Bitcoin Weeklyビットコイン週報
No change.
Nothing to do this week.
One call at the top. The evidence under it. The judgement at the end.
Value is 37.8 and nothing is flashing danger — the band where deployment has been rewarded.
Scheduled contributions continue either way — that never changes. This is about a lump you already have.
Where value has been — 12 months価値の軌跡
Each day ranked against its own history; the current value percentile is 37.8. It peaked at 72.3 on 2025-10-06 and bottomed at 14.9 on 2026-06-29. Weekly points.
The model context模型
Descriptions of price, not targets.
The power law
expanding log-log fit · refit through 2026-10-06
−39.2%
Bitcoin at $85,716 is 39.2% below the long-run power-law line at $140,934. The line is an imaginary description of where price has been, not a target or a forecast. Its coefficients move as observations arrive.
1592 of 4297 modern-era days were cheaper relative to the line — the 37.0 percentile. This is cheap, but it is not historically extreme.
| cycle trough | vs line |
|---|---|
| 2015-01-14 | −76.3% |
| 2018-12-15 | −51.0% |
| 2020-03-12 | −58.3% |
| 2022-11-21 | −63.9% |
| 2026-06-30 | −56.0% |
| cycle peak | vs line |
|---|---|
| 2013-12-04 | +766.6% |
| 2017-12-16 | +566.1% |
| 2019-06-26 | +57.0% |
| 2021-11-08 | +134.2% |
| 2025-10-06 | +13.6% |
Every trough shown went deeper than today, while peak overshoots have contracted sharply. The history is the hedge against reading “39.2% below” as a promise.
The valuation ladder
spot $85,716 · rank 3 · 5 of 8 rungs above spot
| rung | value | spot vs |
|---|---|---|
| cointime priceharvested · derive to 1.41% | $53,172 | +61.2% |
| realised priceharvested · derive to 0.17% | $53,732 | +59.5% |
| true market meanharvested · derive to 0.25% | $77,387 | +10.8% |
| power lawours | $140,934 | −39.2% |
| vaulted priceharvested · derive to 0.98% | $146,676 | −41.6% |
| MVRV +1.0sdharvested · cannot verify | $154,325 | −44.5% |
| MVRV +1.5sdharvested · cannot verify | $180,801 | −52.6% |
| MVRV +2.0sdharvested · cannot verify | $207,277 | −58.6% |
Five of eight rungs are independently derivable: four harvested base-price models can be rebuilt to the stated tolerances, and the power law is ours. The three MVRV bands are copied from an external chart and cannot be independently verified. A ladder is context; no rung, including the upper bands, is a price objective.
Anything unusual?異常
- Puell multiple moved +12 percentile points in a week
Do anything?行動
- No boundary crossed this week.
An item appears here only when a level crosses a boundary, the engine changes state, the sell guard lights, or an input goes stale. Steady readings produce no items on purpose — a report that always finds something will eventually invent something.
The Panel計器盤
Three indicators, three different jobs.
MVRV
the picture, and the buy side
1.596
43.1 percentile ▲ 2.0
Cost basis of the average coin. Below 1.0 the average holder is underwater. This is the buy-side number; do not average it with anything.
Puell multiple
the danger side
1.341
67.4 percentile ▲ 11.9
Miner revenue against its own trend. The panel's danger marker is meaningful only when it reaches the upper band and the other instruments agree.
price ÷ 200-week avg
the confluence check
1.295
26.2 percentile ▲ was 25.2
Price against its 200-week average. Third leg of the guard; only meaningful when it agrees with the other two.
Sell guard — CLEAR · 0 / 3
Nothing in its top fifth. The danger confluence is not lit. It lights only when all three panel members sit in their top fifth at once. That is a reason to stop adding and check sizing — never a sell trigger.
If you were deploying capital today資本
The distribution you are stepping into
days since 2018 with MVRV in the 33–53 percentile band · n=653
Historical distribution conditional on today's reading. Not a forecast. It says what this zone has looked like, not what happens next.
The rule that survived the tests. Keep buying on schedule regardless of this number, including when the panel is not cheap.
Use the panel to size a lump you already have — faster below 40, slower above 60 — and never to decide whether a scheduled contribution goes in.
