How the BTC cycle terminal thinks.

01What this is

cryptowatch.id is a live BTC cycle intelligence terminal. It synthesizes macro, cycle, on-chain, and derivatives signals into a single composite gauge that outputs one of seven verdicts — from STRONG SELL to STRONG BUY — and shows you exactly which indicators are driving it.

The terminal exists because most "crypto dashboards" pick one lens (just on-chain, just TA, just sentiment) and miss the rest. Real cycle turns happen when macro, cycle position, on-chain cohort behavior, and derivatives positioning all line up. We wanted a single surface that shows whether they do.

Every number on the dashboard is computed live from primary sources. Nothing is hand-entered. Nothing is hardcoded. If the backend pipeline goes down, the number goes stale and the UI says so.

02How to use the terminal

A 6-step session workflow. Open the terminal, go through these in order. Each step narrows the decision space until you have a concrete action (or confirm there's nothing to do).

Step 1 · Read the composite verdict (SIGNAL)

The first thing you see. The gauge sums 21 signed scores into a single number from −24 to +24 and maps it to one of 7 verdict bands. Read two things:

• Verdict + score. "MILD BUY +7" means conditions lean bullish but conviction is partial.
• Confidence %. If it's above 80%, indicators agree. If it's below 60%, the total is being driven by a few strong signals while the rest disagree — act smaller.

If the verdict is WAIT (score −2 to +1), stop here. Preserve dry powder. Don't force a trade out of boredom.

Step 2 · Check L4 Execution first (FUNDING + OPTIONS)

L4 is the trigger layer. It tells you whether the derivatives market is already crowded in one direction. This matters more than anything else for timing, because extreme positioning resolves in 1-4 weeks, not months.

• Funding < −54% annualized → shorts are overloaded. Contrarian long setup in place.
• Funding > +54% annualized → longs are overloaded. Contrarian short / top warning.
• DVOL > 90 → panic priced in. Historically coincides with bottoms.
• IV term backwardation → immediate fear priced in. 2020 COVID, 2022 FTX both had it at the low.

Step 3 · Verify with L3 On-Chain (TIER 1 + TIER 2)

L3 is the cohort layer. It answers: "is the on-chain population actually at a stress point, or is this just derivatives noise?" Open the Tier 1 panel and read the triggered count.

• 0/4 Tier 1 → no confirmed bottom. Cap any planned allocation at 30-40%.
• 1-2/4 Tier 1 → approaching. Start watching for the next trigger. Deploy first DCA tranche if Zone A price is hit.
• 3-4/4 Tier 1 → maximum on-chain confluence. Only happens at cycle bottoms in backtest. Deploy full allocation.

Tier 2 is the confirming layer — softer triggers like LTH MVRV < 1, Coinbase Premium positive, Liveliness falling. Tier 2 alone isn't enough to act, but Tier 2 triggering while Tier 1 is still 1-2 of 4 is a leading signal that Tier 1 is about to fire.

Step 4 · Check L2 Cycle Position (CYCLE)

L2 is the slow map. It won't change between sessions — that's the point. Read:

• AVIV Ratio — where we sit in the post-ETF valuation band. <0.75 is historically extreme undervaluation.
• MA Stack Regime Score — 0/4 or 1/4 confirms bear regime. 4/4 confirms full bull.
• Halving regime — what phase the time-based calendar says we're in. If L2 says "late bear" and L3 says Tier 1 fires, conviction doubles.
• Bottom Proximity composite — the one-number summary of how close we are to the bottom zone.

Step 5 · Check L1 Macro permission (L1 MACRO)

L1 is the override layer. If macro is hostile, even extreme undervaluation can get cheaper. Read:

• M2 YoY > +3% → liquidity expanding. Full conviction in crypto-native signals.
• M2 YoY < −2% → liquidity contracting. Cap allocation regardless of how cheap on-chain looks.
• DXY +5% over 3m → dollar strengthening. Crypto fights a headwind. Reduce size.
• BTC-SPX 30d correlation high → macro dominates. Watch equities for direction.

2022 is the textbook reason L1 exists. On-chain screamed "cheap" for 6 months while DXY rallied 20% and crushed BTC another 40%. The framework's L1 layer would have capped your allocation during that window.

Step 6 · Map to DCA zones (DCA)

DCA translates the verdict + confluence into a concrete allocation. The terminal computes 5 zones (A-E) live from Realized Price and historical multipliers. Your job:

• Never deploy all capital at once. Scale in across zones.
• Zone A (closest to current price) gets the smallest tranche. Zones D-E (deepest) get the largest — reserved for maximum confluence.
• When Tier 1 fires 3-4/4 inside Zone C or D, front-load the remaining allocation. That's the highest-conviction buy in the entire cycle.

03The 4 questions

Before any indicator makes sense, internalize the mental model. Every one of the 21 indicators exists to answer exactly one of these four questions. When you know which question an indicator answers, you know how to weight it.

These 4 questions are a priority order when they conflict:

1. L1 wins overrides. Hostile macro caps any bullish call, no matter how strong L2-L4 are.
2. L2 defines the playbook. L3/L4 are read differently depending on cycle phase.
3. L3 provides the evidence. Without L3 confirmation, L4 alone is just "crowded positioning" — not a bottom call.
4. L4 provides the timing. When L2 and L3 both say yes, L4 is what you watch for the entry trigger.

⚠ Out-of-sample validation note

This framework was designed and backtested on BTC data from 2015-2025, covering three full cycles (2015 bottom, 2018 bottom, 2022 bottom) and one cycle top (2025 ATH). Every threshold, every verdict band, every regime boundary was calibrated to fit these cycles. The 2026-2029 cycle is the first true out-of-sample test.

This is more honest than most crypto tools but it means the framework's historical track record — while real — is in-sample. Treat forward readings as probabilistic signals, not guaranteed calls. Per audit v2 #19.

