Price$84,053 6-mo outlook90.4% Cross-check43.2% RegimeMixed
Today’s read · 2026-10-21

Late Bear · valuation 13th percentile · models mixed (90.4% / 43.2%) · turn signals: combined bottom 47 · top 13

Bottom line: late bear (Bear wk 55), valuation at the 13.4th percentile of the trend band after a 32.6% drawdown. The six-month models are mixed (90.4% and 43.2%); the combined bottom signal is below 70 (47).

Where Bitcoin sits in its four-year cycle: valuation, on-chain behaviour and model outlook, refreshed daily.

Price
$84,053
$124,720 ATH · -32.6%
Cycle Stage
Late Bear
Bear wk 55 · past bears lasted 23, 52, 54, 58 wks
Valuation Percentile
13th
0.63x the trend median · -0.78σ vs trend
Days Post-Halving
920
63% of ~4yr cycle
6-Mo Outlook
90.4%
higher than 77% of its past readings
Regime
Mixed
0 of 10 fits · last confirmed Bear

BTC/USD

Daily OHLC · Log scale · Volume (USD)

Up Day Down Day Volume (USD) 200-week MA (cycle floor)

Model Outlook

LEANS LONG
Primary · chance higher in 6 mo
90.4%
Bullish long calls right 72% · base 62%
Cross-check · linear model
43.2%
Bearish long calls right 74% · base 62%
Cycle Turn Signals
Linear · bear end
57
Linear · bull end
24
Tree · bear end
38
Tree · bull end
3
0–100 scores, not probabilities · tick = the 70 strong-signal line · a model reading, not advice
How to read these

6-Month Outlook — primary model. Scored only on weeks the model had never seen · in testing, its long calls were right 72% of the time (price rose in 62% of all weeks).

Higher than 77% of this model’s own past readings (its typical reading is 85%, so it runs high by nature) · it can move about ±10 points on noise alone — read the level, not the week-to-week change

6-Month Outlook — cross-check model. A simpler linear model on the same inputs, tested the same way. It is trained as if rising and falling six-month periods were equally common (about 6 in 10 have been rising), so it reads lower than the primary model for the same view · in testing, its long calls were right 74% of the time (price rose in 62% of all weeks).

Cycle Turn Signals — linear model. A 0–100 signal for whether the current bear or bull phase is ending, using only what was known at the time. A score, not a probability: turns are rare, and the model is trained as if they were not, so readings run high.

Cycle Turn Signals — tree model. The same question asked of a non-linear model that also reads on-chain and macro data, using only what was known at the time. Also a score, not a probability.

Model stance. The primary model puts the chance of a higher price in 6 months at 90%. Above 50% it leans long; below, neutral. A model reading, not advice.

Market Regime (statistical regime model, using only data available at the time): Mixed (last confirmed: Bear) · Fit agreement 0 of 10 fits · Bear wk 55 · past bears lasted 23, 52, 54, 58 wks. Confirms a top 5–10 weeks after it happens and cannot see bottoms. Use it for context, not timing.

Capitulation & Cycle Context

on-chain + derivatives · context, not signals · marker = where each reading sits on its zone scale

Cycle Indexmid-cycle
42
0 bottom · 100 top
MVRVfair value
1.52
→ cap 1.0 · -34%
In Profitmixed
69%
of coins above cost
NUPLoptimism
0.34
<0 = capitulation
Fundingneutral
0.004%
perp, per 8h
DVOLcalm
37
implied vol; 80+ = blow-off
Wk RSIneutral
59
div: none
Trend σfair
-0.78σ
vs long-run trend
Fear/Greedgreed
68
0 fear · 100 greed
Mayernormal
1.16
price ÷ 200-DMA
Puellnormal
1.22
miner rev vs 1-yr avg
Exch Flowoutflow / accum
-83k
net 30d BTC
Guides, not triggers — thresholds drift across cycles (MVRV<3 wasn’t a top in 2021; MVRV<1 may not mark this bottom).
More

Hover any cell for a definition; full history + glossary in the On-chain & Derivatives tab.

Trend Bands · today

2026-10-21

1%$67k
5%$72k
16%$86k
LP$92k
50%$132k
OLS$144k
84%$172k
95%$261k
99%$312k
Spot $84k · 13th percentile · midpoint $132k · outlined = band nearest spot

Projection checkpoints

median · 16–84% · a guide, not a target

DateMedian16–84%
2027-01$141k$92k–$182k
2028-01$192k$126k–$236k
2029-01$258k$170k–$304k
2030-01$342k$225k–$385k
2031-01$446k$295k–$482k
2031-10$548k$363k–$574k

Valuation Ladder

distance from the long-run trend · historical zones, not a top call

now -0.78σ · Hold
−2σ
historically cheap −1σ 0σ
on trend +1σ
past trim zone 10–25% +2σ
past trim zone 50–75% +3σ
σ measures how far price sits above or below its long-run trend. Illustrative, not advice.
More

Replayed over the 2021 top, selling in steps as σ rose would have exited roughly a third to a half of a position near the peak: a stepped approach lightens into strength, it does not call the exact top.

Macro & Liquidity Backdrop

is BTC weakness macro-wide or crypto-specific?

DXYdollar
101.7
higher = BTC headwind
Real Yldrestrictive
2.80%
10y TIPS
VIXcalm
18
equity vol / contagion
M2 YoYexpanding
+5.7%
USD liquidity
Stablesdry powder
$304B
sideline capital
ETF 30dinflow / demand
$+2.1B
spot ETF net flow
Context for whether a move is the whole risk complex or BTC alone — e.g. BTC down hard while VIX stays calm = crypto-idiosyncratic deleveraging, not a macro risk-off.
More

Hover any cell for detail.

