The Bitcoin Halving and the 4-Year Cycle: How Strong Is the Evidence?

The halving is the point where the reward for bitcoin miners is cut in half, every 210,000 blocks or roughly every 4 years. Many people believe it drives a “4-year cycle”: a strong rally after the halving, then a deep crash before the next one. This article checks the evidence layer by layer: how much supply actually disappears, the problem of having only a handful of cycles, the stock-to-flow model, and finally how to size a position. There are no price forecasts here, and nothing about when to buy or sell.

What is a halving

Bitcoin produces a new block about every 10 minutes. The miner who finds it receives newly created coins, called the block subsidy, plus the fees in that block. The code halves the subsidy every 210,000 blocks. At 10 minutes per block, 210,000 blocks is about 1,458 days, almost exactly 4 years.

Halving	Block	Date (UTC)	Subsidy before → after (BTC)
Start (genesis)	0	3 Jan 2009	50
1	210,000	28 Nov 2012	50 → 25
2	420,000	9 Jul 2016	25 → 12.5
3	630,000	11 May 2020	12.5 → 6.25
4	840,000	20 Apr 2024	6.25 → 3.125

Across all halvings the total stays under 21 million, with the last coins issued around the year 2140. Mining difficulty adjusts every 2,016 blocks to pull block times back towards 10 minutes, so halving dates can be estimated weeks or days ahead.

The supply removed shrinks every time

The main argument for the cycle is a supply shock: the new coins that miners usually have to sell partly to pay for electricity are cut in half overnight. If demand stays the same, price should rise. Here is how big that effect actually was in 2024.

1. Coins in existence at the 2024 halving = 210,000 × (50 + 25 + 12.5 + 6.25) = 19,687,500 BTC, or 93.75% of 21 million
2. Before the halving: 6.25 BTC × 144 blocks a day = 900 BTC a day × 365 = 328,500 BTC a year = 1.67% of existing supply
3. After the halving: 3.125 × 144 = 450 BTC a day × 365 = 164,250 BTC a year = 0.83%
4. New supply removed = 1.67 − 0.83 = about 0.83 percentage points of existing supply per year

Doing the same for every halving, the share removed was 12.5 points in 2012, 4.2 points in 2016, 1.8 points in 2020 and 0.83 points in 2024 (calculated at the 10-minute target; real blocks have come slightly faster, so the true figures are a little higher).

Two more things weaken the argument.

• Everyone knows in advance. The date and size of every halving have been written in the code since 2009. In a market where many people share the same information, something known ahead of time tends to be priced in early. If everyone expects a rise after the halving, some will buy before it, and the timing keeps moving earlier.
• Price is set by all coins, not just new ones. Existing holders have almost 19.7 million coins they can sell at any second. A single day’s trading among them is many times larger than 450 coins.

The halving changes only a small and shrinking slice of supply. Whether price rises or falls is still driven mainly by demand.

There are only 4 data points

The history people point to: price peaked roughly 12–18 months after the halving three times (late 2013, late 2017 and late 2021), then fell more than 70% from the peak each time. It sounds regular, but count the actual data.

• There have been only 4 halvings and only 3 completed cycles. For a trading system that would be 3 trades, which How many trades before you can trust the results explains is far too few to separate skill from luck.
• The pattern was defined after seeing the data. “12–18 months after the halving” is a window chosen to contain the peaks that happened. Had a peak come at 10 or 22 months, someone would have drawn a window that fits that too. This is hindsight bias, the same thing that makes support lines look perfect once the chart is finished.
• The 2024 cycle already broke the old pattern. Price made a new all-time high in March 2024, before the halving, which had not happened in earlier cycles.

A random walk can produce a “cycle” too

If price had no cycle at all, how often would a pattern like this appear? We simulated a random walk with drift: a price that moves randomly every week, with an expected return (drift) of 40% a year and volatility of 80% a year. Both are assumptions chosen to swing like crypto, not values measured from bitcoin. The model has no halving and nothing happens every 4 years. We generated 10,000 paths of 16 years each and counted the falls of “70% or more from a previous peak before a new peak”.

What we counted	Share of 10,000 paths
At least 1 fall of 70% or more	99.8%
At least 3 falls of 70% or more	26.9%
At least 3, with every pair of peaks 3–5 years apart	2.6%

Read honestly: crashes of 70% or more happen in almost every path, and about 1 in 4 paths crashes hard at least 3 times in 16 years. Boom and bust needs no halving. But peaks spaced a regular 3–5 years apart every time showed up in only about 1 path in 40. We tried 4 other sets of assumptions (drift 30–60%, volatility 60–100%): the share with 3 or more hard crashes ranged from 5.8% to 47%, and the share with peaks 3–5 years apart from 0.6% to 3.2%.

So bitcoin’s 4-year rhythm is not something a random walk produces easily. It is still not evidence that the halving is the cause, because:

• The pattern was picked after seeing the data. People around the world search for patterns across many assets and many time spans, and some will look regular by chance. This is multiple testing.
• Other forces run on similar multi-year rhythms (next section). Three cycles can’t tell them apart.
• Belief itself can create the pattern. If enough people buy because they believe in the cycle, price can follow the belief for a while, and stop as soon as people stop believing or start front-running each other.

