How this works, and where it stops working.

What a treasury company's share is worth

A bitcoin treasury company is a pile of bitcoin with claims on it. Take the bitcoin, divide by the shares, and you have bitcoin per share. Multiply by the bitcoin price and you have what each share owns. Subtract the senior claims per share, debt and preferred stock ranking ahead of the common, and you have what each share is worth if the pile were distributed tomorrow.

The market rarely pays exactly that. The ratio between the share price and that value is the premium, and it is the single most important number in this model, because it is the part that has nothing to do with bitcoin. A company can accumulate coin per share while its premium collapses, and the shareholder loses anyway.

Convertible debt makes this circular. A note that is in the money converts, which adds shares and extinguishes its own principal, and both change the price that decided whether it converted. The model resolves that by iterating to a fixed point rather than by ignoring one side of it, which is the common error in both directions.

Why the breakeven is the whole argument

Selling near a top and buying back lower is not free. A ladder selling into a top gets a blend of prices around it, not the peak. A ladder buying back across a trough pays a blend above the low. Measured on the historical cycles, that is 82.2% of the top on the way out and 1.36× the low on the way back.

So the round trip ends with more coin than holding only if

capture ÷ premium × ( top price ÷ trough price ) > 1

which rearranges to a single threshold: the fall has to be deeper than 1 − capture ÷ premium. At the measured values that is 39.6%. Every verdict on the front page is that comparison and nothing else.

The two inputs are also the weakest numbers in the model. The capture rests on three historical tops and the premium on two usable bottoms, because earlier cycles ran before the pattern settled and the current one has not finished. Two observations is not a distribution. Both are dials on the front page for that reason.

The scenarios are a ladder, not a forecast

The default path is one person's reading, labelled as such, and the 6 scenarios stretch it rather than replacing it. Each applies a long-run difference that phases in across the first cycle, so no scenario reprices today: a regime change that had already happened by next year would not be a scenario, it would be a different market.

• Compression continues. Every cycle has topped at a fraction of the last. This is that trend continuing.
• The cycle disappoints. Amplitude keeps compressing and the premium never recovers.
• The cycle repeats. Roughly what the last cycle did, with the maturing already priced in.
• New buyers arrive. Demand from a class of buyer that is not in the market today.
• The cycle runs full size. Amplitude stops compressing and the cycle behaves like an earlier one.
• Mania. The premium over net asset value amplifies rather than the price alone.

A scenario changes what the money is worth and never which years the model trades. That separation is deliberate: if a more optimistic assumption also moved the trade dates, the comparison between scenarios would be measuring two things at once.

One of them is derived rather than authored

Compression continues is the exception, and it is the only number on this ladder that comes out of the record instead of out of my head. Bitcoin's bottom-to-top multiple has fallen at every cycle without exception: 539x, then 112x, then 21.2x, then 7.9x, which is a factor of about 4.1 each time. One more step gives 1.94x, and 1.94 times the 2026 low of $58,525 puts the next top near $114,000. The scenario is calibrated to land there.

It extrapolates exactly one cycle, and I would rather say so than hide it. Apply the same decay again and the multiple comes out at 0.48, which would put a cycle top below the low before it. That is incoherent, so the later tops in this scenario are the same compression applied evenly, the way every other scenario here works, and they are not separately derived. Inventing a rate at which the decay itself decays would have produced a smoother-looking answer and a less honest one.

Worth noticing, if you think the trend is too pessimistic: drawdowns have compressed just as hard, from -94% through -87%, -84% and -78% to -53% in the current bear. The argument that this cycle fell less because the asset has matured is the same argument for the next rise being smaller. Taking the compressed downside and an uncompressed upside is the one combination the record does not support.

What it cannot know

It works in annual closes. A cycle top is a spike lasting days and this model sees only the year. It cannot capture the best price available, and it cannot capture the worst either. The effect is conservative on both ends, which is not the same as being right.

There have been five cycles. The halving effect attenuates each time and the amplitude of each cycle has compressed: the top multiple has fallen at every one. A model fitted to five observations of a maturing process is a weak instrument for the sixth, and this model is exactly that.

The premium is not forecastable. The scenarios move it because it has moved, not because anything here predicts it. It has been above three and below one within a single cycle.

Company figures lag. Coin holdings, share counts and debt change monthly, sometimes weekly. Prices here refresh daily; balance sheets are entered by hand from filings and are stale between them. That is why every field is editable rather than fixed.

A reserve holding anything other than bitcoin grows at a rate you set. Gold in particular is a drift assumption, not a forecast. There is no gold model here, and inventing one would be the least defensible number on the page.

Where the default path came from

The first row is observation, not forecast. 2026 is mostly behind us: the low printed at $58,525 on 30 June. That row used to carry a year-end estimate written before any of it happened, which assumed a 29% recovery off the low and priced the share about 37% above where it was actually trading, before any scenario had been chosen. It now sits at spot, so a recovery has to be argued for in the years that are still forecasts.

The rest is halving-timed, with tops 2029, 2033, 2037 and the bear the year after each, and it sits deliberately below an extrapolated power-law curve rather than on it. It is my own reading and it is in the code rather than hidden: change it and everything downstream changes with it.

If you disagree with it, that disagreement is the most useful thing you can do with this tool. The calculator exists to be argued with.