Cycle Commentary

AI

Bitcoin is at $84,053 as of 2026-10-21, 32.6% below its October 2025 all-time high of $124,720. It trades above the lower edge of its long-run trend band ($79,918, +5.2% vs spot) and sits at the 13.4th percentile of that band, where the midpoint today is $132,384.

Market regime: mixed. The regime model's 10 fits disagree about the last complete week (9 Distribution, 1 Accumulation), so the regime stays at the last confirmed one, Bear. Bear wk 55; past bears lasted 23, 52, 54, 58 wks.

Bottom line: late bear (Bear wk 55), valuation at the 13.4th percentile of the trend band after a 32.6% drawdown. The six-month models are mixed (90.4% and 43.2%); the combined bottom signal is below 70 (47).

Full commentary

A caution on that reading: the regime model confirms a top five to ten weeks after it happens and has never identified a bottom in real time. It is context, not a timing signal.

Six-month outlook: the primary model puts the chance of a higher price in 26 weeks at 90.4%; the simpler cross-check model says 43.2%. The two models disagree. In testing on weeks the models had never seen, the primary model's long calls were right 72.1% of the time over 333 weeks (price rose in 61.9% of all weeks), the cross-check model's 73.9% (through 2026-04-26).

Cycle turn signals (0-100 scores, not probabilities): the linear model scores a forming top 24 and a forming bottom 57; the tree model scores them 3 and 38. Combined: top 13, bottom 47, so the case for a bottom is stronger than the case for a top, and none reaches the 70 level we treat as a strong signal.

[Generated automatically from today's data. A factual summary, not hand-written analysis and not financial advice.]

Generated 2026-10-21 · Not financial advice.

Track record · long calls right

primary model, weeks it had never seen · design chosen on this period (see Methodology)

2018: 30% of 27 long calls right'182019: 61% of 23 long calls right'192020: 100% of 52 long calls right'202021: 56% of 39 long calls right'212022: 65% of 34 long calls right'222023: 91% of 53 long calls right'232024: 96% of 47 long calls right'242025: 46% of 41 long calls right'252026: 59% of 17 long calls right'26base 62%
right 72% of 333 long calls vs 62% base · ~16 independent six-month windows · through 2026-04-26
Year-by-year table
YearLong callsRight
20182730%
20192361%
202052100%
20213956%
20223465%
20235391%
20244796%
20254146%
20261759%
Price Bands Bitcoin’s price has grown along a remarkably steady long-run curve.
About this view
This chart fits that curve to the full price history, refits it every day, and draws bands around it showing how far above or below trend price has historically traded. The grey line is the midpoint: half of history sits above it, half below. Blue bands are below trend (the deeper the blue, the cheaper relative to history) and orange bands are above (the deeper the orange, the richer). The infobar names the zone price is in today.

The lower pane shows the same bands as a multiple of the midpoint, so 1.0× is on trend, 0.5× is half of trend and 2× is double. It is the easier place to read where price sits. The small ladder on the right is a snapshot of today’s bands. Notice that the upper bands narrow over time: each cycle top has been less extreme than the last (the 2021 top reached about 4× the midpoint, the 2025 top about 1.4×).

The dashed section is a 5-year projection of the bands, with yearly checkpoints above the chart. Treat it as scaffolding, not a price target. Small, defensible changes in how the curve is fitted move the implied 2029 level by almost 4×, and the projection stops at the point where the model’s own bands would cross, because past there it no longer makes sense.

The lavender line is a second estimate of the same trend, fitted to every daily close instead of starting 900 days in. The two are drawn together so you can see that “the trend” is an estimate: the infobar shows the gap between them, and that gap comes from the choice of start date, not from anything Bitcoin did. One detail: the percentile on this chart ranks today’s price among the fitted bands, while the Distribution Planner uses a stricter version that only ever looks backwards in time. Both are valid and they can differ; the planner’s is the one that triggers its steps.

The teal cycle-wave overlay (off by default; click it in the legend) adds 2 repeating waves to the second trend line. Each wave is longer than the one before (by factors of 2.09 and 1.31), which is one published way of describing why Bitcoin’s cycles stretch over time. We rebuilt it from our own data, then froze it on 2026-08-23 and do not refit it, so its dated calls (next bottom 2026-10-26, next top 2030-01-23) are an open test that anyone can check later. Price sits -8.9% from it today. Two reasons not to trade on it: the predicted bottom moves by years depending on how many waves are included (anywhere from late 2026 to 2039), and when we froze it at earlier dates and replayed 2018–2024 it placed 0 of 4 tops within a year and 1 of 3 bottoms within six months. It describes the past well. It is not an input to any model here and triggers nothing.

BTC Price — Long-Run Trend Bands

Daily price on a log scale inside its long-run trend bands · blue = below trend, orange = above · dashed = 5-year projection, a guide not a target

BTC/USD trend midpoint (median) second trend estimate (all data) cycle-wave overlay (frozen 2026-08-23) 1–5% · cheapest5–16% · deep value16–50% · below trend50–84% · above trend84–95% · extended95–99% · richest projection
Projection checkpoints · median · 16–84%2027-01$141k$92k–$182k2028-01$192k$126k–$236k2029-01$258k$170k–$304k2030-01$342k$225k–$385k2031-01$446k$295k–$482k2031-10$548k$363k–$574k
today · 2026-10-21 1% $67k 5% $72k16% $86k LP $92k50% $132kOLS $144k84% $172k95% $261k99% $312k$84k · 13th
Cycle Map (hindsight) A map of past cycles, drawn with the benefit of knowing how each one ended.
About this view
The top pane shows weekly price inside its long-run trend channel with the likelihood that a cycle top or bottom was forming; the bottom pane shows the market regime (bear, accumulation, markup, distribution). Solid lines are smoothed, dashed lines are the raw weekly values. Because it is fitted on the whole history, use it for orientation only. For readings made without hindsight, see Top & Bottom Signals.

