A new data-driven Bitcoin cycle framework, built on four full halving eras and stress-tested across multiple backtests, is projecting a deep but historically consistent drawdown into the next bear-market low. The model’s median path places the next cycle bottom in the mid-$30,000s around December 2026, implying a roughly 72.5% decline from the current cycle’s $126,219 bull-market high.
The same modeling approach previously identified the timing windows for the 2021 and 2025 tops, and its author has now released a “v2” version that focuses on three variables: how far Bitcoin falls from a bull-market high, how long it takes to reach the low after a halving, and how strongly it recovers into the following halving.
What the new model is actually forecasting
At the core of the new “Akiba Cycle Model v2” is a simple headline forecast: starting from a cycle peak of $126,219, Bitcoin’s next major low is expected to land near $35,000. In model terms, that corresponds to a 72.5% drawdown from top to bottom.
That drawdown estimate is not a single-point guess. The model runs a 50,000-iteration Monte Carlo simulation to generate a distribution of outcomes around that figure. For the drawdown itself, most simulations cluster tightly between 71.9% and 73.1%. When applied to the $126,219 high, that translates into an implied low in the mid-$30,000s, with a median simulated level around $34,700 and a 10th–90th percentile band from $33,900 to $35,500.
Unlike many price targets in the crypto space, this model is explicit about how confident it is in each leg of the forecast. It treats the cycle low as the most forecastable element and frames the subsequent halving price not as a single bullish or bearish target but as a broad scenario range.
How the halving-cycle framework is built
The v2 framework operates inside a halving-cycle taxonomy that splits Bitcoin’s history into four completed eras and a fifth in progress. For each cycle, it tracks:
Drawdown: percentage decline from the bull-market high to the subsequent cycle low.
Timing: number of days from the halving date to that cycle low.
Recovery: the price multiple from the cycle low into the next halving.
The model is trained on data from the four completed halving cycles:
Cycle 1 (H1, Nov 2012 halving): peak-to-trough drawdown of 94.1%.
Cycle 2 (H2, Jul 2016 halving): 88.2% drawdown.
Cycle 3 (H3, May 2020 halving): 83.7% drawdown.
Cycle 4 (H4, Apr 2024 halving): 77.6% drawdown, with a bull high at $68,998 and a cycle low at $15,474.
Across these four observations, the peak-to-trough damage has consistently eased from cycle to cycle while still remaining extremely deep. That monotonic decay is what allows the model to fit a relatively tight projection for cycle 5’s drawdown centered near 72.5%.
From a methods perspective, the v2 model uses an ensemble of simple functional forms—linear relationships, exponential decay, and average-decrement variants—and evaluates them through walk-forward validation and leave-one-out cross-validation (LOOCV). The Monte Carlo engine then samples across these forms, injecting noise calibrated to past residual errors and “jackknifing” the limited dataset by dropping one cycle at a time to test sensitivity to outliers.
Why the model targets a December 2026 low
Beyond the size of the drawdown, the model is also concerned with when the low occurs in each halving era. Historically, Bitcoin has taken longer to find its floor after each halving:
Cycle 1: 778 days from halving to cycle low.
Cycle 2: 784 days.
Cycle 3: 890 days.
Cycle 4: 923 days.
This is another monotone pattern: the time from halving to low has lengthened each era, though not perfectly smoothly. Using this trend, the model estimates that in the fifth cycle it will take around 980 days from the April 2024 halving to reach the next bottom. That maps to December 2026.
As with price, the model expresses timing as a probability band rather than a fixed date. The simulated 10th–90th percentile window for the next cycle low runs from November 2026 through January 2027. The LOOCV root-mean-square error on timing is 37 days—much wider than the 0.63 percentage-point error on the drawdown—so the author emphasizes that timing is inherently noisier than depth.
A condensed table of cycle history in the model shows how these pieces line up:
Halvings climbed from a price of $12.56 in 2012 to around $65,000 in April 2024.
Bull highs lagged each halving and delivered compressing returns: the cycle 4 high at $68,998 was roughly 632% above the halving price, compared with much larger multiples in earlier years.
Days to the bull-market high have also lengthened, from 363 days after the first halving to 555 days after the third, with the current cycle’s top placed at 537 days post-2024 halving.
These observations underpin the model’s focus on drawdowns and elapsed time rather than on peak upside, where historical gains have already compressed dramatically.
Recovery multiples and the wide range for the next halving price
The most uncertain leg of the framework is the recovery from the cycle low into the next halving. Historically, those low-to-next-halving multiples have been very large but shrinking:
From cycle 1 low into the H2 halving: 347.8x.
