How can the Power Law channel itself become a discipline?
The Power Law page makes a structural argument: bitcoin's price moves inside a channel most of the time, with a conservative floor at 0.42× trend, a central trend, and an upper band at 3× trend that past spikes crossed briefly, the early ones by a wide margin (12× trend in 2013). Across fifteen years and nine orders of magnitude, that channel has held with R² above 0.95. Most readers encounter this as a model to believe or reject.
There is another reading. The channel can be treated not as a forecast but as a protocol. If price oscillates between bounds, then bounded triggers — sell when the historical distribution says you're in a high-percentile zone, rebuy when it says you're in a low-percentile zone — transform the model from a thing-you-believe-in into a thing-you-act-on. The channel becomes a discipline.
This isn't active trading. It isn't market timing. It isn't speculation about where price is going. It is a disciplined response to where price is, relative to a structural channel. The triggers are observable, the actions are bounded, and the framework is articulable in two sentences. Inside a retirement-account wrapper where buys and sells generate no tax events, the only cost of being wrong about a sell-and-rebuy bet is the bitcoin you fail to capture if the rebuy doesn't materialize.
The structural argument
The framework rests on three observations.
First, the channel's asymmetry is real. The floor at 0.42× trend has been a structural backstop — bitcoin has never sustained a daily close below it across multiple regulatory shocks, exchange collapses, and macro crises. The upper band at 3× trend is different in character: a brief-spike zone visited only at cycle euphoria peaks. The two bounds are not symmetric; they are not analogous; and the discipline that lives between them must respect that asymmetry.
Second, the historical distribution of price relative to trend is well-defined. Bitcoin has spent roughly half its history at or below trend, half above it. The 80th percentile of historical position corresponds to a specific multiple of trend; the 20th percentile to another. These are facts about the data, not predictions about the future. The Calculator sets its levels as multiples of trend and prints these historical-frequency anchors beside each one, so the user can see, in real time, how often a given trigger condition has actually existed.
Third, inside a retirement-account wrapper, the rebalancing protocol generates no tax events. Sells trigger no capital gains; rebuys reset no holding-period clocks; the only friction is the bitcoin you fail to capture if a sell-and-rebuy bet doesn't complete. Outside that wrapper, every speculative sale is a taxable event, and the framework collapses into something far less defensible — the Calculator's account-type toggle makes that cost explicit.
These three observations combine into the discipline: Sell a defined fraction of the stack when price reaches a high-percentile zone. Re-buy when it returns to a lower-percentile zone. Accept that some cycles won't trigger; accept that the discipline only works inside a wrapper that absorbs the friction.
How much you sell matters as much as when. At the same triggers (sell on the way up through 1.75× trend, buy back at 0.85×), an IRA that sold the whole stack each time would have ended at 0.61× HODL since 2011, and one that sold half at 0.95× (hypothetical backtests on the recorded prices). Partial sales cut both the damage and the upside. That is the honest trade, and it is why the calculator starts at selling a quarter, not the whole stack.
Audience and entry point
The framework explicitly assumes one structural prerequisite: bitcoin held inside a retirement-account wrapper, or funds that could be moved into one. Self-directed IRAs, Roth conversions, solo-401(k) plans — the mechanism varies; the wrapper is what matters. The Calculator's account toggle exists to make this concrete: selecting Taxable surfaces the tax drag that turns the discipline from defensible into questionable.
The audience explicitly includes users who don't yet hold bitcoin in a retirement account but could move funds into one. The framework itself is part of the argument for why that move might be considered. The page is not gated; anyone can read it. The retirement-account context is surfaced as a structural prerequisite, not as a credentialing check.
What ‘rebalancing’ means here
The conventional finance framing of rebalancing — periodically restoring a target allocation across uncorrelated risk factors — doesn’t describe what most Bitcoiners thinking about this strategy are actually considering. The relevant question isn’t how to rotate between bitcoin and other assets. It’s whether a once-in-a-decade peak is the moment to fund a long-deferred lifestyle purchase — a house, a vehicle, a sabbatical, kids’ education — that years of accumulation finally make possible.
With that framing, selling at the most extreme peaks isn’t aggressive. It’s the moment years of accumulation were leading toward. The ‘rebuy’ is then just deploying whatever savings remain, gradually returned to bitcoin over the following years. The discipline’s default state at any reasonable sell percentile is HODLing — sell triggers sit far above current price most of the time, and the strategy simply tells you to keep holding until they don’t.