Cycle Map (hindsight)

Fitted on the full history, so it knows how each cycle ended · top and bottom likelihood · long-run trend channel · market regime

BTC P(Bull End) P(Bear End) Accumulation Markup Distribution Bear
6-Month Outlook Two models estimate the chance that Bitcoin is higher 26 weeks from now.
About this view
The primary model is a gradient-boosted tree ensemble; the cross-check is a simpler linear model fed the same seven inputs. Both are retrained every week, and every point on these lines uses only data from before its own date: it is what today's models would have said at that point. Today's models were designed with this history in view, though, so the line flatters them somewhat; the Methodology page says how much. When the two agree the read is firmer; when they split, treat it as mixed. The dashed grey line is the primary model fitted with hindsight, shown only as a reference. Track record and limits are on the Methodology page.

6-Month Outlook

Chance BTC is higher 26 weeks ahead · two models, each scored only on weeks it had never seen

BTC Price Primary model Cross-check model fitted with hindsight (reference only)
Top & Bottom Signals Four panes: price, then three attempts to spot a cycle turn.
About this view
The tree model and the linear model each give a 0–100 signal that a top or a bottom is forming; these are scores that rank how turn-like a week looks, not probabilities. The proximity score estimates how many weeks remain until the next turn and rescales that to 0–100. Every line uses only what was known at its date, including when a past turn could first be called: a top counts as known once price had fallen 40% from it, a bottom once price had risen two-thirds from it, so each line starts only after the first turn it could learn from. Since 2016, the period with complete data, Bitcoin has had three tops and two bottoms, so these models are exploratory: read them as supporting evidence for the 6-month outlook, never on their own. The proximity scores are shown here for reference only and are not used anywhere else on the site. The bottom-proximity score in particular reads the calendar more than the market: both bottoms it learned from came about a year after their tops, so once a bear market passes its first year without a bottom, the score drops sharply as if the bottom had already happened. Dashed grey = fitted with hindsight, reference only.

Top & Bottom Signals

Three models' reads on whether a cycle top or bottom is near. Each uses only data available at the time, and counts a past top or bottom only once it could have been called (price 40% away from it)

BTC Price Tree model: top Tree model: bottom Linear model: top Linear model: bottom Top proximity Bottom proximity fitted with hindsight (reference only)
Realized Price The aggregate cost basis of all BTC holders, computed as Price / MVRV ratio.
About this view
When spot price trades below Realized Price (P/RP < 1.0x) the average holder is underwater — historically a reliable cycle-floor zone. Above 2.5x P/RP signals euphoric distribution. Single pane, log scale, 4-year default zoom.

Realized Price

Price vs aggregate holder cost basis · Daily · Log scale · RP = Price / MVRV

BTC Price Realized Price
On-chain & Derivatives Seven indicators for reading where we are in the cycle.
About this view
They are shown for a human to read and none of them feeds the models: each was tested as a model input and added nothing the models did not already have. MVRV (market vs realized value; 1.0 = aggregate cost basis), NUPL (net unrealized P/L; <0 capitulation → >0.75 euphoria), Funding (perp leverage lean, %/8h), DVOL (implied vol; 80+ = blow-off), Weekly RSI(14) (the one signal that flagged the 2021 top in real time, via bearish divergence), Pi-Cycle (111d÷2×350d SMA; a cross above 1.0 historically marked tops within a day — but it failed the 2021 Nov double-top, shown here as a caution not a promise), and Distance from trend (σ) (how far price sits from its long-run trend, where every past point is calculated using only the data that existed at that date — the honest history; the Price Bands chart, refitted on everything, flatters the past). Dashed reference lines mark the classic zones — but those thresholds drift across cycles, so read them as guides, not triggers. Coverage: funding from 2019, DVOL from 2021, MVRV & Pi-Cycle from 2013; NUPL, weekly RSI and distance from trend cover the full history.
What do these mean?
MVRV — Market Value to Realized Value
Price ÷ the average cost basis of every coin (the “realized” price). >1 = holders in profit, <1 = underwater. ~1.0 = value/capitulation zone, 3+ = euphoria. The top level drifts down each cycle (4.7 in 2017 → 2.7 in 2021), so it’s a guide, not a fixed trigger.
NUPL — Net Unrealized Profit/Loss
The whole supply’s paper profit as a share of market cap. Bands: <0 capitulation · 0–0.25 hope/fear · 0.25–0.5 optimism · 0.5–0.75 belief · >0.75 euphoria. Derived as 1 − 1/MVRV (the exact identity) from CoinMetrics MVRV, so it spans every cycle — the BGeometrics NUPL feed only starts 2022. Matches that feed at 0.9999 corr over the overlap. Same signal as MVRV; same cross-cycle drift.
Funding — perpetual funding rate (per 8h)
The recurring payment between perp longs and shorts. Positive = longs pay (bull leverage crowded); negative = shorts pay (bears crowded / longs flushed). Deep-negative near washouts, persistently high near froth. A positioning gauge, not valuation.
DVOL — Deribit implied volatility
The options market’s expected 30-day BTC volatility — a “VIX for Bitcoin.” <40 calm · 40–60 elevated · 60–80 fear · 80+ a vol blow-off that usually accompanies capitulation. Measures fear, not direction.
Weekly RSI(14)
Momentum oscillator on weekly closes. >70 overbought, <30 oversold. The value is divergence: price higher-high while RSI lower-high = fading momentum — the one signal that flagged the Nov-2021 top in real time. Discretionary; can be early or late.
Power-law σ (and the causal pane)
Where price sits vs BTC’s long-run power-law trend, in standard deviations. <−1σ deep value, >+2σ extreme. The gauge uses the production channel (fine for today); the tab pane uses a causal expanding-window fit (refit only on past data) — the honest version, because the live channel is refit on all history and over-states how extreme past dates looked.
Pi-Cycle Top — 111d MA ÷ 2×350d MA
A cross above 1.0 has marked tops within ~1 day (2013 / 2017 / Apr-2021). But it failed the Nov-2021 top — the summer −50% dip reset the 350-DMA so high the slower second top never crossed. Famous but fallible; shown here as a caution.
Trim / Add Ladder
Not an indicator — a decision aid. Maps current power-law σ onto a pre-committed plan: accumulate below −1σ, hold in the middle, trim in steps (10 → 25 → 50 → 75%) above +0.5σ. You can’t call the exact top, but a rules-based ladder set before euphoria lightens you into strength. Illustrative, not advice.

