Bitcoin runs on a ≈4-year clock. Bottoms have struck ≈524 days before each halving, tops ≈535 days after. This site measures that clock on real data — and shows you, in one live chart, two simple scores and a daily report, exactly where today sits.
No account. No install. Your settings stay in your browser.
Bitcoin's supply halves roughly every four years. Around each halving, price has traced the same anatomy — measured here on real data, not folklore.
Start with a clean TradingView-style chart. Layer on exactly the evidence you want — every toggle remembers your choice, privately, in your own browser.
Halving lines, BUY/SELL dates with every historical move measured close-to-close, technical confirmations, and the research tables — over live candles that tick with every Bitstamp trade. Layer any of 18 evidence overlays, sketch your own ideas with the built-in drawing tools, and read it on four timeframes. The current bar forms in front of you, with a countdown to the close.
One click marks every measured bottom and top with its distance from the halving, shades the fitted windows for the 2028 cycle, and opens the CYCLE MATH panel — anchors, decay law, rhythm constants. The Decay Ceiling draws each cycle's causally-projected top zone (both testable tops printed at 81–83% of it), and a fitted floor box brackets the forming bear bottom. All labeled for what they are: hindsight fits, not forecasts.
The classic ±500-day rule ships untouched. Next to it: a research preset fitted to the measured anchors — buy 450 days before, sell 550 after — with its full backtest shown right on the chart, honestly compared against buy & hold.
Reduced-form implementations of five candidate mathematical laws of Bitcoin — cost-basis elasticity (network cost basis with dispersion-normalized bands: tops are measurably more stable in Z than in raw MVRV), halving absorption (net coins leaving exchanges per unit of new issuance — flips positive after every modern halving; ×2.5 right now), a volatility wedge, a security-settlement manifold, and a dormancy-release hazard — graded on the full on-chain history: NUPL separated every measured top (≥0.65) from every measured bottom (≤−0.28), n = 3+3. The MARKET LAWS panel shows each law's live state and its verdict — including the ones our data can't confirm. A research tool that grades its own homework.
The Cycle Regime Map splits the 4-year cycle into ten named phases and checks the calendar's claim against four independent confirmation engines. Everything rolls up into two scores anyone can read — live, below.
Ten phases, anchored to the halving clock. The lit pill is today.
fetching the live score…
How attractive current conditions score against the measured past, 0–100. A weighted read of timing, valuation, technicals, holder behavior, liquidity and leverage — each weight published.
fetching the live score…
How overheated the market looks, 0–100. High heat is a risk reading, never an entry signal — the site tells you that on every surface where the number appears.
Live snapshot · confidence —/100 · every component and weight is on the dashboard
Price trend and momentum: Mayer Multiple, weekly RSI, MACD, drawdown, volatility.
What holders are actually doing: MVRV, NUPL, reserve risk, cost-basis floors, miners.
Money flowing in or out: exchange flows, issuance absorption, ETF demand, macro.
Leverage and froth: funding rates, open interest, implied volatility.
The calendar is only one opinion. Two more systems read the market with no calendar at all — and the site shows you where all three agree, and where they don't.
Counts days around the halving. Simple, transparent, testable — the ten-phase clock you saw above.
Phase table + the ±500-day rule, versioned and backtested.Names the market's state purely from the data's behavior — and flags the exact moment that behavior shifts.
Hidden Markov regime probabilities + Bayesian change-point detection, no lookahead.Hunts the past for stretches that look most like today — then shows what happened next, with sample sizes.
Multivariate DTW historical analogues + phase similarity.The live three-way read loads a moment after the page. Divergence is a finding, not an error: it means a transition may be under way — or this cycle is behaving unlike the measured past.
combined confidence —/100 · never averaged away, always shown
Every model here is tested the hard way before you ever see it: trained only on the past, graded only on data it never saw. Its report card is public.
0%
directional accuracy — walk-forward, out of sample
371 ninety-day forecasts · 24 folds across 15 years · beats every naive baseline it was tested against
A forecast here is a distribution — quantiles with probabilities, resampled from every comparable moment in history. Never a single magic number. Under 30 samples, the page says so and widens honestly.
Before anything else, the lab states the measured chance of a painful drop inside the forecast window — not just where price might end.
Data freshness, drift, out-of-distribution and structural-break monitors — one Model Health dashboard, refreshed daily. When inputs degrade, confidence drops and the page tells you why.
New models predict in public but cannot touch the production scores until their out-of-sample record earns promotion. Predictions are stored before their outcomes exist — no backtest subtlety can contaminate them.