04The 4-layer framework

Every composite on the dashboard rolls up through four layers. Each layer answers a different question. Each contributes signed scores that sum into a single composite.

L1 · Macro Regime

Question: is the global environment supportive of risk assets, or is it pulling capital out?

Bitcoin does not exist in a vacuum. When the US dollar strengthens and global liquidity contracts, even a perfect on-chain setup struggles. When M2 is expanding and DXY is rolling over, the same on-chain setup becomes a launchpad. L1 is the permission slip.

Signals: US M2 year-over-year · DXY 3-month change · BTC–SPX 30-day correlation regime.

L2 · Cycle Position

Question: where are we in the 4-year Bitcoin cycle — capitulation, accumulation, markup, or distribution?

Cycle position is the slow-moving map underneath everything else. It blends time-based features (halving framework), technical regime (weekly MA stack), valuation (AVIV Ratio from Cointime Economics), and composite bottom-proximity scoring. This layer is what separates a "dip" from a "bottom" and a "pullback" from a "top."

Signals: MA Stack Regime Score · AVIV Ratio · Halving regime · Bottom proximity composite · Drawdown from ATH.

L3 · On-Chain Stress

Question: is the on-chain population under stress, taking profits, or still distributing?

L3 is the cohort layer. It reads what long-term holders and short-term holders are actually doing on-chain, separates them by age, measures profit/loss behavior, and tracks institutional flow through spot ETF activity and Coinbase premium. This layer catches real bottoms before price does.

Signals: Zero-FP Tier 1 confluence · LTH MVRV · STH MVRV · Liveliness / Vaultedness · Realized Loss · Coinbase Premium · IBIT activity multiplier · LTH SOPR.

L4 · Execution Timing

Question: is now the actual moment — is derivatives positioning primed to squeeze, or is the crowd already one-sided?

L3 says "a bottom is close." L4 says "and shorts are max-crowded at -70% annualized funding with backwardated IV." That's the difference between a six-week early call and a same-week entry.

Signals: BTC funding annualized · DVOL · IV term structure shape · 25Δ skew proxy · OI / market cap.

06The composite gauge

Every indicator contributes a signed score from −2 to +2. The layers sum up, the layers sum together, and the total lands on a gauge from −24 to +24. That total maps to one of seven verdict bands.

The scoring convention

+2 = strongly bullish (extreme undervaluation, capitulation, panic, or crowded short positioning).
+1 = mildly bullish (undervalued or supportive).
0 = neutral / no signal.
−1 = mildly bearish (overvalued or stretched).
−2 = strongly bearish (euphoric distribution, crowded long positioning, or extreme complacency).

Verdict bands

Confidence

Alongside the score, the gauge shows a confidence percentage. This is a coverage + agreement score:

• 50% from coverage — what fraction of the 21 indicators have live data right now (if a data feed is stale, its indicator drops out).
• 50% from directional agreement — of the indicators that have data, what fraction are scoring in the same direction as the total.

A +8 score with 90% confidence means "most indicators agree and most of them have fresh data." A +8 score with 55% confidence means "the total is being driven by a few strong signals while the rest are neutral or conflicting."

What the gauge is not

07Confluence decision rules

The gauge is the one-number summary. The decision rules are what you do when you actually have to act. This section is what separates reading the terminal from using it.

When layers agree — act with full size

All 4 layers aligned is the highest-conviction setup. It happens maybe 3-5 times per 4-year cycle. When it does, the signal is unambiguous.

When L4 conflicts with L2+L3 — reduce size, wait

The most common edge case: on-chain and cycle position agree there's an accumulation zone approaching, but derivatives positioning is on the wrong side. This is usually resolved by a liquidation event — don't deploy before it happens.

When L1 is hostile — override everything

Macro headwind delays even extreme undervaluation. This is the hardest rule to follow emotionally because on-chain will be screaming "cheap" while price keeps dropping.

Layer priority hierarchy

05Framework design rationale

Why 21 indicators and not 10. Why ±2 scoring and not raw values. Why some indicators are symmetric and some aren't. Why the verdict bands land at ±16 and ±8. These are not arbitrary choices — they're deliberate decisions with backtest rationale.

Why 21 indicators

The framework needs to cover 4 independent dimensions (macro, cycle, on-chain, execution) without missing signal and without becoming noise. We landed on 21 after testing configurations from 8 to 35. The distribution:

• 3 macro (M2, DXY, BTC-SPX correlation) — thin, because macro signals move slowly. More macro indicators adds redundancy without signal. 3 is enough to detect the regime.
• 5 cycle (MA Stack, AVIV, Halving regime, Bottom Proximity, Drawdown) — our highest-credit layer. Each indicator uses a different methodology (technical / valuation / time-based / composite / price) so they're uncorrelated enough that all 5 contribute independent signal.
• 8 on-chain (Tier 1 confluence, LTH MVRV, STH MVRV, Liveliness, Realized Loss, Coinbase Premium, IBIT activity, LTH SOPR) — biggest layer because on-chain is BTC's native data advantage. We want cohort behavior, valuation, miner stress, institutional flow, and profit-taking dynamics all covered.
• 5 execution (Funding, DVOL, IV term, 25Δ skew, OI/Mcap) — enough to capture derivatives positioning without making L4 dominate.

Fewer than 21 leaves blind spots. In testing, 15-indicator versions consistently missed either miner capitulation signals or IV term structure reversals. More than 21 crosses into diminishing returns — each additional indicator contributes less signal while adding correlated noise and computation load. 21 is where backtest accuracy plateaus.

Why ±2 scoring instead of raw values

Every indicator uses different units. M2 YoY is a percentage. MVRV is a dimensionless ratio. Realized Loss is dollars per day. Funding is an annualized percentage. You can't sum them directly. A normalized score per indicator (−2 extreme bearish to +2 extreme bullish, 0 neutral) makes them addable and comparable.