MVRV

Market value ÷ realized value · cap 1.0 / euphoria 3.0

MVRVBTC price (L)

NUPL

Net Unrealized P/L · <0 capitulation → >0.75 euphoria

NUPLBTC price (L)

Funding rate

Perp funding %/8h · <0 = shorts pay (longs flushed)

FundingBTC price (L)

DVOL

Deribit implied volatility · 80+ = vol blow-off / capitulation

DVOLBTC price (L)

Weekly RSI(14)

Momentum · 70 overbought / 30 oversold · watch divergence

Wk RSIBTC price (L)

Pi-Cycle Top

111d ÷ 2×350d MA · cross >1.0 = top (failed the 2021 double-top)

Pi-CycleBTC price (L)

Distance from trend (σ)

Each point uses only data available at that date · <−1 deep value

σ vs trendBTC price (L)
Sentiment, Miners & Floor Four more cycle lenses (context, not model inputs).
About this view
Fear & Greed (crowd sentiment; extreme fear clusters near bottoms), Mayer Multiple (price ÷ 200-day MA; <1 below trend, >2.4 overheated), Puell Multiple (miner revenue vs its 1-year average; <0.5 = miner capitulation), and Hash Ribbons (30-day ÷ 60-day hashrate; a cross back above 1.0 marks miner recovery). The macro/liquidity backdrop (DXY, real yield, VIX, M2, stablecoins) lives in the strip on the landing. Same caveat as everything here: guides, not triggers.
What do these mean?
Fear & Greed Index
Alternative.me’s 0–100 crowd-sentiment gauge (volatility, momentum, social, surveys, dominance). <25 extreme fear has historically clustered near bottoms; >75 extreme greed near tops. Sentiment, not valuation.
Mayer Multiple — price ÷ 200-day MA
How far price sits above/below its 200-day moving average. <1.0 = below the 200-DMA (historically cheap), ~2.4+ = historically overheated. A simple, widely-watched valuation oscillator that complements power-law σ.
Puell Multiple — miner revenue stress
Daily miner revenue (USD) ÷ its 365-day average. <0.5 = miner capitulation (revenue far below normal — clusters near cycle bottoms); >4 = miner euphoria near tops. A different actor (miners) than holder cost-basis.
Hash Ribbons — 30-day ÷ 60-day hashrate
When the 30-day hashrate MA falls below the 60-day, miners are capitulating (switching off); the cross back above 1.0 has historically marked recoveries / good accumulation windows. A miner-health signal.
Macro & Liquidity backdrop (landing strip)
DXY (dollar), 10y real yield, VIX (equity vol), M2 YoY (USD liquidity), and stablecoin supply (dry powder). Together they answer “is BTC weakness the whole risk complex or crypto alone?” — e.g. BTC down with VIX calm = crypto-idiosyncratic, not macro contagion.

Fear & Greed

Crowd sentiment 0–100 · <25 extreme fear / >75 extreme greed

Fear/GreedBTC price (L)

Mayer Multiple

Price ÷ 200-day MA · <1 below trend / >2.4 overheated

MayerBTC price (L)

Puell Multiple

Miner revenue vs 1-yr avg · <0.5 miner capitulation / >4 euphoria

PuellBTC price (L)

