Hold
The rules for adding and for reducing are not met.
Know what each label means. Definitions, not current signals.
The rules for adding and for reducing are not met.
No supported direction: evidence is late, conflicting, or no rule is met.
How many of six recovery conditions are met.
Not a probability or confirmation of a final low.
Schematic only · no current reading. Checks can share evidence; six conditions are not six independent confirmations. A final low can remain unconfirmed.
Read recovery methodologySummaries of supporting evidence.
CryptoWatch reads Bitcoin’s on-chain, fund-flow, cycle and macro data once a day and turns it into one posture, with the evidence behind it and what would change it. It does not place trades, size positions or predict prices.
The Dashboard has five views, each led by one main panel; on Now the short-term score sits inside the evidence balance. The rest of each view is evidence for that panel. Each section below ends with cards for its panels.
It is rule-based, dated and fail-closed: a missing or stale input reads Unavailable, never a guess. It is not a signal service, investment advice or a price forecast, and Action copy never becomes an order.
Home and Bitcoin State show the same daily read at two depths. The read describes the last completed day, so its date is usually yesterday.
State reports a change only after a new state holds for two daily reads in a row; one-day blips are dropped. Ranked metric moves are the Dashboard’s Since Yesterday.
Two context rows on State: miner revenue against its yearly average, and Bitcoin’s 30-day return against traditional assets. Neither changes the posture.
Open Bitcoin State →The posture is a research label set by fixed rules. It is not an order and it does not set a position size. Today’s plan splits one buying budget into three parts; the posture decides which are eligible.
Both accumulation postures also need nothing that weakens the read to be triggered. Both accumulation postures and aggressive distribution need complete supporting reads and a recognized macro mode.
Each part reads Wait or Eligible; under Hold, Wait or distribution every part reads Wait. When distribution evidence permits, a planned de-risk budget can show the same 20% / 50% / 30% split, labeled PLANNED, WAIT or ELIGIBLE. Nothing executes.
Process reads outrank calendar day-counts, live readings outrank analogs, and missing data fails closed. A positive score and a late-bear calendar can coexist with Hold: the posture owns permission.
Open today’s plan →The phase says where the market sits today. The clocks are history, not a forecast. Recovery checks count how many of six fixed conditions are true now. None of them dates a bottom.
The scorecard names one stage; "Recovering" means the low is not confirmed. Neither it nor the price-trend gauge is a forecast.
The cycle engine names its phases Capitulation, Accumulation, Markup and Distribution; the Dashboard's price-trend gauge shows them as Falling, Basing, Rising and Topping, and the public cycle stage comes from the scorecard. Valuation, holder behavior, drawdown, momentum and the halving calendar (one vote) add points to a phase, and the most points wins. Overheated means MVRV Z-Score 3.0 or NUPL 0.55, levels this cycle has actually reached. The current phase holds unless another leads by two points or more, and the long-term-holder SOPR input is a seven-day mean, so one day cannot flip it.
Last halving: 20 Apr 2024, block 840,000; the next is block 1,050,000. Past tops fell 480–560 days after a halving and past lows 900–1,100 days after (HIST. BOTTOM). October 2025 is a local high in an unfinished cycle, not a confirmed top.
Bitcoin State sets the price read beside three clocks: days since the cycle high, since the 2022 low, and since the halving, each against where past lows fell. When price and calendar disagree, the tile says so and counts the days left in the low window. A midterm-year September hike adds a fourth bar for 180 days; two past cases, so it never sets agreement. History, not a forecast.
State’s watch list leads with a weekly close under the 20-week band, then the 50-week average: its level, its four-week move and the weeks since the latest breakout over it (after at least eight closes below) without a close back below, with four cases since 2015. A level and count, not a trigger.
Recovery needs price above the True Market Mean, four or more checks including the recent-buyer reclaim, a prior fall of 20% or more from the dated local high, and no break condition firing. In the one closed episode since spot funds launched, three of six was followed by a further 31% fall, so the bar is four. Even 6/6 is not a guaranteed low.