This approach uses the Power Law to define its sell and rebuy triggers. If the Power Law isn't a model you find useful for bitcoin's price trajectory, this framework probably isn't either. The Power Law page lays out the model and its evidence — including the out-of-sample validation and the channel asymmetry — for readers who want to assess it directly before using it.
What this is not
- Not active trading. The triggers fire on percentile crossings of a defined channel, not chart patterns or sentiment shifts. Cycles complete on the order of years, not weeks.
- Not market timing speculation. The framework makes no claim about when the next trigger will fire — only about how to act if and when one does. The Calculator tab grounds the framework in fifteen years of actual price data; it does not project forward.
- Not signal-driven decision making. The triggers are observable percentile positions in a structurally-defined channel, not interpretive judgments about cycle tops, sentiment regimes, or technical levels.
What gets settled — and what doesn't
The Calculator tab walks the discipline through fifteen years of actual bitcoin price history. The user sets a sell level and a buy-back level, each as a multiple of trend, and a share to sell; the page finds every point in the record where the rule would have fired and tracks the stack across the cycles that resulted. The output is a record — "starting from one bitcoin in 2011, or in 2014, here is what this rule would have done across the cycles bitcoin actually went through" — not a forecast.
What it does not settle: whether to follow the framework. Whether retirement-wrapped bitcoin makes sense for your situation. Whether your tax bracket warrants the slider you've chosen. Whether the next cycle will materialize, or in what shape. The calculator computes outcomes under your assumptions. The decision is yours.
When would you sell, and when would you buy back?
Both levels are multiples of bitcoin’s power law trend: 1× is the trend, 0.42× the floor. The strategy only pays if both rules work: the sell rule has to fire, and the buy-back rule has to bring you back in lower. How this strategy can fail covers what happens if price stops returning toward the trend.
Next peak if spikes keep shrinking (about 1.1×)If 2024–25 was unusually small (up to about 2.2×)Your buy-back levelYour round trip
- History
- Next cycle
Buy back when price falls to this level.
Where every past cycle bottomed (0.40–0.56×)Your sell levelYour round trip
- History
- Next cycle
More tax and timing settings
Timing
Tax, in a taxable account
What your rules did each time
Hypothetical results, not a forecast: the rules applied to the recorded prices, before trading costs. The early cycles had spikes of 5–14× and a trend growing several times faster than today’s, which made both rules easier to profit from and, in 2013, easier to lose with. Future cycles can play out very differently.
| Cycle · spike · low | Sold | Bought back | Tax paid | Round trip | Your stack | vs HODL | Worth today |
|---|
“Round trip” is the bitcoin held after buying back, as a multiple of what you held before the sale. “Your stack” carries every earlier cycle forward, with this cycle’s change below it. “vs HODL” is the bitcoin you hold after the cycle minus the bitcoin you started with, and “Worth today” is that difference at today’s price, before tax in a taxable account. Prices are the site’s ~12-day series, so a rule fires on the first sample past its level.
Where your rules sold and bought back
Your stack over time
While the rule holds cash, the line counts that cash in bitcoin at each day’s price, so it moves even though nothing was traded.
What your rules might do in future cycles
Nobody knows how big the next spikes will be, so this doesn’t pick one. It replays the four past cycles in future order, resizes every spike to the height along the bottom of the chart, and runs your rules on each replay. Results are in bitcoin only, against HODLing the same stack, by about 2040.
How this is built, and why it flatters the rules
The replays. The four cycles from the 2012 low to the 2026 low, laid end to end from today in four rotations. Every spike is resized so its peak equals the height on the chart. Lows are run two ways: kept where they were (near the 0.42× floor, as in every past cycle) and shrunk in proportion to the spike. The band spans all eight runs, to about 2040.
The projection. Each spike’s height above trend has shrunk cycle after cycle. If it keeps shrinking at any of its past rates, the next spike peaks around 1.1× trend. If 2024–25 (1.19×) was unusually small and the 2017-to-2021 rate resumes, around 2.2×.
Why it flatters the rules. The trend now grows about 37% a year, so a buy back below trend nearly always costs less than the sale did. Past cycles also repeat here on schedule and in the same shape, which real cycles won’t. This illustrates how your rules behave; it doesn’t forecast bitcoin’s price.