Why specifically ±2 and not ±1 or ±3?

• ±1 is too coarse. Only 3 states per indicator (bearish / neutral / bullish). Loses the "extreme" signal that is exactly what we care about at cycle turns.
• ±2 gives 5 states: extreme bearish, mildly bearish, neutral, mildly bullish, extreme bullish. This matches how humans naturally categorize conditions. "Cold, cool, mild, warm, hot" is the psychological template.
• ±3 or more over-weights noisy indicators. Funding can move 100% in a single day. Rewarding it +3 for a temporary spike would let it dominate the composite. The ±2 cap keeps any single indicator from owning more than ~8% of the theoretical range.

Why some indicators are asymmetric

Not every indicator scores symmetrically. Some push harder in one direction than the other. These aren't bugs — they're deliberate.

How the verdict bands were calibrated

The 7 bands (−24 to +24) aren't round numbers. They were set by replaying the framework through Time Machine against every historical cycle extreme since 2015 and checking where the composite score landed.

The ±16 threshold is where the framework has >80% historical precision for "this is a cycle turn." The ±8 threshold is the "conviction grows" line. ±2 is the noise floor — anything inside is a WAIT.

08Indicator reference

Every scored indicator, in order, with its endpoint, logic, and scoring bands. This is the full source-of-truth for what the terminal is computing.

L1 · Macro Regime

L2 · Cycle Position

L3 · On-Chain Stress

L4 · Execution Timing

09How the Market Heat thermometer works

The heat thermometer is a separate composite from the ±24 gauge. Where the gauge outputs a verdict, the thermometer outputs a 0-100 temperature. They measure different things. This section explains what it is, what it blends, and when it diverges from the main gauge.

What the thermometer measures

The thermometer is a valuation + positioning state, not a directional signal. It answers the question "how stretched is the market right now?" on a scale from frozen (0) to overheated (100). The gauge answers "is the setup bullish or bearish?" The two are related but not identical.

Think of it this way: the gauge is a verdict, the thermometer is a pressure reading. A neutral gauge (WAIT) can coexist with a very high or very low thermometer reading — the gauge is waiting for confluence, but the thermometer is already telling you the market is tightly wound one way or the other.

Components and weights

The thermometer blends four subscores, each normalized 0-100:

• Valuation (35% weight) — MVRV ratio, NUPL, Puell Multiple, Supply-in-Profit. Maps "how far above realized price are we?"
• Momentum (25% weight) — BTC vs 20w/50w/100w/200w MAs, weekly RSI, 30d price change. Maps "how stretched is trend?"
• Derivatives (25% weight) — annualized funding rate, OI/market-cap ratio, 25Δ skew. Maps "how leveraged is positioning?"
• Activity (15% weight) — active addresses, transaction volume, realized cap velocity, Coinbase premium. Maps "how engaged is the user base?"

Reading the zones

When the thermometer diverges from the gauge

Divergences are signal, not noise. The two most common patterns:

Warm thermometer + MILD BUY gauge

The market is stretched but indicators disagree on direction. Usually means momentum and derivatives are frothy while on-chain hasn't confirmed distribution yet. Read: don't add new longs, trail stops on existing. The thermometer is warning about a pullback the gauge hasn't seen yet.

Cool thermometer + WAIT gauge

The market is cold but the framework isn't calling a bottom. This is the classic "early is indistinguishable from wrong" zone. Valuation is attractive, derivatives are clean, but Tier 1 hasn't triggered and AVIV hasn't hit accumulation territory. Read: stay patient. The thermometer is showing the setup is building, but the gauge needs evidence of actual cohort stress before it signals buy.

10Reading cycle patterns

BTC's 4-year halving cycle has been the most predictive structural rhythm in the asset's history. This section explains the rhythm, the typical phase durations, the diminishing extremes thesis, and how to spot which phase you're currently in.

The 4-phase rhythm

Every halving cycle since 2012 has followed the same structural sequence, though the amplitudes have changed. A full cycle is ~1,460 days (4 years) post-halving, divided into:

The diminishing extremes thesis

Each successive cycle has had shallower extremes than the last. This is not a bug — it's maturation. Lower volatility at cycle turns is what institutional adoption looks like.

Notice the consistency: each peak MVRV is ~60-70% of the previous cycle's peak. Each drawdown is ~7-10 percentage points shallower. This is why the framework uses post-ETF thresholds (Tier 1 fires at MVRV < 1.0, not < 0.75 like pre-2022 backtests). The thresholds are calibrated to this decay curve.

The implication for traders: you will never again see 2015 or 2018 style extremes. Waiting for MVRV < 0.80 before deploying means you miss the bottom entirely. The framework's job is to calibrate what "extreme" means in the current regime, not the 2015 regime.

Spotting which phase you're in

Three fast reads:

1. Days post-halving — open the Halving Regime panel. Time alone puts you in a ±3 month window.
2. MA Stack score — 4/4 = markup/distribution. 0/4 = late bear. 1-2/4 = early bear or early markup. 3/4 = mid bull.
3. AVIV Ratio — >2.5 = distribution. 1.5-2.5 = bull. 1.0-1.5 = mid. 0.75-1.0 = accumulation. <0.75 = deep value.

When all three agree on a phase, you know where you are. When they disagree, check the Halving regime calendar — time-based context usually wins in the disagreement (history teaches that cycle timing is more reliable than cycle magnitude).

11Predicting the next bottom

The framework can't tell you the exact bottom price or day. It can tell you when the conditions match historical bottom setups so precisely that acting becomes defensible. This section translates the framework into a forward-looking checklist.