The whole pipeline is deterministic and inspectable — the same numbers every time, from public data you can check yourself.
Full Bitstamp BTC/USD daily history, extended back to 2010 with Coin Metrics community data. The chart re-syncs itself minutes after every UTC bar close.
BUY = halving − offset · SELL = halving + offset.
Every completed BUY→SELL and SELL→BUY move is measured close-to-close and
listed in the performance table — including the drawdowns.
Only completed weekly bars count — no repainting. BUY: weekly RSI(14) crosses above 35 while the Mayer Multiple ≤ 0.90. SELL: weekly MACD turns bearish while RSI ≥ 45 and Mayer ≥ 1.00, within ±90 days of a calendar date.
Bottoms and tops located per cycle → anchor statistics, the ×0.27 decay law, a log-log power-law corridor (R² 0.96), and projected 2028 windows — every fit shown with its sample size and dispersion.
Models train only on the past and are graded on data they never saw — 371 out-of-sample 90-day forecasts behind the accuracy number above. New models wait in shadow mode until their record earns promotion.
No refresh-watching required. The lab publishes its read once a day — short, visual, and free — and the full report is one click away at any hour.
Every morning at 08:10 Berlin time, the bottom checklist and both scores go out as one clean card on X — the same numbers this page reads, drawn by the same code.
next post in —
Follow @Derik_vt on XThe bottom checklist and the model intelligence read, generated fresh from live data the moment you click — a clean PDF you can file, print or forward.
Generate analysis report (PDF)Each layer on the chart in one plain sentence, with the exact math and its intended use one click away. Nothing is a black box, and nothing here is financial advice. See them all combined on the Cycle Dashboard.
Buy 500 days before each halving, sell 500 days after. That is the entire rule — and it beat holding in every completed cycle.
BUY = halving − 500d · SELL = halving + 500d. Moves are measured close-to-close;
the backtest is long BTC inside the windows, cash outside, 0.1% cost per switch:
112.6%/yr, Sharpe 1.40 vs buy-and-hold 76.1%/yr, Sharpe 1.10.
Use: the baseline. Everything else on the chart exists to confirm or challenge these two dates.
The same idea, tuned by a backtest scan around the measured anchors: buy a little later (after the average bottom has already printed), sell a little later.
Offsets from an offset-sensitivity scan over the measured anchors:
BUY = halving − 450d · SELL = halving + 550d →
137.6%/yr, Sharpe 1.55, max drawdown −74.8%. An in-sample fit on three cycles —
versioned, never silently replacing the classic rule.
Use: a research preset to compare against ±500, with its backtest shown on the chart.
The calendar says “around here”; momentum says “now”. The aqua and fuchsia flags mark when weekly momentum agreed with a BUY or SELL date.
Completed Sunday-close weekly bars only — no repainting. BUY: weekly RSI(14) crosses above 35 while the Mayer Multiple (price ÷ 200-day SMA) ≤ 0.90. SELL: weekly MACD(12,26,9) crosses bearish while RSI ≥ 45 and Mayer ≥ 1.00 — each within ±90 days of a calendar date; only the first hit per window is kept.
Use: timing refinement inside the calendar windows.
In the last three cycles the bottom struck ~524 days before the halving and the top ~535 days after — and bottom-to-top took almost exactly 1,059 days every time.
Extremes located per cycle, then averaged: bottoms −524d ±16, tops +535d ±11, bottom→top 1,059d ±9 (0.8% variation), bears 383d ±21 at −77…−85%, and the halving sits at the temporal midpoint (ratio 1.02 ±0.05). Future zones are drawn at mean ±1σ.
Use: the “when” scaffolding — fitted bottom Oct–Nov 2026, fitted top around Oct 2029.
Each cycle’s pump is a shrinking fraction of the previous one. Projecting that shrinkage gives this cycle a price ceiling — known on halving day, before it happens.
Ceiling = halving-day price × decay-projected multiple, where the multiple is a
log-linear extrapolation of prior completed cycles’ halving→top multiples
(×92.6 → ×29.6 → ×7.9, decaying ×0.27 per cycle). Causally testable twice:
the 2021 top hit 83.2% of its ceiling, the 2025 top 81.0%.
The dotted line marks the 0.75× terminal zone; amber boxes show where the zone crosses
the calendar top window — both real tops died inside their box.
Use: a level zone, not a timing signal — it means most inside the calendar top window.
The same shrinking-cycle logic pointed downward: a falsifiable target box for where this bear could bottom.