Hash Ribbons

30d ÷ 60d hashrate · cross back >1.0 = miner recovery

HashRibbonBTC price (L)
Holder Cohorts (by coin age) Cohorts split by how long coins have sat still (NOT wallet size — that’s the next tab).
About this view
LTH (long-term holders) = coins unmoved >155 days; STH = the rest. % supply held >1 year rises through bears (accumulation) and falls into tops (distribution). Realized price by cohort = each group’s aggregate cost basis — the whole market (Realized), long-term holders (a slow support floor), and short-term holders (recent buyers, the line that snaps to a top and breaks first in a downturn); toggle each line on the chart. Dormancy = average age of coins moving each day; spikes = old coins waking up = distribution. All computed from our own Bitcoin node (the long-term-holder split reads too high before 2017). Guides, not triggers.
What do these mean?
LTH / STH supply
Long-Term Holder supply = coins last moved >155 days ago (the classic "smart money" threshold); Short-Term Holder = everything younger. LTH supply rising = accumulation/conviction; falling = old coins distributing (often near tops).
% supply held > 1 year
Share of circulating supply that hasn’t moved in over a year (from HODL waves). Climbs during bears as holders wait, peaks near cycle bottoms, and rolls over as coins are spent into a bull top. A patience/illiquidity gauge.
Realized price by cohort
Each cohort’s aggregate cost basis (value-weighted last-move price), all from a single pass over every coin on the blockchain, so aggregate = LTH + STH exactly. Realized = the whole market’s basis (= market cap ÷ MVRV). LTH cost = long-term holders’ basis, a slow-moving support that spot falling below has marked deep capitulation. STH cost = short-term (<155d) holders’ basis, which tracks near recent price — spot losing it often front-runs a trend break, regaining it a recovery. The card badge tracks the STH–LTH crossover: when STH cost falls below LTH cost, recent buyers are underwater relative to old holders — a late-stage capitulation signature that has clustered near cycle bottoms. Context only (it was tested and does not add to the bottom model — the aggregate already carries it). Click a legend item to toggle any line.
Dormancy
Average age (in days) of the coins that moved each day, 30-day smoothed. Low = mostly young coins changing hands; spikes = long-dormant coins waking up, i.e. old supply distributing.

% Supply Held > 1 Year

Age-based HODL signal · rises in bears, rolls over into tops

>1y heldBTC price (L)

LTH vs STH Supply

Long-term (>155d) vs short-term holder supply

LTHSTHBTC price (L)

Realized Price by Cohort vs Spot STH 1.48× LTH cost

Holder cost basis — aggregate, long-term (>155d) & short-term · click a legend item to toggle · log

PriceRealizedLTH costSTH cost

Dormancy

Avg age of coins moving (30d) · spikes = old supply distributing

DormancyBTC price (L)
Supply by Wallet Size Circulating supply split by address balance — shrimp (<1) · crab (1–10) · fish (10–100) · dolphin (100–1k) · shark (1k–10k) · whale (>10k BTC).
About this view
Click a band in the legend to show/hide it and isolate cohorts. The question this answers: are large holders accumulating while retail capitulates (or vice-versa)? Computed from our own Bitcoin node (no free data service offers this), as of its last sync. Caveat: early-era P2PK / non-standard outputs have no decodable address (~8% of supply, mostly pre-2014), so totals run below true circulating supply.
What do these mean?
Wallet-size bands (address balance)
Each on-chain address is bucketed by its BTC balance at each weekly snapshot, and the supply it holds is summed per band: shrimp <1 · crab 1–10 · fish 10–100 · dolphin 100–1,000 · shark 1,000–10,000 · whale >10,000. Rising large-band supply into weakness = accumulation by big holders.
How it’s built
From our own Bitcoin node: per-address running balance over time → size band → per-band weekly supply. Address ≠ entity (one exchange = many addresses; one person = many addresses), so read it as on-chain balance distribution, not literal headcount.
Caveat
Outputs with no decodable address (early P2PK incl. Satoshi-era coins, bare multisig, non-standard) can’t be balance-attributed, so band totals are ~8% below true circulating supply — concentrated in the early era. Recent-cycle coverage is near-complete.

Supply by Wallet Size

Address-balance cohorts · click a band to filter · computed from our own Bitcoin node

Shrimp <1 Crab 1–10 Fish 10–100 Dolphin 100–1k Shark 1k–10k Whale >10k
Exchange Flows The supply-side lens (CoinMetrics, genesis+).
About this view
Net Flow: BTC moving onto exchanges (+, sell-side pressure) vs off (−, accumulation / cold storage), per week. Reserve Trend: cumulative net flow ≈ relative balance held on exchanges — the multi-year decline (coins leaving for self-custody) is the structural supply-squeeze backdrop; upticks = supply returning to sell. Exchange-wallet tagging improves over time so absolute levels are approximate — the sign and trend are the signal. Context, not a model input.
What do these mean?
Exchange Net Flow
Weekly BTC inflow minus outflow for exchange-tagged wallets. Positive = net inflow (coins arriving to be sold — sell-side pressure, often spikes at capitulation); negative = net outflow (coins leaving to self-custody — accumulation / reduced sell-side supply). Noisy week-to-week; read the trend.
Exchange Reserve Trend
Cumulative net flow = a relative proxy for the balance held on exchanges. A sustained fall means supply is leaving exchanges (illiquidity / squeeze backdrop, broadly bullish); a rise means supply is building up ready to sell. The absolute level is arbitrary (anchored at the series start); the slope is what matters.

Exchange Net Flow

Weekly net BTC to(+)/from(−) exchanges · >0 inflow (sell pressure) / <0 outflow (accumulation)

Net flowBTC price (L)

Exchange Reserve Trend

Cumulative net flow ≈ relative balance on exchanges · falling = supply leaving exchanges

Reserve trendBTC price (L)
Spot ETF Flows Real-money institutional demand — daily net creations/redemptions across all US spot Bitcoin ETFs (Farside Investors, USD).
About this view
A genuinely new signal in the ETF era: the marginal buyer changed. Net inflows = ETFs absorbing supply (demand); outflows = distribution. The cumulative line is total net institutional capital since the Jan-2024 launch. Last ~30 days: $+2.1B (inflows); $58B cumulative since launch. Read the trend, not single days. Context, not a model input.
What do these mean?
ETF Net Flow (weekly)
Weekly sum of daily net creations minus redemptions across all US spot BTC ETFs, in USD. Positive = net new money buying BTC through ETFs (demand); negative = money leaving. Price-independent (it's fund-share flows, not AUM), so unlike "ETF holdings $" it isn't distorted by BTC's price moving. Spiky day-to-day; the trend is the signal.
Cumulative ETF Flow
Running total of all net flows since the Jan-2024 launch ($B). A steadily rising line = sustained institutional accumulation absorbing supply; a flattening/falling line = demand stalling or reversing. One of the clearest structural-demand signals unique to this cycle.