Timing projection

The next cycle bottom, projected purely from the halving calendar:

• Last halving: April 20, 2024
• Historical bottom window: day 930-980 post-halving
• Projected bottom dates: November 2026 - January 2027

This is a time window, not a date. The real bottom could arrive anywhere in this 90-day range, and in exceptional cycles could be 2-3 months early or late. Don't anchor to a single date — anchor to the window.

Checklist: what needs to be true at the real bottom

When all of these are simultaneously true, the framework will read STRONG BUY (+16 to +20). This has only happened three times in BTC's history (2015-01, 2018-12, 2022-11). A fourth occurrence is the target.

What early warning looks like

You won't see all 12 conditions trigger at once. The final low forms over 4-8 weeks, with indicators filling in one by one. The early warning sequence, in historical order:

1. Months out: AVIV drops below 1.0. MA Stack drops to 2/4 or lower. Drawdown passes −50%. Halving regime enters "late bear."
2. Weeks out: Tier 1 starts firing one signal at a time. Usually MVRV < 1 first (fastest to trigger), then NUPL < 0, then SIP < 50%, then Puell < 0.5 last (miner capitulation is always last).
3. Days out: Funding goes deeply negative. DVOL spikes. 25Δ skew goes negative. IV term briefly backwardates. L4 layer goes from +2 to +6 in a single week.
4. At the bottom: Tier 1 4/4 triggered simultaneously with L4 at max bullish confluence. Composite gauge hits +17 or higher. Duration: often 3-7 days of extreme confluence before price turns.

What to do in each phase

12Predicting the next top

Tops are harder than bottoms because distribution is slow and grinding, not a single capitulation event. By the time the top is obvious, 20-30% of the drawdown has already happened. This section explains what to watch for, when to start selling, and how to pace distribution.

Timing projection

The next cycle top, projected from the halving calendar:

• Last halving: April 20, 2024
• Historical top window: day 480-560 post-halving
• Projected top dates: August - October 2025
• Actual top (in this cycle): October 6, 2025 at $124,714 — inside the projected window at day 534

The 2025 top validated the framework's timing model. Day 534 post-halving matched the historical 480-560 window. All prior cycle tops (Dec 2017 = day 558, Nov 2021 = day 576) also fell in this window. This is one of the few deterministic patterns in crypto — use it.

Checklist: what needs to be true at a real top

How distribution unfolds

Tops don't happen in a single day. They unfold over 4-12 weeks as one indicator at a time moves into distribution territory. The sequence is remarkably consistent:

1. 6-8 weeks before top: AVIV crosses 2.0. Funding starts printing persistently positive. LTH SOPR climbs above 1.3. First MILD SELL readings on the gauge.
2. 3-4 weeks before top: NUPL crosses 0.55. MA Stack 4/4 locked. Coinbase premium flips negative. Framework goes to SELL.
3. 1-2 weeks before top: LTH SOPR crosses 1.5 (heavy distribution confirmed). Funding gets stretched above +100% annualized. OI/Mcap climbs above 3.5%. Framework approaches STRONG SELL.
4. Final 3-7 days: 25Δ skew goes extreme positive. DVOL low (complacency). Halving regime clock reads "distribution danger." All indicators at or near max bearish scores. STRONG SELL confirmed.

How to sell — the distribution playbook

13Case studies

Four full post-mortems: the 2022 cycle bottom with its June false signal, the 2018 bottom, the 2015 bottom, and the 2025 top. Each one shows the framework's reading evolving through the turn so you can see what confluence looks like in practice.

Case 1 · 2022 bottom (with the June false signal)

The 2022 cycle bottom is the single most instructive case in BTC history because it included a clear false signal 5 months before the real bottom. The framework's confluence requirement is what separated them.

June 2022 — the false signal

The temptation: "MVRV is below 1! Historical bottom signal!" The framework: "Only 1 of 4 Tier 1 signals fired. Not confluence. Wait."

What happened next: Price fell from $18,500 to $15,476 by November 21. Another −17% for anyone who bought the false signal.

November 2022 — the real bottom

All 4 Tier 1 signals fired simultaneously. L4 at max contrarian setup. L2 at max bear confluence. L1 neutral (M2 had stabilized). Composite +17 — the framework's maximum historical reading.

What happened next: BTC rallied from $15,479 to $73,000 over the next 14 months (+370%). The framework called it within 2 days of the price low.

The lesson

Case 2 · 2018 bottom

2018's bottom was cleaner than 2022 — no notable false signal, no macro headwind, just a long grinding bear that resolved in December.

Day 952 post-halving (July 2016 halving + 952 days). Right inside the historical 930-980 window.

What happened next: BTC rallied to $13,800 over 6 months (+330%), then retraced in a mid-cycle dip before continuing to the 2021 top.

Case 3 · 2015 bottom

The earliest backtest anchor. Halving was in November 2012, so the 2015 bottom at day 974 aligns perfectly with the pattern.

What happened next: +20× in 35 months (to $20,000 at the 2017 top).

Case 4 · 2025 cycle top

The most recent data point. The framework called the top within one week — but with a lower composite score than prior cycles because extremes have been diminishing.

September 2025 — first warnings

October 6, 2025 — the ATH

The diminishing extremes caveat

What happened next: BTC fell from $124,714 to $74,339 by April 2026 (−40.3%). Anyone following the distribution playbook who scaled out 20-30-40 across MILD SELL → SELL → STRONG SELL would have captured ~$112K-$118K average exit price.

14The 155-day cohort convention

This section exists because it keeps causing confusion: how do "LTH MVRV" and "STH MVRV" actually get computed, and why does the endpoint name look backwards?

Glassnode's naming rule

Every Glassnode cohort metric uses the suffix _less_155 or _more_155. The suffix refers to the age of the coins included in the calculation, not the count of coins or anything else:

• _less_155 = coins held less than 155 days = Short-Term Holder cohort (STH)
• _more_155 = coins held more than 155 days = Long-Term Holder cohort (LTH)

So market/mvrv_less_155 is the MVRV ratio computed using only coins that have been held fewer than 155 days — that's the STH cohort, and on our terminal it displays as STH MVRV. Conversely market/mvrv_more_155 is LTH MVRV.