Forces that coincided with every cycle

Every strong post-halving rally came with other big events.

• 2017: the ICO boom, where new projects raised money by selling tokens, pulled a wave of new money into crypto
• 2020–2021: major central banks cut rates to near zero and flooded markets with liquidity during Covid, and many risk assets rose together
• 2022: the US Federal Reserve began raising rates quickly from March, followed by the collapses of a major stablecoin and a major exchange, and price crashed the same year
• 2024: spot Bitcoin ETFs began trading in the US in January, three months before the halving

None of these happened because of the halving, but they overlapped in time. Global liquidity and the interest-rate cycle also run over several years. If rates were the main driver, the result would look just like a “halving cycle” even if the halving did nothing. For more on rates and risk assets, read How interest rates move gold and currencies.

Stock-to-flow: a model that looked right, then missed

Stock-to-flow (S2F) is existing supply ÷ new supply per year, a measure of scarcity used for commodities like gold. The higher it is, the scarcer the asset. A model that became very popular in 2019 compared the log of bitcoin’s market value with the log of S2F, got a line that fitted past prices closely, and used it to project prices after halvings.

Bitcoin’s S2F follows directly from the table above: about 8 after 2012, about 24 after 2016, about 56 after 2020 and about 120 after 2024. It at least doubles at every halving.

The problems come in three layers.

1. S2F rises on a fixed schedule. It is almost a clock that ticks more slowly over time. Anything that grows over time will appear to fit it. We took 5,000 simulated random-walk prices (drift 40%, volatility 80%), monthly from mid-2010 to March 2024, and compared them with log S2F from the real schedule. R² (the share the line explains, from 0 to 1) was above 0.5 for 44% of paths and above 0.8 for 12%, even though the simulated prices know nothing about S2F. Statisticians call this a spurious regression: two series that wander independently but look related.
2. There is no demand side. The model looks only at supply, while the earlier section showed new supply has become too small to matter much.
3. It failed out of sample. In 2022 the actual price fell far below the model’s line, and nothing in the model warned of that crash. A model that explains the past well but fails on new data has the same problem as an overfit trading system.

This doesn’t mean scarcity is irrelevant. It means a single-variable model whose input is entirely known in advance should not be the basis for money decisions.

In practice: size the position instead of guessing the timing

What the evidence shows far more clearly than any cycle is that bitcoin has fallen more than 70% from its peak several times: in 2011, 2014–2015, 2018 and 2022, and most other coins fell harder than BTC over the same periods. You can plan around that without knowing when the next one comes.

Maximum crypto share of the portfolio = the portfolio loss you can accept ÷ the fall you assume

Example: a portfolio of 100,000 baht, where you accept the whole portfolio losing at most 10,000 baht (10%) from crypto.

1. Assume an 80% fall: 10% ÷ 80% = 12.5% of the portfolio, or 12,500 baht. If the fall happens, you lose 12,500 × 0.8 = 10,000 baht, exactly the limit you set
2. Assume a 90% fall (allowing for altcoins or an exchange problem): 10% ÷ 90% = 11.1%, or about 11,100 baht
3. After an 80% fall, what’s left needs 1 ÷ 0.2 − 1 = +400% to get back. After a 70% fall it needs +233%

Rules that work whether or not you believe in the cycle:

• Size from an 80–90% fall, not from the belief that it will rise. If the cycle repeats, this size still captures the upside. If it doesn’t, your portfolio survives
• Don’t use leverage because you think you know the halving timing. At 5x leverage, a fall of about 20% or less can trigger liquidation, many times closer than the 70–80% falls that have already happened. Read Perpetual futures and funding rates
• For short-term trades, control risk with a Stop loss and volatility-based size. Use the method in Crypto position sizing and the crypto position size calculator
• Rebalance on a schedule. If a rally makes crypto the bulk of your portfolio, your risk has grown without you noticing. Rebalancing by a rule set in advance doesn’t require guessing where the top is
• Count the risks that aren’t price. An exchange failure or a lost wallet can cost 100% whatever the cycle does. See Crypto custody and stablecoin risk

If you’re coming from Forex, this level of volatility is very different from the major pairs. Read Crypto vs Forex alongside this.

Common mistakes

• Treating 3 cycles as a law. Three occurrences is a trading system with three trades. However regular it looks, it can’t be called a rule
• Believing it because the cycle chart looks neat. Shared charts are usually cropped and scaled to fit the story, like the three hand-picked random paths above
• Using the halving date as an entry or exit. A date everyone has known for years gives the market no new information
• Trusting a model because its R² is high. Two series that grow over time will always look related. Judge it on data the model has never seen
• Sizing from a profit target instead of from the fall you can survive. Ask first whether your life and your portfolio would be fine after an 80% drop. If not, the position is too big