Spot ETF Net Flow (weekly)

Net creations−redemptions across US spot BTC ETFs · >0 inflow (demand) / <0 outflow

Net flow $MBTC price (L)

Cumulative ETF Flow

Total net institutional capital since Jan-2024 launch ($B) · slope = demand trend

Cumulative $BBTC price (L)
● cost-basis snapshot: 2026-10-18 · 0w old · recomputed weekly from the full blockchain
Supply in Profit / Loss The share of circulating supply whose on-chain cost basis (the price when each coin last moved) sits below the current price.
About this view
Computed coin by coin from our own Bitcoin node (the realized-price distribution), so it is a coin-headcount view — distinct from NUPL, which is value-weighted. The empirical reads: <50% = capitulation / bottom (most of the market underwater, historically a buy zone), >95% = euphoria / top-risk (almost everyone in profit → fragile). Currently 69.4% of supply is in profit (recovering). Early Satoshi-era P2PK coins are excluded (a small, always-in-profit dormant bias). Context, not a model input.
What do these mean?
% Supply in Profit
Of all spendable supply, the fraction whose cost basis (price at the coin’s last on-chain move) is below the current price. Falls in bear markets as recent buyers go underwater; <50% has marked every major bottom and >95% every blow-off top — thresholds far more stable across cycles than MVRV’s. Recomputed daily against the live price from the latest cost-basis distribution; the weekly history comes from a full per-UTXO chain walk.
Supply in Profit vs Loss (BTC)
The same split in absolute BTC: coins above water (profit) vs underwater (loss). The loss line climbing into the millions is the supply an eventual capitulation would have to flush; it collapses toward zero as price makes new highs and nearly all coins return to profit.

% Supply in Profit

Share of supply above its on-chain cost basis · <50 capitulation / >95 euphoria

% in profitBTC price (L)

Supply in Profit vs Loss (BTC)

Coins above water (profit) vs underwater (loss)

In profitIn loss
● history accumulates forward from 2026-06-09 · no earlier history is available from the source, so this chart grows a day at a time
Options Skew (25Δ Risk Reversal) What options traders pay for downside vs upside: implied vol of 25-delta puts minus 25-delta calls on the ~30-day Deribit expiry.
About this view
Positive = puts richer = demand for crash protection (fear — clusters near capitulation/bottoms); negative = calls bid = upside chasing (greed — clusters near tops). The shape of the vol surface, distinct from DVOL (the level). Currently +3.1 vol pts (mild put skew). Context, not a model input.
What does this mean?
25Δ Risk Reversal (put IV − call IV)
Take the ~30-day options equally far out-of-the-money on each side (25-delta) and subtract the call's implied volatility from the put's. Positive = downside protection costs more than upside calls — traders paying up to hedge a fall (fear). Negative = calls are bid — traders chasing upside (greed). Tops typically show flat/negative skew (nobody fears a drop); bottoms show elevated positive skew (everyone's hedged). A positioning read the spot-only indicators can't see.

Options 25Δ Skew (put IV − call IV)

Downside-protection demand · >0 = fear (puts bid) / <0 = greed (calls bid)

25Δ skew
Cycle Navigation Objective bull/bear cycle position (context, not model inputs).
About this view
The Composite Cycle Index averages the percentile ranks of 9 cycle metrics (MVRV, NUPL, Puell, Pi-Cycle, Fear & Greed, RSI, Mayer, 2Y-MA, power-law σ) into a single 0–100 reading — <15 = cycle-bottom zone, >85 = cycle-top zone. Alongside: MVRV Z-Score (std-normalized market-vs-realized value), SSR (market cap ÷ stablecoin supply — low = more sideline buying power), and the 2-Year MA Multiplier (price ÷ 730-day MA). All normalized over full history, so read them relative — every one of these under-read the 2021 top, so they are guides, not fixed triggers.
What do these mean?
Composite Cycle Index (0–100)
The mean of the percentile ranks of 9 cycle-position metrics (each scored 0–1 vs its own full history, then averaged ×100). All inputs read high at tops / low at bottoms, so the average is directly "where in the cycle." <15 = cycle-bottom zone, >85 = cycle-top zone. Averaging cancels any single metric's noise — the one objective bull/bear navigator. (Full-history normalization, like CBBI; read relative, not as fixed lines.)
MVRV Z-Score
(Market cap − realized cap) ÷ the standard deviation of market cap. The classic cycle top/bottom oscillator: deeply negative = price below aggregate cost basis (capitulation), high = euphoric overvaluation. Std-normalization is provider-dependent, so the absolute scale here is its own — read the shape and extremes.
Stablecoin Supply Ratio (SSR)
BTC market cap ÷ total stablecoin supply. A "buying power" gauge: low SSR = lots of stablecoins relative to BTC = more dry powder on the sidelines (historically bullish); high SSR = sideline capital exhausted. From 2018 (stablecoins barely existed before).
2-Year MA Multiplier
Price ÷ its 730-day moving average. A long-horizon cousin of the Mayer Multiple: <1 = historically cheap (below the 2-year mean), >~5 = historically overheated. One of the simplest, most-watched cycle top/bottom references.