Why 155 days

The 155-day threshold comes from Glassnode's Realized HODL Ratio research and was formalized in the ARK Invest × Glassnode Cointime Economics paper (August 2023). It's the statistical point at which a coin's behavior transitions from short-term-holder-like (volatile, responsive to price) to long-term-holder-like (dormant, price-insensitive). The same threshold shows up in supply/lth_sum and supply/sth_sum with the cohort names spelled out explicitly — Glassnode considers _more_155 synonymous with "LTH" and _less_155 synonymous with "STH" across their entire catalog.

The economic sanity check

If the labels were ever swapped, the numbers would be mathematically impossible given the current price. Consider a world where BTC is trading below the prior cycle's ATH:

• LTHs bought during the prior bear market and early markup — their average cost basis is far below current price, so LTH MVRV should read well above 1.0.
• STHs bought recently, often near the ATH — their average cost basis can be above current price, so STH MVRV can drop below 1.0.

If LTH MVRV ever appears to drop below 1.0, that's a genuine signal — it means even the patient long-term cohort is now underwater, which historically only happens at the deepest capitulation lows (2015, 2018-12, 2022-11). If STH MVRV ever prints above LTH MVRV, something is wrong.

Framework neutrality verification (v2 #18)

Audit v2 #18 questioned whether the framework is structurally bullish-biased. We ran the simplified composite backwards across the full Glassnode history (2010-2026, 5,751 daily rows) to measure the actual average composite per year.

Result: full-period average = −0.285 (median 0). The framework is empirically neutral. Per-year peaks align correctly with past cycle turns: 2015 avg +8.49 (bottom year), 2017 −7.38 (top), 2022 +5.27 (bottom), 2025 −4.08 (ATH year).

Effective range (audit v2 #3): STRONG BUY (≥+16) and STRONG SELL (≤−16) were never hit in 11 years. Practical range is ~−12 to +14, not the theoretical ±24. The verdict bands are calibrated against this effective range in practice.

Band distribution (full period): SELL 7.1% · MILD SELL 28.5% · WAIT 29.0% · MILD BUY 23.3% · BUY 12.1%. Sell-side 35.6% / buy-side 35.4% — within 0.2 percentage points of symmetric.

Verification script: verify_bias.py. Re-run after any scorer change. Last run 2026-04-15.

15Historical validation

The Time Machine (date picker in the header of the terminal) rewinds the entire 4-layer composite framework to any date since 2015. These are the readings the terminal produces at confirmed cycle extremes. Click any card to open the terminal at that date.

The framework reads STRONG BUY at every confirmed cycle bottom and SELL at every confirmed cycle top without any tuning per-date — the scoring bands above are the same bands used for the current live score. This is the point of the Time Machine: the framework is the same at every historical extreme as it is today, which means its current reading is directly comparable to prior cycle extremes.

16Market Pulse Scorecard

A 7-category × 2-axis scorecard borrowed from Glassnode's weekly "Market Pulse" reports. Each category gets a Level (Low / Moderate / High) AND a Direction (Rising / Declining / Flat), giving you a 2-dimensional read per layer rather than a single-number collapse.

Why a second view alongside the composite gauge

The composite gauge collapses 21 indicators into a single −24 to +24 score. That's great for "should I act?" but it loses the cross-layer nuance. The Pulse Scorecard preserves it — you can see at a glance that Fundamental is High/Rising while Capital Flows is Low/Flat, and that disagreement is information that a single-number composite would hide.

The 7 categories

Level scoring: 2-year percentile bands

Each metric's "Level" is computed against its own 2-year rolling percentile distribution using High/Low bands at the 20th/80th percentiles:

• Low — current reading is below the 20th percentile of the last 2 years (statistically depressed)
• Moderate — between 20th and 80th percentile
• High — above the 80th percentile (statistically elevated)

This adapts to regime changes. A funding rate of 60% annualized is "high" in 2025 but was "moderate" in 2021. The percentile band captures that, a fixed threshold wouldn't.

Direction: 7-day delta

Every metric's "Direction" is the sign of its 7-day change: Rising if >+3%, Declining if <−3%, Flat otherwise. Direction catches the velocity that Level alone misses.

Polarity tinting

Different categories have different bullish/bearish polarities at High vs Low:

• Spot · ETF · Fundamental · Capital Flows — High is bullish (green border), Low is bearish (red border)
• Options — High is bullish (contrarian bottom-ish, fear priced in), Low is bearish (complacent top)
• P&L States — Low is bullish (capitulation bottom), High is bearish (profit-taking top)
• Futures — polarity-neutral, direction-dependent

Each category's label also shows a context tag so you don't have to remember the polarity: "High · Demand" (Spot) vs "High · Vol/Fear" (Options) vs "High · Activity" (Fundamental) vs "High · Profit" (P&L States). Same badge color, different meanings explicitly labeled.

How to read it

• All 7 cards bull-tinted → strong alignment, high conviction
• All 7 cards bear-tinted → strong alignment the other way
• Mixed → disagreement itself is information. Read the Divergence Watch panel for the conflict explanation.

17Cohort framework (LTH · ETF · Smart Money)

Post-ETF, "follow smart money" is no longer a single-cohort question. You now have three orthogonal smart-money lenses: on-chain Long-Term Holders (LTH), TradFi institutional via US spot BTC ETFs, and Nansen's historically-profitable wallet cohort. These 3 have different behaviors, different data granularity, and different leading/lagging characteristics.