A new lower close resets the observed low; missing or conflicting days withhold it. MVRV, NUPL and Realized Price share profit information, so their agreement is not three confirmations. The Pi series here is Pi Cycle Top; the separate Pi Bottom idea was withdrawn by its author in July 2022. Proximity is historical similarity on a 0–100 scale; the 180-day likelihood stays Unavailable until walk-forward gates pass. November 2022 is the textbook analog, as context only.
Top risk is a process (NONE → EARLY → ACTIVE → EXHAUSTION → POST_TOP_CONFIRMATION) shown on CW Risk; a terminal top is an event, never merged into one score. Top percentages stay unavailable until their gates pass, which does not mean no risk.
The indicator ledger shows well-known top and bottom readings against four kinds of mark. Classic is the level the indicator is known for. ETF-era is the engine’s rolling band, four years for tops and two halving epochs for bottoms: a top band can only be easier than classic, a bottom band only stricter. Definition is where the indicator’s own construction crosses a line. Published is a level an author or publisher put on the chart; it is drawn and never fires. Only engine marks fire, and none is a forecast. A rank chip on a row says where today’s value sits among the last four years of that indicator (top, bottom or mid-range). It describes the reading; it is not a threshold and not a signal. The crossing line names the last stored day the value moved to the other side of the line the distance names, and the days since; the sparkline above the scale is two years of weekly readings with the marks drawn across it.
Past readings are the values on the last cycle’s high and low and on this cycle’s highest close, found from the daily price history. A row hides while its data is not current. CVDD, balanced price, adjusted SOPR and dormancy flow were added on 29 Oct 2026 and show once collected. Binary CDD and fee momentum were added on 8 Nov 2026 as companion readings inside their host rows, from data already pulled. Reserve Risk, RHODL, thermocap and the Mayer Multiple wait for an engine rule. The full ledger is /api/public/v1/indicators.
Panels on this view
Research models · not on the Dashboard
Recovery checks lead: six fixed conditions from the canonical bottoming receipt and what would break the read. They never date the final low.
The price-trend gauge uses the phase vote above. The moving-average stack adds a small bonus from compression, the spread of the 20-, 50-, 100- and 200-week averages as a share of price, and velocity divergence, price over its 50-week average against that ratio’s 20-week mean.
Zones are observation ranges, not forecasts. The power-law trend distance is descriptive and uncalibrated, and the wave candidate failed its out-of-sample promotion gate, so it stays research only.
Off the Cycle view (research only): CW Risk valuation bands, the prices at the 95th, 50th and 1st risk percentiles, and the separation statistic, a rank correlation of past scores with 30-day returns. Network risks (mining, fee share, custody concentration) describe the network, not a trade; Now’s Long run carries that pillar.
The bottoming likelihood counts, over all-time-high drawdown episodes, how often no lower daily close followed within 180 days. Its validation history stays in the research record; the Dashboard shows only the current, publishable line.
The Cycle verdict card carries the recovery count, the open check and what would break the read; the full gate list is its own card. The bottom-proximity history, how price behaved after similar readings, is thin, overlapping evidence and stays research only.
Money shows who is adding or selling Bitcoin (long-term holders, US spot funds, companies) and where the coins sit. When the buyers agree, conviction is highest. “Split” means CryptoWatch will not force one story.
Panels on this view
Each flow keeps its own unit and window; ETF flow is measured from 11 Jan 2024. New cash is a deployable-liquidity proxy, not committed Bitcoin purchases.
Long-term-holder coverage counts the live holder inputs; clarity and cross-validation describe how strongly and how consistently they agree. None is a price probability.
The ETF week is “within the normal range” when the 7-day average daily flow is between −$50M and +$50M. Treasuries show the 30-day change and top holder; the holder-type split and top-five list are research detail.
Research only: the BTC and ETH rotation cross-checks (Bitcoin dominance, the ETH/BTC trend) feed no verdict.
The per-fund ETF table, the holder coverage diagnostics (coverage ring, override rules, evidence list) and past long-term-holder inflections are research detail and stay off the Dashboard.
Holder cohort references use a fixed 155-day boundary with z-scores and percentiles. The long-term-holder share is that cohort over circulating supply, and the market-cap-to-flow ratio appears only when the 30-day net flow tops $100M; both are descriptive.