These rules assume price keeps returning toward the Power Law trend. The record shows that it has, which is not a guarantee that it will. The rules can fail in two ways.
The Power Law has fit fifteen years of price, across nine orders of magnitude, with R² above 0.95. If price leaves the channel and stays out, the levels stop meaning anything.
If price stays below the floor, the sell rule never fires, and you take the whole fall with no cash set aside. If price stays above the upper band, the sell rule fires and the buy-back level never comes. You hold cash while bitcoin keeps rising, until the deadline buys back higher, or for good if there is no deadline. Neither rule can tell you which case you are in until well after the fact.
Both rules can fire as designed, selling high against trend and buying back low against trend, and still leave you with less bitcoin than if you had held. It has happened, and the table above marks it “Lost bitcoin”.
The levels are multiples of the trend, but you sell and buy back in dollars. When the trend grows fast between the sale and the buy back, a price below trend can still be higher than the sale price. In 2013 the trend was growing about 219% a year; today it grows about 37%, so the same failure now needs a much longer wait. It can still happen.
Related
Figures throughout are in USD. Reading from outside the US? Read why →
a = 1.6×10−17, b = 5.77. See The Channel for the reference view.
The Math
The Calculator tab runs a standing rule through the actual bitcoin price record. This page walks through how the rule’s levels are set, how a trigger fires, how a round trip changes the stack, how tax and lots work, why selling as a spike fades beat selling on the way up, what the buy-back deadline does, why 2013 hurts, how the Looking ahead replays and the next-spike projection work, and what isn’t modelled. The goal is not to convince you the discipline is right for you; it’s to make the calculator’s reasoning auditable.
1. Levels are multiples of trend
The rule’s levels are multiples of the Power Law trend: sell at 2× trend, buy back at 1×. A multiple is a fixed line. It means the same thing next year as it did in 2017.
Each level shows a percentile as context: the share of the record since 2011 that sat at or above it, counted on the site’s ~12-day samples. That is the modern set The Bitcoin Floor and How Much Cash use. Percentiles drift as data arrives. The 80th percentile is 1.75× on the full record, 1.78× since 2011 and 1.60× since 2017, so a rule written as “sell at the 80th percentile” moves its own trigger. That is why the levels are multiples and the percentiles are context.
| Percentile | Full record | Since 2011 | Since 2017 |
|---|
Links shared before this version used percentiles. They still work: ?sell=80&rebuy=50 lands on the since-2011 levels, about 1.78× and 0.87× trend.
2. Trigger detection
The rule has two states: holding the stack, watching for a sale; and holding cash beside what’s left, watching for the buy back.
A sale fires one of two ways. On the way up, it fires when price crosses upward through the sell level: below it on one sample, at or above it on the next. As the spike fades, the rule arms when price reaches the sell level, then fires when price falls back below the fade level. The sale takes your chosen share of the bitcoin held at that moment.
A buy back fires when price crosses downward through the buy-back level, or, if a deadline is set, at market once that many months have passed since the sale. It spends all the cash.
- One open sale at a time. A sale does not re-fire while the rule holds cash, even if price crosses the level again. This stops oscillation around a level from generating trades.
- Record ends mid-cycle. If the record ends while the rule holds cash, the table marks it “still in cash” and values the cash at the last price.
- One event per sample. On a sample where the rule sells, it does not also buy back. The next sample can.
3. Cycle accumulation — worked example
Each completed round trip changes your stack by the ratio of the two prices, applied to the share you sold:
Per-cycle multiplier, before tax:
new_stack = old_stack × (1 − f + f × P_sell / P_rebuy)
where f is the share sold, P_sell the sale price and P_rebuy the buy-back price.
The round trip grows the stack if and only if P_sell > P_rebuy. Against trend this is always true (the sell level sits above the buy-back level). In dollars it is only sometimes true, which is the failure mode in How this strategy can fail on the Calculator tab.
Sell at 1.75× trend and buy back at 0.85×, with the trend standing still: the sold share comes back 1.75 / 0.85 = 2.06×. Selling half, the stack ends at 0.5 + 0.5 × 2.06 = 1.53×. In practice the trend grows between the sale and the buy back, often for years, so the buy back costs more in dollars than that arithmetic suggests. The cycle table on the Calculator tab shows each real round trip, including the ones that ended below 1.00×.
4. Tax, and which purchases are sold first
In an IRA or other tax-advantaged account, nothing is taxed. In a taxable account, each sale is taxed on its gain.