Composite Cycle Index

Mean percentile-rank of 9 cycle metrics · <15 bottom zone / >85 top zone · the objective navigator

Cycle IndexBTC price (L)

MVRV Z-Score

Market vs realized value, std-normalized · <0 below cost basis / high = top

MVRV-ZBTC price (L)

Stablecoin Supply Ratio (SSR)

Market cap ÷ stablecoin supply · low = more dry powder (bullish)

SSRBTC price (L)

2-Year MA Multiplier

Price ÷ 730-day MA · <1 historically cheap / >5 historically hot

2Y-MA multBTC price (L)
Where we are right now updated from today's readings · a plan written now for the post-2028-halving bull, top ~2029
Trend percentile 29th · top signal 19th. The next step in the plan is Leg 1 (First chips), which needs the trend percentile to reach the 62nd (now 29th, 33 points away). Plan status: hold, nothing is due.
Two signals drive it. Legs 1–2 use the trend percentile (29th; it rises steadily through a bull market). Legs 3–4 use the top signal (19th percentile of its own history; today the linear model scores 24 and the tree model 3 out of 100, average 13).
Late Bear. Bear wk 55 | past bears lasted 23, 52, 54, 58 wks · Everything below is dormant by design until the next cycle's top approaches.
Distribution Planner the author's own plan, shown as a worked example · rules, not price targets
This is the plan the author of this site has written for selling part of a Bitcoin position across the next cycle top. It was written now, in the bear market, and it runs on rules so that euphoria cannot rewrite it later. It is published as a worked example of how to build such a plan. It is not a recommendation, and your own situation will differ.

Why rules instead of price levels: the classic sell signals have faded. Each cycle top has been less extreme than the last (the Mayer Multiple at the top went 8.2→3.7→2.8→1.4, and the 2025 top printed almost exactly on the long-run trend). So each step triggers on a percentile rank that recalibrates every cycle, never on a fixed level. Two signals are used: the early steps follow the trend percentile, which climbs steadily through a bull market, and the later steps follow the models' top signal, which only spikes near a top. The plan is back-weighted on purpose: only about 28% of the tradeable amount is sold early, and most is kept for the top zone and sold across it.
Position calculator try the plan on any position size · runs entirely in your browser: nothing is stored or sent, and it resets to 0 on refresh · illustrative
Tradeable sleeve
enter a position → BTC per leg
In this plan the lowest cost-basis coins stay in the core, where the largest unrealised gain is never triggered, and the highest-basis lots are sold first so each sale realises the smallest gain. Tax treatment varies by country and by person; this is not tax advice.
The ladder status updates with each daily refresh
Leg Primary trigger Confirmation Share sold / BTC Status
Leg 1 · First chips
A first small sale as price stretches above its long-run trend.
 trend ≥ 62nd pctile price above the prior all-time high 11%
DORMANT
Leg 2 · Froth
A larger sale as valuation stretches further on the way up.
 trend ≥ 68th pctile price well above trend / weekly RSI hot 17%
DORMANT
Leg 3 · Top zone
The models' top signal fires. The heaviest fixed step.
 top signal ≥ 90th pctile trend percentile ≥ 82nd 22%
DORMANT
Leg 4 · Model screams top
Only the most extreme tops. It will feel early; the plan sells anyway.
 top signal ≥ 98th pctile regime model still reads bull 22%
DORMANT
Leg 5 · Trailing tail
Catches any overshoot without trying to pick the exact top.
 trailing (see rule) arms after Leg 3, then exits on the first weekly close below the 20-week average, or when the regime model flips to bear, or when the top signal reaches its 98th percentile 28%
DORMANT
Price context (soft): two geometric estimates of the ~2029 top that disagree: a conservative ~$166k floor and a higher Fibonacci zone measured up from this bear market's low (see below). They size the steps; the percentile rules decide when to act. How it would have done (2023–25 bull, top $124,720): the plan would have sold at about $69k and $98k on the way up, sold the top steps near $119k, and trailed out of the last portion at about $109k, for a weighted average exit of about 86% of the top, well spread out (a version using only the top signal bunched 72% of its selling into one two-week window). Honest limits: that is one bull market, and the early thresholds are lightly fitted to it; the top-signal steps are validated on only two tops, which is why Leg 5 exists as a trailing backstop. If ETF demand really does bring back larger cycle tops, the plan's answer is a bigger core, not looser rules.
Projection band where the top might land · sizes the steps · never a trigger
Two geometric estimates of the ~2029 top that disagree by about 1.5–2×: a conservative $166k floor (each cycle top has gained less over the previous one, and that fade has been steady; this estimate does not depend on where the bear market bottoms) and a higher Fibonacci zone ~$211K–$246K, measured up from the bear-market low and computed live from the current $58K bear-market low. The lower edge (~$211K, the 0.618 level) is one the top has passed in every cycle, by 9% to 137%; the upper edge (~$246K, the 0.69 level) is where tops have actually peaked. It is provisional until the bottom is confirmed and recomputes each day as the low updates. Both estimates are built from the same handful of past turning points, so they are not two independent sources of evidence. The percentile rules decide when; the band suggests how much and where the peak might be.
$166K start ~$211K · passed every cycle ~$246K backstop
Spot $84,053 is below the band (+97% to the $166K floor) — the plan is dormant, nothing is due.
ZonePriceWhat the plan does
A · begin ~$166K The conservative low estimate. A weak cycle can top near here, so the plan starts selling if its percentile rules trigger and does not wait for more.
B · core $166K–$211K Up to the level the top has passed in every cycle. The main selling zone: the plan keeps selling through it, front-loaded, as the top signal confirms.
C · peak & backstop $211K–$246K Where tops have actually peaked. The final steps and the trailing step act here; past ~$246K the plan sells down to the core. Only a 2017-style runaway cycle reaches the ~$280K tail.
In short: $166K is where the plan starts selling, never a level to wait above, and it is the start, not the top. The top has always at least passed ~$211K, so the plan keeps selling through zone B, with the final steps and the backstop near ~$246K. It does not finish early: there is a real chance price runs above the conservative floor. The floor is firm; the Fibonacci zone recomputes automatically as the bear-market low updates.
Fibonacci cross-check context only · based on three past cycles · never a trigger
A cross-check on the ladder: each cycle's turning point compared with a Fibonacci projection of the previous cycle's shape. It uses only turning points that were already confirmed, so nothing is redrawn after the fact. It has worked on the sell side (the last two tops landed within about 9% and under 1% of the projected zone) and not on the buy side: two of three bottoms came in shallower than the band, so it missed them.
BUY zone · retrace of the last bull: 0.382 $56,616 · 0.5 $44,188 · 0.618 $34,488
spot $84,053 = 0.194 retrace — still above the buy band (it starts at 0.382)
SELL zone · next top: pending — it can be computed once this bear market's bottom is confirmed, as bottom × 8.2×0.5–0.786
Context only — never a model input, and it moves no step of the ladder.