Who is under stress売り手
What the week rules out.
| cohort | level | percentile |
|---|---|---|
| long-term holders | 1.398 | 46.2 |
| short-term holders | 1.002 | 55.6 |
| supply in profit | 0.735 | 54.3 |
SOPR below 1.0 means that cohort is realising losses on average. This is the one thing price cannot tell you: who is doing the selling. Use it to kill stories — if someone claims long-term holders are capitulating, this either shows it or it does not.
Week, month, quarter, year四つの地平
| span | BTC | value then | Δ value | Δ MVRV | Δ Puell |
|---|---|---|---|---|---|
| weeksince 2026-09-29 | +2.5% | 36.0 | +1.8 | +2.0 | +11.9 |
| monthsince 2026-09-06 | +6.7% | 34.2 | +3.6 | +4.1 | +14.6 |
| quartersince 2026-07-07 | +35.4% | 17.0 | +20.8 | +23.2 | +48.1 |
| YTDsince 2026-01-01 | −3.4% | 41.3 | −3.5 | +3.2 | +30.7 |
One week is noise. The same direction across four horizons is the difference between drifting and moving. Δ columns are percentile points.
Where we are in the cycle周期
This drawdown so far
Day 365 since the all-time high.
Completed cycles, for scale
peak to trough · the analogue set, all four of it
| cycle | duration | depth |
|---|---|---|
| 2011-06 → 2011-11 | 162d | −93% |
| 2013-12 → 2015-01 | 406d | −85% |
| 2017-12 → 2018-12 | 364d | −84% |
| 2021-11 → 2022-11 | 378d | −77% |
The Analyst Note所見
No change. Nothing to do this week.
The question this week is what “cheap” is measured against. Bitcoin at $85,715.8 is above some of the ladder and below the rest. Value is 37.8, the regime engine remains in RECOVERY with a rule firing, and the sell guard is CLEAR at 0 / 3. The BUY sizing posture continues. No boundary crossed; the disagreement between models does not create an action item.
The discount depends on the denominator
Spot is 10.8% above true market mean and 39.2% below the expanding power-law line. These descriptions answer different questions. The first does not make bitcoin universally expensive; the second does not establish a destination for price. 1592 of 4297 modern-era days were cheaper relative to the power-law line, putting today at its 37.0 percentile. My reading is that cheapness remains usable for sizing, but the large model discount supplies little evidence about when it closes. Averaging the ladder into a target would hide the disagreement that makes it useful.
Recovery has repriced without a fresh decision
Bitcoin gained 35.4% over the quarter, yet remains down 3.4% year to date. Value rose from 36.0 to 37.8 this week. Puell's 11.9-point rise to 67.4 is the anomaly, but the guard remains clear. Long-term-holder SOPR is 1.398, and short-term-holder SOPR is 1.002: neither shows average realised losses. Their percentiles, 46.2 and 55.6, do not identify a dominant source of buying. I would not turn the recovery label into a claim that the market's next buyer has been found.
The branches concern the discount's persistence
Branch A — recovery carries through. Price holds the 2026-06-30 low and the engine progresses toward expansion while the guard stays clear. The power-law discount can remain open during that process; closing it is not required to validate recovery. A return to BEAR_DECLINE would invalidate this continuation thesis, and a guard that ceases to be CLEAR would invalidate its uncomplicated sizing premise.
Branch B — the discount survives a failed recovery. Capitulation is 0 of 8, and the clock has recorded 0 accumulation days since the provisional low. Those absences leave this branch open; they do not prove it. A break below $58,525.05 would reset the clock and confirm that the June low was provisional. A rule-fired BULL_EXPANSION while that low holds would invalidate the failed-recovery reading.
For now, the engine gives a sizing condition and leaves the timing question open. The empty action list is consistent with a rising market: movement within the same condition does not require a new decision. The model gap is a reason to keep the thesis conditional, not to promote either line into a forecast.
Standing caveat. The completed-cycle sample contains four observations. Historical distributions are context, not forecasts. Nothing here is a reason to change a scheduled contribution.
Read this before trusting any number above. Every input is checked for staleness before it is rendered, using the regime engine's own dropout cutoff. The readout reports no stale inputs this week.
The engine's label is RECOVERY, sourced fired — a rule fired and retained the previous state. Its edge is at naming the present, not predicting it. Treat everything here as context for sizing and for killing stories, not as timing. Generated against the regime engine · value percentile is expanding-window, no look-ahead. The full method, every refuted strategy, and the evidence for this panel are in the engine atlas.