The calculator tracks every purchase as a lot: the first bitcoin at the price on the start date, then each buy back at its own price. A sale draws on lots oldest first (FIFO), the default under US rules, or highest cost first, a form of specific identification that needs records. A lot held under a year is taxed at 32%, the federal ordinary rate, plus NIIT and state if set; otherwise at the long-term rate you choose (0%, 15% or 20%) plus NIIT and state. Tax is charged on the net gain across the lots sold, never below zero, and it comes out of the cash available to buy back. The previous version averaged the cost of all the bitcoin held. Average cost isn’t an allowed method for bitcoin held directly.
Source: IRS Rev. Proc. 2024-28 and the per-wallet cost-basis rules that apply from January 2025. Bitcoin held through spot ETFs is taxed as shares: check with your broker; most default to FIFO.
5. Selling on the way up, or as the spike fades
A rule can sell as price climbs through a level, or wait for the spike to reach a level and sell when it turns back down. Waiting did better. Armed at 2× trend, selling half when price fell back below 2× and buying back at 0.85×, an IRA ended at 1.35× HODL since 2011 and 2.25× since 2014. Selling half on the way up through 1.75× ended at 0.95× and 1.32×. On the way up, the run-up kept going after the sale.
The cost of waiting: a fade rule fires only if the spike reaches the level first. The 2024–25 spike peaked at 1.19× trend, so no 2× rule fired. Bitcoin’s Spikes found the same pattern for funding a bill: selling early on the way up was costly, selling as the spike rolled over was the practical option.
6. The buy-back deadline
A buy-back level set too deep can strand a sale in cash. Selling everything at 2× trend and waiting for 0.5× ended at 0.13× HODL since 2011, because price never came back down far enough.
A deadline buys back at market after a set number of months. Twelve months hurt: it tended to buy back in the middle of a bear market. Twenty-four months was neutral when the buy-back level came anyway, and it rescued the stranded cases: the same rule with a 24-month deadline ended at 0.61×. That is why the starting rules use 24 months. With no deadline, the cycle table marks a sale still waiting as “still in cash”, valued at today’s price.
7. 2013, and how fast the trend grows
Most of the gap between the since-2011 and since-2014 results is one cycle. Rules that sold in March or April 2013, at about $70 to $131, bought back in January 2015 at about $234 to $270. Both trades were right against trend: high, then low. But the trend was then growing about 219% a year, so “back below trend” was a higher price than the sale.
Today the trend grows about 37% a year. The same failure now needs a much longer wait. “Since 2014” starts after that cycle, which is why the hero shows both, and it matches the record How Much Cash uses.
8. Sampling: a trigger fires on the first sample past its level
The record is the site’s price series, one close about every 12 days. A rule fires on the first sample past its level, not at the level itself. In 2011 price went from below 1.5× trend to 2.94× between two samples, so a rule that sells on the way up through 1.5× sold at 2.94×. Daily closes would have sold lower.
Every result on this page carries that sampling error. It was larger in the fast early years than it is now, when price moves less between samples.
9. Looking ahead: how the replays work
The Looking ahead section replays the record; it doesn’t forecast a price. A cycle runs from one low against trend to the next. The 2012, 2015, 2019, 2023 and 2026 lows give four complete cycles, which I lay end to end from the latest sample, in four rotations. For each spike height on the chart, every spike is resized so its peak equals that height: each ratio above 1× is raised to the power ln(height) / ln(that cycle’s own peak). The lows run two ways, unchanged and resized with the same power, so each height gets eight runs. Price on each run is the ratio times the power law trend on that future day, and your rules and tax settings go through the same engine as the record. The breakaway case is a separate path: price climbs from today’s level to 2.3× trend over three years, then swings around 2.2× for the rest of 14 years.
10. The next-spike projection
Each spike’s height above trend has shrunk cycle after cycle: about 14× in 2011, 12× in 2013, 5.4× in 2017, 3.2× in 2021 and 1.19× in 2024–25. I measure the shrinking on the logarithm of each peak. Four past rates are applied to the latest spike: the average rate per cycle from 2011, from 2013 and from 2017 to 2024–25, and the single step from 2017 to 2021. They put the next peak at 1.06× to 1.13× trend, which the sliders mark as about 1.1×. If 2024–25 was unusually small and the 2017-to-2021 step resumes from 2021’s peak, the next spike reaches about 2.23×. Five cycles are too few for odds, so the page shows ranges and counts, not probabilities.