The 3 cohorts compared

16.1 · LTH Behavior Tracker

The LTH Tracker panel classifies the LTH cohort into one of 6 phases based on 30-day net position change and SOPR:

LTH Net Position Change is the canonical accumulation/distribution metric. Positive 30d = cohort net growing = more STH coins aging into LTH than LTH coins being sold = accumulation. Negative = distribution.

The panel shows a 90-day supply sparkline, 1d/7d/30d/90d deltas, a 5-metric grid (Supply, In Loss %, SOPR, MVRV, NUPL), the current phase classification with historical context, and a validation table running the classifier retroactively against all past cycle inflections so you can verify it calls the right signals at the right times.

16.2 · Cohort Confluence

A derivative panel that reads all 3 cohort stances (LTH from the tracker above, ETF from the Capital Pipeline, Smart Money from Nansen) and computes a single conviction level:

• HIGHEST · 3/3 aligned — rarest, highest-signal configuration. All three orthogonal lenses agree on direction.
• MEDIUM · 2/3 aligned — two cohorts agree, one disagrees. Check the odd-one-out. Usually LTH + SM lead and ETF lags; when ETF is the odd one, the 2/3 is usually early (the leaders are ahead). When LTH is the odd one, the disagreement is a warning (the contrarian cohort isn't confirming).
• LOW · divergent — all three split. No directional signal. Wait for resolution.

The panel shows 3 cards side-by-side (LTH / ETF / SM) with each cohort's current label (e.g., "Accumulation through capitulation", "TradFi buying", "SM deploying stables"), their stance (accumulating / distributing / neutral), and underlying metric. Plus a single verdict line like "3 of 3 cohorts accumulating — contrarian-bull setup".

16.3 · Capital Pipeline

Consolidated "is money flowing INTO or OUT OF crypto?" composite. Only two flows count as true dollar-in/dollar-out events:

1. Stablecoin Supply Δ (USDT + USDC) — mint/burn at the issuer. 1:1 dollar wrapping/unwrapping.
2. US Spot BTC ETF Net Flows — creation/redemption via AP. 1:1 dollar conversion to BTC custody.

Composite = Stablecoin Δ + ETF Flow over 24h / 7d / 30d windows. Shown as a big hero number (the 7d composite) with direction ("capital flowing INTO crypto" / "flowing OUT" / "flat") and 3 time-horizon columns. Context rows below the composite show Futures OI Δ and BTC Exchange Reserve Δ as separate orthogonal signals (not summed into composite because OI mixes dollar + coin margin and exchange reserve is coin custody not dollar flow).

This panel answers one question: "Did new dollars enter crypto this week?" Most serious analysts use stablecoin supply growth as the single cleanest proxy for net dollar inflow. ETF adds TradFi confirmation. Together they're the closest thing to a wholesale capital-in/capital-out measure.

18Bear Market Value Zone (BMVZ)

A price zone between Realized Price (RP) and True Market Mean (TMM) that historically defines where BTC spends most of its accumulation and redistribution time. Breaking above TMM = first structural confirmation of cycle recovery.

The two anchors

• Realized Price (RP) = Realized Cap / Circulating Supply = the aggregate cost basis of the network. Historically the "everyone paid" reference level. Floor of the BMVZ.
• True Market Mean (TMM) = price / AVIV ratio (Cointime Economics). Adjusts RP for dormant supply, giving the "active economic mean" rather than the raw cost basis. Ceiling of the BMVZ.

The AVIV ratio (Active Value to Investor Value) is Glassnode's Cointime Economics metric. When AVIV is low (< 1), TMM is above spot → market is undervalued vs its economic mean. When AVIV is high (> 1), TMM is below spot → market is overvalued.

TMM is live-derived now

Previously the dashboard used a hardcoded TMM = 1.44 × RP multiplier (a common heuristic). That's been replaced with live TMM = current_price / aviv_ratio, which is the actual Cointime Economics definition. The old multiplier was off by 5-10% at any given time; the new calculation is exact.

LTH Cooldown Tracker

Alongside the BMVZ band, the same panel tracks two LTH-specific cool-down metrics:

1. Realized Loss 30d SMA — the 30-day smoothed total realized loss. Historically needs to drop below ~$50M/day for LTH capitulation to be considered cooled. (Note: this is a total-network proxy because Glassnode Professional doesn't expose LTH-specific realized loss flow cleanly.)
2. LTH Supply in Loss — total BTC currently held by LTH wallets that are underwater. Needs to drop below ~1M BTC for full cooldown.

When both cooldown thresholds fire, the LTH cohort has finished its capitulation phase and the cycle floor is confirmed. This is a lagging confirmation, not a leading signal — it tells you "the capitulation is done" after the bottom is already in.

19Divergence & trap detection

Detects market-reversal traps by cross-layer divergence. When price moves one direction but orthogonal signals (Smart Money, exchange flows, sentiment, funding, capital flows, institutional flows) disagree, that's a trap signature. The panel lists active divergences with full evidence — transparent, not opaque scoring.

4 detectable patterns

Evidence-based design

Each flag requires 3+ orthogonal conditions agreeing before firing (severity scales with how many conditions agree). The panel shows the full evidence list so you can see WHICH signals fired. You can read the bullets and decide whether you trust each one for your specific trade — no opaque risk score.

When no flag is active, the panel explicitly says "No active divergences — price and orthogonal signals are aligned. The current move is clean — no trap signature detected in the last 7 days." Silence is useful too.

Caveats

• These are 7-day-horizon signals based on daily data. Not for intraday scalping.
• Multi-source divergence increases probability of reversal but never guarantees it.
• Use as a "conflict checker" alongside the Composite Signal and Tier 1/2 panels — not as a standalone decision.
• Stop-hunt / liquidity-grab detection is not yet implemented (requires hourly data).

20Daily brief system

The terminal ships ONE scheduled alert per day: the morning brief delivered via Telegram at 08:30 WIB. Everything else is intentional silence.