What this site is

Bitcoin Intel tracks where Bitcoin sits in its roughly four-year cycle. It combines a long-run valuation trend, on-chain behaviour read directly from the blockchain, market and macro data, and a small set of statistical models. It is rebuilt from fresh data every day. Nothing here is a recommendation to buy or sell.

How the models are tested

Every model reading shown as a track record is out of sample: the model is trained only on data up to a given week, asked for a forecast, and scored later against what actually happened. It is then retrained with one more week and the process repeats. A model never gets credit for a week it had already seen.

Over the test period (through 2026-04-26), the primary model leaned long in 333 weeks and price was higher 26 weeks later in 72% of them. For comparison, price was higher 26 weeks later in 62% of all weeks, so the model’s edge is the gap between those two numbers. Each forecast covers 26 weeks, so neighbouring weeks overlap heavily: the whole period holds only about 16 independent six-month windows.

Run as a strategy that holds Bitcoin for 26 weeks after each long call (26 staggered positions), its risk-adjusted return (Sharpe ratio) was 1.24, against 0.95 for simply holding Bitcoin over the same weeks. Since the late-2021 top the figures are 0.92 against 0.48. Measured instead on separate, non-overlapping 26-week windows, the Sharpe ratio ranges from 0.36 to 1.20 depending on which week the windows start.

The cross-check model was right 74% of the time when it leaned long (base rate 62%).

One caution covers all of these numbers: the models’ inputs and settings were chosen after looking at this same period, so the figures flatter them. The clean test is forecasts made after the design was frozen on 2026-06-10. Those are being recorded as they are made; the first resolve on 2026-12-09.

The average hides a lot. Year by year, the share of long calls that were right:

YearLong callsRight
20182730%
20192361%
202052100%
20213956%
20223465%
20235391%
20244796%
20254146%
20261759%

It does well in established trends and poorly around turns, which is exactly when a reader most wants it to work.

Why a high reading is not a promise

  • The primary model runs high by nature: its typical reading sits well above 50%. The home page therefore shows how today’s reading ranks against the model’s own history, not just the raw number.
  • Retraining the same model with a different random seed moves the reading by about ten points. Read the level, not the week-to-week change.
  • Bitcoin has had only four complete cycles, and the period with complete data (since 2016) holds three tops and two bottoms. The cycle-turn models learn from those few turns. That is a thin record. Treat their output as one input, not a verdict.
  • The cycle-turn readings are scores from 0 to 100 that rank how turn-like a week looks. They are not probabilities: turns are rare, and the models are trained as if they were not, so a reading of 40 does not mean a 40% chance.
  • The proximity scores (how many weeks to the next top or bottom) are reference lines on Top & Bottom Signals only. The bottom score learned from two bottoms that both came about a year after their tops, so it mostly tracks the calendar: past that point it drops sharply whatever the market does.

What the regime model can and cannot do

The regime model labels the market as bull, bear or transition from price behaviour. It confirms a top five to ten weeks after the fact and has never identified a bottom in real time. It is context for the other readings, not a timing tool.

The regime shown changes only after two complete weeks in a new state, so a single unusual week cannot flip it. The model is fitted ten times from different starting points; the page shows how many of those fits agree with the regime shown, and calls the regime mixed when fewer than half do.

What is drawn with hindsight

Cycle Map (hindsight) is fitted on the full price history, including the future relative to any point on it. It shows what a cycle looks like once it is over and is useful for orientation. It is not a forecast and is never used to measure performance. Every reading on the home page and on Top & Bottom Signals uses only what was known at the time. That includes knowing when a past top or bottom happened, which is only clear in hindsight: the models count a top as known once price had fallen 40% from it, and a bottom once price had risen two-thirds from it. For the turns since 2015 that took between 7 and 20 weeks, and each line starts only after the first turn it could have learned from.

Context gauges

MVRV, NUPL, supply in profit, funding, implied volatility, Fear & Greed and the other gauges are shown for a human to read. None of them feeds a model, and their historical thresholds drift from cycle to cycle. Each gauge shows the date of its data when that data arrives late.

Data sources

Prices: FMP and a six-exchange volume rollup (Coinbase, Crypto.com, Kraken, Bitstamp, Bitfinex, Gemini). On-chain: a self-hosted Bitcoin node, CoinMetrics and Blockchain.com. Macro: FRED. Derivatives: Deribit. ETF flows: Farside. Stablecoins: DefiLlama. Sentiment: Alternative.me.

Bitcoin Intel · refreshed daily · Methodology · For information only. Not financial advice.