Buried in the percentile tables of Willy Woo's latest bitcoin study is a number no headline carried: $51,900 in 2030. That is the tenth-percentile path out of a 10,000-run Monte Carlo simulation — the pessimistic decile — and it implies a drawdown beyond 90% from the very 2028 cycle top the same model projects. The figure everyone repeated is $1 million, arriving at a median date of 2033.2. Between those two numbers sits the entire investment case, and the coverage has quietly deleted one of them. I have spent enough time reading distributions submitted to me as evidence to know that the gap between the tenth and the fiftieth percentile, not the target itself, determines whether a holder survives the holding period. In a bear market the question is not how high the median goes. It is how deep the floor drops and how long the floor holds.
Woo is not an anonymous dashboard operator. He has been publishing on-chain research since roughly 2013, and his outputs have shaped how a generation of analysts think about realised capitalisation and holder cost basis. That reputation is precisely why the model deserves the harshest reading available rather than the softest. The study belongs to a family of cycle instruments that have dominated bitcoin forecasting for a decade: the stock-to-flow regression, the logarithmic rainbow band, and the halving-window heuristic commonly called the 500-day rule. Each of them shares one structural assumption — that bitcoin's price path is a stationary function of supply schedule and historical trend, and that the parameters calibrated on past cycles will hold on future ones. The rainbow chart, by the study's own admission, was breached through every band by June 2026. That is not a minor embarrassment for one chart. It is an external falsification event for the entire model family, and any new member of that family enters the room carrying the family's record.
The methodology itself is respectable in isolation. Monte Carlo simulation is a mature statistical instrument; running 10,000 random-walk price paths with a power-law decay term of 0.45 applied to returns is computationally trivial and conceptually clean. The decay parameter encodes an intuition that deserves respect: as market capitalisation grows, each additional percentage point of appreciation requires a non-linearly larger inflow of new capital. Anyone who has modelled marginal liquidity against float understands this. The trouble is that a Monte Carlo engine is an amplifier, not a source of truth. Its output is entirely determined by the input distribution, and the study discloses the random walk and the exponent while leaving the tail specification unstated. Normal, lognormal, and heavy-tailed return distributions produce radically different ninetieth and tenth percentiles from identical median assumptions, and crypto returns are famously heavy-tailed. Without that disclosure the model cannot be reproduced, and a model that cannot be reproduced cannot be falsified. Code compiles, but context reveals the exploit. The exploit here is a missing distribution family.
The sample is the second fracture. The model trains on data from the 2018 bottom to the 2022 bottom — roughly two complete cycles. Fitting a decay exponent on two observations is not estimation; it is interpolation with a decimal point attached. A parameter of 0.45 derived from that window is almost certainly a back-fit to the historical curve rather than a forward law, and the study offers no sensitivity grid. Shift the exponent to 0.40 or 0.50 and the arrival date for $1 million moves by years, which means the headline number has an unquantified error bar wider than the forecast horizon it describes. When a client hands me a valuation model with one free parameter and no sensitivity table, I return it. Not because the parameter is wrong, but because nobody, including the author, knows how wrong it is allowed to be.
The structural break is the third and most serious fracture. Woo states plainly that derivatives, ETFs, and fast liquidation cascades better describe today's market than the spot-driven cycles of the sample window. He is correct, and the admission is honest, and it destroys the model's own foundation. Spot bitcoin ETFs did not exist inside the training data. The marginal buyer in 2024 through 2026 is an allocator responding to macro liquidity and portfolio construction mandates, not a retail cohort responding to a halving countdown. A regime change in the demand function that occurred after the sample closed cannot be captured by parameters fitted before it closed. Formalising that contradiction is not pedantry; it is the difference between extrapolation and fantasy.
Run the arithmetic anyway, because arithmetic keeps everyone honest. From an assumed spot price in the $60,000–$80,000 band, $1 million by mid-2033 requires a 12x to 16x move over roughly seven years — an annualised 45% to 55%. That is aggressive but not absurd for an asset with bitcoin's historical record. The median path, however, is not a straight line: the study's own numbers embed a 43% drawdown along the way, and the tenth percentile bottoms at that $51,900 print in 2030. Meanwhile the twenty-ninth percentile never reaches $1 million at all. Nearly one in three simulated futures fails the headline. Framing that as a "minority pessimistic scenario" is a rhetorical choice, not a statistical one — 29% is a live branch, and any position sized as if the median were a promise is mis-sized by construction.
The halving convergence deserves the same scrutiny, not more awe. The model's projected cycle top sits between $241,000 and $413,000 from December 2028 through March 2029; the 500-day rule points its sell window at 2029; the next halving lands in April 2028. Three independent instruments pointing at the same twelve-month window looks like confirmation until you notice that all three were calibrated on the same four halvings, and that the model's training set literally contains them. Using halvings to predict halving-driven tops is circular reasoning wearing a lab coat. The convergence tells you the instruments agree with each other. It does not tell you they agree with the future.
Here is what the bulls got right, and it is not nothing. Woo voluntarily excluded pre-2018 data on the grounds that 100x gains from a tiny base are trivially easy and therefore uninformative — that is structural-break awareness, and most forecasters never demonstrate it. He repeatedly labels the output a simulation rather than a prediction, which is a rarer form of restraint than the industry rewards. And the model's actual verdict on the near term is bearish relative to consensus: a 2028 top in the low-to-mid $300,000s is a cold shower for anyone holding a $500,000 target for the same window, and a 2033 median is a decade-long sentence for anyone who thought the next cycle delivers seven figures. A study that lowers short-term expectations while its headlines raise them is a study whose author is more careful than its distribution. That distinction matters.
What I want from this model is not a target but three disclosures: the tail specification behind the random walk, a sensitivity grid around 0.45, and Woo's historical hit rate on prior published calls. Give me those and the simulation becomes an instrument. Withhold them and it remains a narrative object with a probability axis bolted on — useful for calibrating the emotional temperature of a market that badly wants a reason, useless for sizing a position. The only line in the study with operational value is the one nobody quoted: a median path that first climbs, then loses 43%, then spends three more years recovering. That is the shape of the trade. Anyone who cannot hold through the trough of that curve should not be modelling the peak of it.