The morning brief

At 08:05 WIB, a single cron runs the full daily pipeline:

1. 08:00 WIB — Wolf Howl nightly retro logs yesterday's autopilot trades (trader-ops only, not surfaced on this dashboard)
2. 08:05 WIB — Glassnode pipeline: collector fetches 60+ BTC metrics, scorer computes composite, Surf collector pulls F&G / ETF per-fund / mindshare / options, Nansen collector pulls Smart Money balances / segmented flows / HL perp trades, Perplexity research runs 5 queries, Perplexity panels fetches catalysts + trending, Grok X-sentiment fetched, and finally Opus 4.6 with extended thinking composes the daily brief
3. 08:20 WIB — Macro layer refresh (US M2 / DXY / SPX / IBIT)
4. 08:25 WIB — Bottom proximity logger appends today's score
5. 08:30 WIB — telegram_morning_brief.py pushes the Opus brief to the private Telegram channel

What the brief contains

• TLDR — 2-3 sentences with the bottom line
• Macro backdrop — M2 / DXY / SPX / Gold / VIX / 10Y context
• Cycle position — MVRV Z, NUPL, Puell, Supply-in-Profit, realized price anchors
• Technical structure — TradingView multi-TF for BTC / ETH / SOL / HYPE
• Risk & derivatives — funding rates, OI, leverage, options skew, VRP
• Institutional flow — per-ETF breakdown, IBIT / FBTC / BITB / etc.
• Network activity — on-chain usage health
• Social sentiment — Grok X-aggregate mood
• Altcoin pulse — brief alt rotation read
• Positioning guidance — per-asset bias / conviction / size / key levels / catalysts for BTC / ETH / SOL / HYPE (machine-readable block)
• Smart money flows (Nansen) — stable rotation, top accumulation, top distribution, per-asset flow decomposition, HL perp trades
• Market context integration — Perplexity-researched catalysts, macro calendar, regulatory events

Average length: ~2,000 words, ~10 minute read. Includes a Telegram-native summary card for quick scan.

In-browser freshness awareness

Every panel shows a small pill indicating its data age and expected cadence (e.g., 44m · daily, 2m · 5min). A sticky freshness banner under the nav shows overall pipeline health — green when all sources are within their expected cadence, amber when anything is stale, red when multiple sources are very stale. See Data sources & cadence.

Re-enabling alerts (not recommended)

If you want to re-enable intraday alerts, know what you're signing up for: you need to apply hysteresis to prevent tier-boundary thrashing (e.g., BUY threshold +9 / WAIT threshold +5 instead of both at ±8). You also need to either kill the Perplexity per-alert enrichment or fetch fresh per alert — the cached replay is a trust violation. See feedback_no_intraday_alerts.md in the memory system for the full incident retro.

21Time Machine usage

The Time Machine (date picker in the header) rewinds the entire framework to any date since 2015. This is the feature that validates the whole system — you can see what the gauge would have said at every historical cycle extreme without changing a line of code.

How to use it

1. Click the date picker in the header (cyan-framed "TIME MACHINE" block)
2. Pick a date or use one of the preset buttons (2015↓, 2018↓, 2022↓ for bottoms; 2021↑, 21↑↑, 2025↑ for tops)
3. Click GO
4. The entire dashboard rewinds — composite gauge, KPI row, Tier 1/2 panels, Cycle phase, Cointime layer, all 21 indicators — to their values on that historical date
5. Click ↺ live to return to the current state

What it's actually useful for

1. Validating the framework

The framework has been tuned against historical data by design — it would be strange if the historical readings didn't match cycle turns. But the Time Machine lets you verify that the scoring bands, the Tier 1 confluence, the AVIV thresholds, and the verdict math all hold at every historical extreme. Open 2022-11-21 and confirm the gauge reads +17 STRONG BUY. Open 2018-12-15 and confirm +18. Open 2015-01-14 and confirm +19.

2. Calibrating current readings

This is the subtle use case. When the terminal says "MILD BUY +7 today," you don't know how that compares to history. Open Time Machine to a past date that had a similar setup and see what the framework said then. If today reads +7 and 2019-01 (early accumulation) also read +7, you know today is an "early accumulation" reading — not a bottom call yet.

3. Testing the distribution playbook

Open 2021-04-14 (first local top) and the framework reads SELL. Open 2021-11-09 (cycle top) and the framework reads STRONG SELL. You can see how the indicators evolved between those two dates — which ones deteriorated first (NUPL), which held up until the very end (LTH SOPR), which gave the final confirmation (funding + OI/Mcap). This teaches you what to watch for during a real distribution phase.

4. Checking the June 2022 false signal

Open 2022-06-18. You'll see Tier 1 at 1/4 (not 4/4), framework reading MILD BUY +3 (not STRONG BUY), and all the reasons the confluence requirement held the framework back from calling the bottom. This is the single best demonstration of why confluence matters — you can see the framework being right about being cautious.

Limitations

22Data sources & refresh cadence

Every feed on the terminal is listed here with its source, cadence, and what it powers. If a feed is stale, the indicator it powers drops out of the composite and the confidence score reflects it. The freshness banner at the top of the terminal + per-panel age badges surface staleness in real time.

Daily pipeline · 08:05 WIB

Live feeds · 5-minute cron

On-demand

Frontend polling

The terminal frontend calls /api/onchain/* on every 30-second interval. This is a read of already-cached backend data — no external APIs are hit on frontend poll. A user who refreshes the page does NOT trigger a fresh Glassnode/Nansen/Surf fetch; they just reload whatever the last daily pipeline wrote to disk.