11. The historical record is not a forecast
The Calculator tab shows what the discipline would have done across actual bitcoin price history at the user's chosen settings. Every result there is hypothetical: the rules applied to recorded prices, not trades anyone made, and before exchange fees and spreads. It is arithmetic about the past, not a prediction about the future. Three concrete ways the future will differ from the historical record:
- Volatility compression. Bitcoin’s cycle amplitudes have decreased over time. Cycle 1 peaked near 8× trend in 2011; recent cycles peak closer to 2× trend. The triggers that fired in 2011 or 2017 against extreme price excursions will likely fire less often in future cycles, even at the same levels. The Power Law Theory tab documents this compression in detail.
- Regime change is unmodeled. If the Power Law channel’s shape shifts — if the slope changes, if mean reversion fails, if bitcoin breaks free of the channel and stays out — the levels become unreliable. The historical record does not exhibit this; future records might.
- The strategy can fail without doing anything wrong. The historical record at certain settings shows cycles where both triggers fired correctly (high sell, low rebuy) and yet ended with less bitcoin than before, because the rebuy executed above the prior sell price. Detailed in the How this strategy can fail section on the Calculator tab.
When a rule never fires across the record, the Calculator says so, and the discipline reduces to HODL across that span. The starting rules didn’t fire in the 2024–25 cycle. That is information about the framework, not a bug.
12. Historical channel positions
The chart below shows bitcoin's price/trend ratio across all of PL_DATA. Horizontal reference lines mark the 50th, 70th, 80th and 90th percentiles since 2011. You can see where each excursion sat, how long it lasted, and how often each level was crossed.
Above-80th excursions occurred on roughly 20% of historical days, with average excursion duration around 88 days. Above-90th excursions occurred on roughly 10% of days, with shorter average duration. These are facts about the historical record, not predictions about the future shape of cycles.
13. What I’m not modelling
The calculator focuses tightly on the cycle protocol. The following are deliberately out of scope:
- Varying cash returns. Cash waiting to buy back earns 0% or a flat 4% a year. Real money-market rates ran from 0% to about 5% over the record.
- Tax-loss harvesting. In a taxable account, capital losses elsewhere could offset cycle gains. The backtest doesn't model this; gains are taxed in isolation.
- State short-term rates and thresholds. The state rate is each state’s top long-term rate, applied flat. Washington is the exception, as two options: gains under about $270k a year (0%), and gains above it (7%), applied flat to every sale. Its 9.9% rate above about $1.27M a year, from 2025, isn’t modelled.
- Retirement-account contribution and withdrawal rules. Annual contribution limits, RMDs, Roth conversion mechanics, Backdoor / Mega-Backdoor pathways — none of this is modeled. The page is a cycle protocol calculator, not a retirement-account tax-strategy advisor.
- Multi-account splits. A user with bitcoin partly in a retirement account and partly in a taxable account would benefit from running the discipline only on the retirement portion. This is a future page, not a feature of this one.
- Future-cycle uncertainty. The historical record is one realization. A Monte Carlo treatment with stochastic volatility would surface a distribution of possible outcomes rather than a single line through the past — but it would also hide the fact that the past is the actual evidence. I treat the historical record as the primary thing worth showing, and the volatility-compression note above as the acknowledgment that future cycles will look different.
- Behavioral cost. The discipline requires action at exactly the moments most users find hardest — selling into euphoria, rebuying into fear. The calculator assumes perfect execution of the protocol.
Methodology disclosure
- Power Law coefficients used:
a = 1.6 × 10−¹⁷,b = 5.77(reference Porkopolis values — see The Channel for attribution) PL_DATA: the site’s ~12-day price samples from mid-2010 to the present- Percentiles: share of samples since January 2011 (also shown for the full record and since 2017)
- Backtest: one pass through
PL_DATAfrom the first sample of the start month (2011, 2014 or 2017), 1 BTC at that sample’s price; crossing triggers, one open sale at a time - Cash returns: 0% or 4% a year, compounded between samples
- Trading costs: not modelled. Exchange fees and spreads would lower every round trip’s result
- Tax: in a taxable account, net gain on the lots sold (FIFO or highest cost first); 32% plus NIIT and state on lots held under a year, the chosen long-term rate plus NIIT and state otherwise