market had reasons to reflect it before the event. A design such as a synthetic control tries to address this, and even it found an effect for 2024 and not for 2020.
  6. Using only the three completed post-halving drawdowns (84.5%, 83.8%, 76.7%), compute the 90% confidence interval for the mean, given the 90% t value with 2 degrees of freedom of 2.920.
    Show solution
    Mean = 81.67, standard error = 2.49 (from the worked example). Half-width = 2.920 × 2.49 = 7.27. Interval: 74.4% to 88.9%. Even at 90% confidence the interval is about 15 points wide, and the method assumes the three values come from one stable distribution, which the shrinking trend contradicts.

References

  1. Nakamoto, S. (2008). Bitcoin: A Peer-to-Peer Electronic Cash System. Source
  2. Bitcoin Core. src/validation.cpp (GetBlockSubsidy) and src/kernel/chainparams.cpp (nSubsidyHalvingInterval, nPowTargetTimespan, nPowTargetSpacing). Source code, read November 2026. Source
  3. Coin Metrics. Community API, asset metrics PriceUSD, SplyCur and IssTotNtv for BTC, daily, retrieved November 21, 2026. Source
  4. PlanB (March 22, 2019). Modeling Bitcoin’s Value with Scarcity. Medium. Source
  5. Asmakov, A. (June 21, 2022). Vitalik Buterin: Stock-to-Flow Bitcoin Price Model ‘Really Not Looking Good Now’. Decrypt. Source
  6. Virtonen, V. (2025). Estimating the Impact of the Bitcoin Halving on Its Price Using Synthetic Control. arXiv:2511.05512 (preprint, not peer-reviewed). Source
  7. Veloso, V., Gatsios, R. C., Magnani, V. M., and Lima, F. G. (2025). Is Bitcoin’s Market Maturing? Cumulative Abnormal Returns and Volatility in the 2024 Halving and Past Cycles. Journal of Risk and Financial Management, 18(5). Source
  8. Chen, L. (2025). A Game-Theoretic Foundation for Bitcoin’s Price: A Security-Utility Equilibrium. arXiv:2508.06071 (preprint). Source
  9. Che, N. X., Copestake, A., Furceri, D., and Terracciano, T. (2023). The Crypto Cycle and US Monetary Policy. IMF Working Paper 2023/163. Source
  10. Adrian, T., Iyer, T., and Qureshi, M. S. (January 11, 2022). Crypto Prices Move More in Sync With Stocks, Posing New Risks. IMF Blog. Source
  11. Kristoufek, L. (2013). BitCoin meets Google Trends and Wikipedia: Quantifying the relationship between phenomena of the Internet era. Scientific Reports, 3, 3415. Source
  12. Gensler, G. (January 10, 2024). Statement on the Approval of Spot Bitcoin Exchange-Traded Products. US Securities and Exchange Commission. Source
  13. Hunt, J. (March 15, 2024). Spot ETF net inflows slow as bitcoin retreats from latest all-time high. The Block. Source
  14. Rowe, N. (March 5, 2024). Bitcoin breaks all-time high of $69,000 as investors flock to spot ETFs ahead of halving. Fortune. Source
  15. Lyons, C. (March 6, 2026). The debate over Bitcoin’s four-year cycle is over: Benjamin Cowen. Cointelegraph Magazine. Source

Exercise

  1. Open block 839,999 and block 840,000. In each, open the coinbase transaction and note its total output. Subtract the block’s total fees to recover the subsidy, and check it against the formula in this lesson.
  2. Note the timestamps of blocks 630,000 and 840,000, and compute the average block interval for that era.
  3. Open the Coin Metrics community API link in the references, change the dates to April 15–25, 2024, and divide each day’s IssTotNtv by that day’s subsidy to get the number of blocks mined.
  4. Using the same API with PriceUSD, find the highest and lowest daily reference prices between two dates of your choice, and compute the drawdown and the recovery needed.
  5. Write down, before looking at any more data, one dated outcome that would count against a four-year cycle. Explain why it has to be stated in advance.
Open mempool.space block explorer →

Check yourself

1. What sets the date of a Bitcoin halving?

The rule is height-based. Because blocks have arrived slightly faster than every ten minutes, halvings have come before the four-year mark.

2. After a halving, the annual issuance rate falls by about half and stock-to-flow about doubles. Why does this make a regression of log price on log S2F hard to interpret?

Any rising series fits a deterministic trend well in sample. The fit cannot tell a supply effect from a common trend.

3. Why do two sources give different values for the same cycle peak?

An intraday high on one venue exceeds a daily reference price built from several venues. State which measure you use.

4. What was unusual about the 2024 halving compared with earlier ones?

Bitcoin passed its November 2021 record on March 5, 2024 and set a new record on March 14, before the April 20 halving.

5. A known future supply cut should, in a market that prices public information quickly, affect prices:

Anticipation moves the effect forward in time. Defenders must explain why the market would fail to price a public schedule.

6. A synthetic-control study finds a halving effect for 2024 but not for 2020. The best reading is:

One significant estimate and one null result, from a preprint, do not establish a general effect.

7. Three cycle peaks fall in a window chosen after seeing them. Why is the 1/512 chance of that misleading?

With many candidate windows or anchors, one fitting all three by luck becomes far more likely. This is a multiple-testing problem.

8. Halvings, US election years and several monetary-policy turns have similar spacing. What does that imply?

When candidate causes coincide, a short record cannot separate them. That calls for designs with a comparison, not more narrative.

Common questions

What is a Bitcoin halving?

The point, every 210,000 blocks, at which the block subsidy paid to miners is cut in half. The fourth halving, at block 840,000 on April 20, 2024, cut it from 6.25 to 3.125 bitcoin per block.

What is the stock-to-flow ratio?

The existing stock of an asset divided by one year of new supply. For bitcoin it is set by the issuance schedule, about 120 after the 2024 halving, so it rises on a known path regardless of demand.

Does the halving cause Bitcoin's price to rise?

The evidence is mixed. One synthetic-control preprint finds an effect for 2024 but none for 2020, the schedule is public in advance, and with three or four cycles the halving cannot be separated from other causes or from chance.

This lesson teaches methods and describes past events. It is not investment advice. Live panels: tracked whale wallets, updated 2026-11-22 07:40 UTC.