From cycle 2 low into the H3 halving: 67.2x.
From cycle 3 low into the H4 halving: 20.8x.
Extrapolating that compression, the model’s central estimate for the fifth cycle’s recovery is about 5.0x from the next cycle low into the halving after H5. But this component rests on only three historical data points and, in the author’s own backtesting, it failed its walk-forward test by underpredicting the previous cycle’s recovery strength.
Because of that, the simulation deliberately uses a wide uncertainty band for the next halving price. When the recovery module is applied to the projected cycle 5 low, the 10th–90th percentile range for the next halving (H5) price runs from $60,000 to $489,000, with a median of $172,000.
The model explicitly treats this as “scenario space,” not a precise forecast. In other words, while it is relatively confident that the next bear-market low will occur in a certain neighborhood of price and time, it is far more cautious about pinning down what Bitcoin could be worth at the halving that follows.
Backtest results: how well did the model do on cycle 4?
To test robustness, the author trained the model on cycles 1 through 3 and asked it to predict the behavior of cycle 4. On the two variables the framework treats as more stable—drawdown and timing—the fit was close:
Drawdown: 78.2% predicted vs. 77.6% observed (a 0.7 percentage-point difference).
Days from halving to low: 929 days predicted vs. 923 observed (a six-day gap).
Implied low price: $15,012 predicted vs. an actual low of $15,474, about a 3% miss.
Where the framework struggled was on the recovery multiple. The backtest projected a 13.0x move from the cycle 3 low into the H4 halving, while the realized multiple was 20.8x—about 38% higher than the model anticipated. That shortfall then propagated into a larger error on the implied H4 halving price.
Those diagnostics shape how the outputs are presented today. The model elevates the cycle-low estimate as its primary, more reliable forecast, and intentionally presents the next-halving price as a wide band rather than a crisp target.
What the distribution says about risk and upside
The Monte Carlo output for the fifth cycle summarizes the joint behavior of drawdown depth, cycle-low timing, and next-halving price. A snapshot of the key metrics shows:
Drawdown from bull high: P10–P90 range from 71.9% to 73.1%, centered at 72.5%.
Cycle low price: P10–P90 range clustered at $34,000–$35,000, with a median of $35,000 in rounded-band terms.
Days from the April 2024 halving to the low: 952 days at the 10th percentile, 980 days at the median, and 1,011 days at the 90th percentile.
Cycle-low calendar window: November 2026 to January 2027, with the central case in December 2026.
From this distribution, the notes derive two explicit conditional probabilities:
A 64.4% chance that the next halving price (H5) exceeds the current cycle’s $126,219 bull high.
A 100% chance, under the model’s structural assumptions, that the next cycle low remains above $20,000—effectively a modeled floor.
Both probabilities are framed as conditional on the design choices: the small-sample calibration of four cycles, the way noise is injected, and the independence assumption that treats drawdown, timing, and recovery as separable random draws even though in real markets these variables can move together.
Key limitations and what this means for Bitcoin investors
The model documentation is direct about its constraints. It is built on only four complete halving cycles, so its statistical tails may understate how extreme future outcomes could be. It does not incorporate regime variables such as spot ETF flows, changes in custody structure, or macro factors like interest rates and liquidity, any of which could alter the behavior of future cycles.
The recovery module is explicitly flagged as the weakest link, given its poor walk-forward performance and the tiny number of past observations. By contrast, the drawdown and timing components have shown smaller historical errors, but they are still based on a small, path-dependent sample of Bitcoin’s history.
For market participants who treat halving-era behavior as a template, the v2 framework formalizes two patterns seen in the data so far: a gradual softening of peak-to-trough drawdowns and a lengthening path from halving to cycle low. It leaves questions around future halving prices and ultimate upside open, framing them as wide scenarios rather than definitive calls.
On its median path, the model suggests that Bitcoin could revisit the mid-$30,000s in late 2026 after a roughly 72.5% retreat from the $126,219 top, before potentially rallying into a broad band of outcomes heading into the fifth halving. The author underscores in the notes that these outputs are not financial advice but a structured way to think about halving-driven cycles using the limited history that exists so far.
Hi, I’m Cary Huang — a tech enthusiast based in Canada. I’ve spent years working with complex production systems and open-source software. Through TechBuddies.io, my team and I share practical engineering insights, curate relevant tech news, and recommend useful tools and products to help developers learn and work more effectively.