Macro asks whether the backdrop helps or hurts a bullish Bitcoin read. Growing money and easier conditions let bullish evidence count; a rising dollar, higher real yields or credit stress can keep even a cheap on-chain setup from paying.
Panels on this view
Research models · not on the Dashboard
Macro drivers lead: money supply, the dollar index, central-bank liquidity and real yields set the scorecard’s Macro line; rates and global markets share the card. Monthly series such as M2 arrive 30–45 days late, so each tile shows its own date.
Copper and the dollar against the yuan are not among the four drivers the Macro line counts, so they are off the view. The 10-year minus 2-year spread labels the yield-curve chart.
Research only: capital rotation, explained below.
Action turns the posture into a research plan: regime, allowed posture, what would invalidate it, then planning levels. Nothing here places or tracks an order.
CryptoWatch does not forecast a floor. It shows where holders’ costs sit (Realized Price, long-term-holder cost, recent-buyer cost, True Market Mean); planning zones below them describe past stress, not a promised rebound. Structural risks are on Tail-Risk.
The Next test line is calculated each day, not written: it names the verified cost level nearest to price, the odds of touching it within a week if price moved as a random walk at this month’s volatility (an odds label, not a forecast), and what followed in past setups like today’s when at least eight exist. A break counts after five straight daily closes, the point where past breaks held about three times in four; one close past a level has held about half the time.
Funding, Options and Liquidations read together as one word, the scorecard's Timing line. The Options row compares one-month implied with realized volatility (DVOL neutral band 35–70), with chips for the premium and the one-to-six-month curve. Timing is not a second cycle call.
Panels on this view
Research only, no execution: setup and cost levels describe preparation, not orders. Timing reads leverage, options and liquidations as one word.
Timing in plain words: the cost of leveraged bets is perpetual funding; expected swings are options-implied volatility (DVOL); near-term fear priced higher is backwardation and calm near term is contango; piles of bets that would be force-closed are modeled liquidation clusters. Setup notes left Action on 3 Nov: they repeated Cycle and Money, and one check had no input.
Volatility curve: near means options expiring within 10 days, far beyond 60; steep contango means far is more than 10% above near. The daily curve is 6-month minus 1-month volatility, flat from 0 to under 5 points.
Scores summarize evidence; they never set the posture. The Composite Score adds signed inputs from four evidence layers, the Bitcoin Thermometer measures heat from 0 to 100, and CW Risk is cycle temperature. None is a probability.
Each input scores from −2 (euphoria, crowded longs) to +2 (stress, forced selling, crowded shorts). The total falls in one of seven fixed bands.
| Score | Evidence label |
|---|---|
| ≥ +16 | Strong positive lean |
| +8 to +16 | Positive lean |
| +2 to +8 | Mild positive lean |
| −2 to +2 | Mixed signals |
| −8 to −2 | Mild negative lean |
| −16 to −8 | Negative lean |
| < −16 | Strong negative lean |
The current plan follows Action. Buy and Sell trigger checklists use native setup checks and confirmed 0–1 CW Risk bands, separate from the signed composite score. Every check on a route must pass; missing or stale checks remain unavailable.
Confidence is 60% coverage (live inputs out of 26) plus 40% agreement with the total’s sign. Serious conflicts lower it and move the posture toward Wait. It is not a calibrated probability.
Inputs fill a fixed 26-slot universe, so a missing feed lowers coverage instead of leaving the denominator. Each weighted input is clamped to ±3 so one loud series cannot dominate, and cycle anchors such as AVIV outweigh short-term noise such as funding. The bands at ±2, ±8 and ±16 are a fixed convention, not a probability scale. The server builds the score from the daily snapshot. A fresh rebuild never renews an old observation. The API’s legacy BUY, WAIT and SELL keys only classify the score; they grant nothing.
Descriptive heat from 0 to 100 across eight valuation, profit, funding, sentiment, miner and holder-age curves: Ice Cold below 20, Cold to 40, Warm to 55, Hot to 70, Burning above. Context only; it neither confirms a bottom nor changes the posture.