Two independent estimators bracket the zone: the bear-depth trend (−84.9% → −83.4% → −76.7% → next ≈ −73.5% of the top) and the bottom ÷ previous-halving-price ratio decay (14.0 → 4.9 → 1.8 → next ≈ 0.66×) → $33k–43k, drawn across the fitted Oct–Nov 2026 window. Never tested out of sample.
Use: a concrete box to be proven right or wrong — note this bear is already shallower than every precedent, so a shallow miss is the live risk.
Across its whole life, Bitcoin’s price has grown along a straight line — once you plot both price and age on logarithmic scales. The corridor brackets that line.
log₁₀(price) = a + k·log₁₀(days since genesis) with k = 5.66, R² = 0.96;
support, mid, and resistance sit at the 1st/50th/99th percentiles of the fit residuals.
Use: long-horizon context — every bear-market bottom (2015, 2018, 2022) held inside the corridor, 36–81% above the support line.
The blue line is what the average coin last changed hands for — the market’s collective cost basis. Price stretches away from it and snaps back.
Realized price = price ÷ MVRV. The deviation is normalized by its own 4-year dispersion:
Z = ln(MVRV) ÷ σ₄ᵧ. Measured result: tops cluster at Z ≈ 2.74 with far less
spread than raw MVRV (CV 0.19 vs 0.33), while bottoms are tightest in raw MVRV (≈0.68).
Bands: cost basis × e2.74σ (top) and × 0.68 (bottom).
Use: how hot or cold price is versus what holders actually paid — the bottom band independently agrees with the fit floor.
Are buyers soaking up more coins than miners create? Above 1 on the histogram means yes — a supply squeeze.
(exchange outflows − inflows, 90d avg) ÷ (new issuance in USD, 90d avg).
It flipped positive in the year after every modern halving (−0.21→+0.09, −0.75→+0.19,
−0.34→+0.20) and currently reads ×2.48 — historically extreme accumulation.
Use: a flows-based leading gauge; strong absorption during weak price is the classic accumulation signature.
Options traders pay a premium over recent realized chaos; the size of that premium says how nervous the market is.
wedge = DVOL (Deribit 30d implied vol) − trailing 30d realized vol, deflated by
annualized |funding|. Currently ≈ +3 points (32nd percentile — sleepy). The full invariant
needs order-book depth and open interest we don’t have, so its stationarity claim is
graded not confirmed — and the panel says so.
Use: context only. A lab that grades its own homework keeps its confirmations credible.
Compares what Bitcoin spends on security (mining) with how much value it settles. Big imbalances should flag stress.
ln(issuance ÷ realized cap) ~ ln(settlement value ÷ realized cap); the regression
residual is the state. On our community-data proxies the relation is weak
(R² 0.08, ~3.5-year half-life) — graded weak, displayed for honesty.
Use: watch, don’t lean on it — it needs adjusted transfer-value data to matter.
One colored strip that reads all of the above and names the season: accumulation, expansion, distribution, or de-risk.
Four agreement-gated blocks: cost-basis Z position between the measured bottom/top thresholds, the cycle leg from the anchors, volatility stress versus expanding percentiles, and absorption > 1. Components that disagree are shown as a “mixed” flag instead of hidden. Example: the Oct 2025 read flipped to DE-RISK — mixed, since absorption still said accumulate; the chart and dashboard always show the live read.
Use: the one-line answer to “where are we?” — the dashboard breaks it into components.
A 0–100 pressure gauge asking three questions at once: how stretched is price above its structural anchor, is that anchor still repricing upward to catch up, and how much of the cycle’s volatility budget is already spent? Stretch is only fragile when the anchor stalls and the budget is gone.
X = 1.2·z(ln P/A) − 0.9·z(Δ₉₀ ln A) + 0.8·(2V−1) + 0.35·z(volume impulse),
ERCO = 50 + 50·tanh(X/2), smoothed by a 7-day EMA. The anchor A is realized
price when on-chain data is fresh, else a price-only hybrid √(200W×730D SMA). V is cumulative
squared daily returns since the halving divided by a causal budget — the median spent
at prior cycles’ tops only. Measured with the report’s default weights (nothing fitted
here): all three measured tops read 71–73 on the realized-price anchor (n = 3), while
bottoms drifted 10 → 20 → 35 as the market matured. If the on-chain feed ever goes stale the
overlay automatically falls back to a price-only hybrid anchor (tops ≈77/62, n = 2) — and the
chart panel always names the active mode. Honestly graded: the report’s prototype
trading rules underperform buy-and-hold
(realized-price anchor: Sharpe 0.65 vs 1.22; hybrid: 0.60 vs 1.09) — so the lab displays
ERCO as a state gauge, never a trade trigger.