Freshness banner methodology

The sticky banner at the top of the terminal inspects file mtimes across every data source and returns a structured freshness map. Each source has an expected cadence (daily / 15min / 5min) and a threshold:

• 5min cadence: fresh if ≤10m old, stale if ≤30m, else very_stale
• 15min cadence: fresh if ≤20m old, stale if ≤1h, else very_stale
• Daily cadence: fresh if ≤28h old, stale if ≤48h, else very_stale

The banner shows the worst status across both groups (daily pipeline + live feeds). Each individual panel ALSO shows a per-panel badge (e.g., 44m · daily or 2m · 5min) so you can see at a glance which panel depends on which cadence. Details accordion (click "details") shows the full per-source breakdown.

Historical mode (Time Machine)

When you rewind via the Time Machine, the terminal reads the actual row at that date from every metric's parquet store — nothing is interpolated or approximated. The same scoring pipeline runs retroactively via cycle_phase.compute_cycle_phase(as_of_date=...), so the composite gauge, Tier 1/2 confluence, phase classification, and all 21 indicators reflect their historical values on that day.

23Glossary

AVIV Ratio

Active Value / Investor Value. A Cointime Economics valuation metric that replaces MVRV Z-Score for the post-ETF era by reweighting supply by coin age.

Backwardation

A state of the implied volatility curve where front-month IV is higher than back-month IV. Traders are paying a premium for immediate protection — fear of a near-term move is priced higher than long-term uncertainty. Historically rare and bottom-coincident (2020-03 COVID, 2022-11 FTX both had brief backwardation at the low).

Bottom Proximity Score

Our composite 0-100% synthesis of bottom-relevant signals: MVRV ratio, NUPL, Puell Multiple, Supply-in-Profit, drawdown depth, LTH MVRV, and Liveliness. Each subscore is weighted by its historical reliability at calling cycle lows. The composite reaches 80-95% at confirmed cycle bottoms (2015, 2018, 2022) and drops to 5-15% at confirmed tops. See §08 Indicator Reference for the component list.

Cointime Economics

Framework developed by ARK Invest and Glassnode (2023) that treats coin-age as the primary dimension for on-chain valuation. Replaces classic MVRV metrics that get distorted by large dormant stashes.

Contango

A state of the implied volatility curve where back-month IV is higher than front-month IV. The normal state. Traders are more worried about long-term uncertainty than immediate moves. Steep contango (back-month IV >> front-month) is a distribution warning.

Delta (Δ)

Options terminology for how much an option's price changes per $1 change in the underlying. A 25-delta call moves 25 cents per $1 BTC move. "25Δ skew" compares the IV of 25-delta calls vs 25-delta puts to measure directional positioning without being distorted by at-the-money traders.

Diminishing Extremes

The observation that each BTC cycle has had milder peaks and shallower troughs than the previous. 2011 cycle drawdown −94%, 2018 −84%, 2022 −77%, 2026 tracking around −40%. Peak MVRV has decayed 5.8 → 4.5 → 3.9 → 2.4. This is why the framework's thresholds are calibrated to post-ETF ranges rather than pre-2022 ranges. See §05 Framework design rationale and §13 Case studies.

DVOL

Deribit's BTC volatility index. The "crypto VIX" — a 30-day forward-looking implied volatility measurement from BTC options.

Liveliness

Ratio of coindays destroyed to coindays created over the lifetime of the Bitcoin supply. Falling liveliness = LTHs hoarding. Rising liveliness = LTHs distributing.

LTH / STH

Long-Term Holder (>155 days) / Short-Term Holder (<155 days). The 155-day threshold comes from Glassnode research identifying when coin behavior statistically transitions from reactive to dormant.

MVRV

Market Value / Realized Value. The ratio of BTC's market cap to its on-chain realized cap (the sum of all UTXOs valued at the price they last moved). MVRV < 1 means the average coin is underwater.

NUPL

Net Unrealized Profit/Loss. (Market cap − realized cap) / market cap. NUPL < 0 means the network is in net unrealized loss territory — historically rare, only at bear-market bottoms.

Open Interest (OI)

The total number of outstanding derivative contracts (futures or options) that have not yet been settled. Each contract has a long and a short side — OI measures the aggregate size of the derivative market. We use BTC futures OI divided by BTC market cap as the leverage regime gauge. See OI / Market Cap.

Puell Multiple

Daily BTC miner issuance in USD divided by its 365-day moving average. Puell < 0.5 means miner revenue is deeply below its historical norm — historically a bottom signal because it implies capitulation in the miner cohort.

Realized Cap

The sum of all BTC valued at the price they last moved on-chain, rather than current market price. If a coin last moved at $30K, it counts as $30K in realized cap — even if BTC is now $75K. Realized cap is the foundation of MVRV, SOPR, LTH/STH cohort metrics, and Realized Price. It's a measure of aggregate cost basis.

SIP (Supply in Profit)

Percentage of circulating BTC supply that is currently held at a profit. SIP < 50% means most coins are underwater.

SOPR

Spent Output Profit Ratio. The average profit ratio of all BTC spent on a given day. SOPR = 1.0 is breakeven. SOPR < 1.0 means the average mover is realizing a loss.

True Market Mean (TMM)

A Cointime Economics variant of Realized Price that reweights supply by dormancy. Unlike raw Realized Price, TMM removes dormant supply from the calculation — so long-lost or deeply-held coins don't drag the mean down. TMM typically sits above Realized Price by 30-80% and acts as a practical support/resistance level on the price chart.

Vaultedness

Cointime metric — the fraction of supply that is "vaulted" (aging without being spent). High vaultedness = hoarding regime.

Zero-FP Tier 1

Four indicators (MVRV < 1, NUPL < 0, Puell < 0.5, SIP < 50%) that have only ever all triggered together at the final lows of Bitcoin cycles (2015, 2018, 2022). "Zero false positive" refers to the all-four-simultaneously confluence being historically unique to cycle bottoms. Partial triggers (1–3 of 4) can occur in mid-cycle corrections.

24Disclaimer