Method bitcoin-thermometer-core8-descriptive-2026-10-15: MVRV Z-Score 12.77%, NUPL 18.09%, Funding EMA 20.21%, Fear and Greed 10.64%, SOPR 15.96%, Puell Multiple 6.38%, Reserve Risk 10.64%, RHODL ratio 5.32%. Funding needs 30 contiguous daily observations. All eight inputs must be current under the 28-hour rule; one missing input withholds the reading and the rest are never reweighted.
Historical grades are context: no live backtest, no public scoreboard, no probability until its gates pass. A price high without holder or fund support is unconfirmed.
Panels on this view
Long-run fundamentals are descriptive; low coverage limits them. The resolver score groups diagnostic evidence.
Negative leans bearish; positive leans bullish; near zero, inputs offset. Bounds sum available indicators’ weighted minimum/maximum points. Unequal downside/upside allowances explain asymmetric bounds, such as −45 to +43. Inputs and weights can change these limits; extremes may not occur together. They are mathematical limits, not percentages or price targets.
Data coverage: reporting indicator groups divided by expected groups. 100% means all reported, without certifying freshness or accuracy.
Directional agreement: reporting groups matching the evidence engine’s combined direction. Each family counts once; related profitability indicators share one family. For a neutral result, only neutral groups agree. This differs from the short-term score.
Evidence strength: round(0.6 × coverage + 0.4 × agreement), minus 5/10/15 points per low/medium/high conflict. Penalties total at most 35; the result stays within 0–100.
Agreement measures consistency; strength includes completeness. Neither predicts returns. Check dates and conflicts; Action owns position decisions.
Conceptual sequence · not a live timeline
Rebuilding analysis does not make an old observation fresh.
Every number comes from the latest snapshot. When an input is late or missing, the screen says Unavailable instead of guessing, and the posture falls back to Wait. Nothing is entered by hand as a live reading.
One Telegram brief goes out each morning after the read is current, opening with Home’s posture sentence. CryptoWatch is not an intraday alert tool. The forecast lock that once ran before the brief was stopped on 2026-09-15.
Software can use the same daily read through a public API and a hosted MCP with ten tools, one per question on this page. The OpenAPI document is the stability contract.
OpenAPI 3.1 specification · Hosted MCP setup guide
Read endpoints allow anonymous access at the lower limit. For higher limits and write operations, sign in to your CryptoWatch account and use Account → Developer keys; the key is shown once and belongs to that account. Keep it out of browser storage, chat messages and source control. To have a coding agent wire the REST API into a project, use Account → Developer credentials → Copy prompt and keep the raw key in CW_API_KEY.
curl -X POST https://cryptowatch.id/api/public/v1/webhooks \
-H "X-API-Key: $CRYPTOWATCH_API_KEY" \
-H 'Content-Type: application/json' \
-d '{"url":"https://example.com/cw","events":["composite.changed"]}'
Research theses are experimental endpoints, not a decision surface.
The short names on the Now cost ladder, and the words CryptoWatch uses across its pages.
The Now ladder’s short names. Usual is where average profit usually sits, not a target.
Retired surfaces, when, and what to use instead. Listed once here so the rest of the page describes only what is live.
The daily BULL / BEAR / NEUTRAL forecast vote was stopped on 2026-09-15. No new call is made; past calls are kept as history. CryptoWatch does not publish a public prediction scoreboard.
The daily forecast was a deterministic vote over the same live inputs shown on Bitcoin Dashboard. It did not run a multi-agent debate or invent price targets, and missing inputs failed the run. Direction-only calls resolved at T+7d (primary) and T+30d. A directional call was correct if Bitcoin moved in the predicted direction by at least 2.5% over the window. Moves inside ±2.5% were inconclusive. NEUTRAL was graded separately. No percentage printed until that horizon had at least one eligible resolved call.
Not financial advice. cryptowatch.id is a research and synthesis tool. It is not a signal service, not investment advice, and it cannot see the future. Cycle frameworks fail and past patterns do not repeat mechanically. Use it as one input among many, and never size a position from a dashboard verdict alone.
Some inputs depend on licensed upstream access, so this page documents scoring logic and the classes of evidence that can be disclosed.