Use: read it against the calendar windows — a high read inside a SELL window is top-side confluence; a low read inside the BUY window is accumulation-consistent. The anchor-band overlay draws A·eμ±σ on the price chart.
Are long-term holders sitting still or starting to spend old coins — and how much paper profit is the market sitting on? In the on-chain literature, tops have come when old coins move while everyone is deep in profit — a pattern our short data window can’t yet verify.
NUPL = (market cap − realized cap) ÷ market cap — below 0 is capitulation, above
0.75 is euphoria. Liveliness = lifetime coin-days destroyed ÷ coin-days created; a rising
90-day change means dormant supply is waking up, and a coin-days-destroyed z-score adds spend
intensity. The grade is computed from the data window: with at least two completed tops
and bottoms in view, the panel reports whether NUPL actually separates them; with a shorter
window it honestly says “context only”. The MARKET LAWS panel always shows the current grade.
Use: a top-side sanity check. Distribution warnings need BOTH high NUPL and rising liveliness; neither alone is enough.
Two more cost-basis lines with different owners: what short-term holders paid (bull support, bear resistance) and what long-term holders paid — the conviction floor. Where price sits on the ladder tells you who is underwater.
STH/LTH realized prices are the average acquisition cost of coins younger/older than ~155 days. Measured on our extremes: every measured bottom printed below the LTH floor — at 0.571×, 0.723× and 0.771× (n = 3, and the ratio rises as the market matures). Alongside Reserve Risk, whose measured bottoms cluster tightly at 0.0004–0.0007 (n = 3; its tops decay, so it grades bottoms only).
Use: structure at a glance — price under the STH line is bear structure; a dip toward or under the LTH floor has marked every generational low.
Two forces the old cycles never had: US spot ETFs absorbing (or dumping) coins, and the global money supply everyone claims Bitcoin follows. One is brand new; the other we actually measured.
ETF net flows (BTC/week) start Jan 2024 — no historical baseline exists, so they are context, never part of a graded law. The M2 thesis we graded properly: weekly year-over-year changes of BTC vs global M2 show only a weak full-sample correlation with no lead-lag advantage at any horizon up to 40 weeks — the popular “M2 leads by 10 weeks” claim does not hold on this data, though the recent era correlates moderately. The dashboard shows the exact numbers.
Use: ETF flows show whether the new demand channel is buying or selling this month; M2 is macro mood — useful context, honestly graded as nothing more.
Three distinct floor estimators plus the pain gauge, all graded on the measured bottoms. When these methods and the calendar converge on one zone, that’s the strongest read this site can give — and it says so with sample sizes attached.
CVDD, Balanced Price and True Market Mean are graded by their measured bottom-ratio bands (price ÷ estimator at the n = 3 bottoms): bottoms never touched CVDD (printed at 2.5–3.4× it), held above Balanced Price at the 2018 and 2022 bottoms (1.23–1.40×) — the 2015 bottom printed 13% below it (0.87×) — and printed at 0.36–0.53× TMM. Implied zone = today’s value × that band. All three derive from the same realized-value / coin-days-destroyed data, so their zones are correlated — methodologically distinct, not independent votes; the calendar is the only independent axis here. The pain gauge: every measured bottom printed with ≥98.6% of short-term holders in loss (n = 3) versus ≤0.5% at tops — the cleanest separation of any gauge on this site. Hash ribbons (miner capitulation) overlapped every bottom ±90d but also fire elsewhere — context only. The dashboard shows the live zones and today’s reads.
Use: watch the dashboard as the consensus bottom window (union of the independent time methods, Oct 2026 – Jan 2027) approaches — the radar shows which conditions are met, never a buy signal.
Most market sites sell certainty. This one sells measurement — and four rules make that difference real on every page.
Every line is computed from public data with its formula shown. Hindsight fits are labeled as hindsight fits — never dressed up as foresight.
Three completed cycles is a sample of three. Each claim ships with its n and its spread, right where the number is.
When evidence is thin or systems disagree, the platform says exactly that. A tool that can't say “I don't know” is a marketing site.
Research and education, nothing else. No calls, no targets, no “buy now”. What you do with the evidence is entirely yours.
Bitcoin has completed three full halving cycles with reliable data. Three of anything is a pattern, not a proof. That is why this chart shows you dispersion, sample sizes, and backtests instead of predictions — and why the current cycle already bends some of the old rules. Research evidence, honestly presented. Never financial advice.
Fitted bottom zone: Oct–Nov 2026 · Fitted top zone: Oct 2029 — and the Opportunity and Heat scores are reading today's tape right now.
Open the live chart — free