THE SOURCE STANDARD

The methodology.

One fused number per coin, from every feed the desk runs. This page names every input and its weight, every feed and the coins it covers, how sources are ranked by reliability, how fresh each input must be, what happens when sources disagree, and how corrections are published. What is proven and what is still being validated are labeled separately.

01 / THE SCORE

Inputs and weights

Each component is mapped to -100..+100 first, then weighted. Weights sum to 115, verified against the scoring engine at page build.

ComponentWeightWhat it measures
patterns30pattern-engine score (-8..+8)
tradingview15TV gauge 1D/4H ratings
etf15BTC/ETH spot ETF streak (Farside)
options10Deribit put/call OI (BTC/ETH)
polymarket10crowd upside odds vs dip odds (XRP, BTC, ETH, SOL, DOGE, LINK where real markets exist)
momentum57d price change
funding5perp funding, contrarian (crowded longs = bearish)
orderbook5Kraken L2 bid/ask imbalance
coinbase5Coinbase premium (US bid vs world)
social5Threads crowd net sentiment, hourly scan. Absent when chatter is stale or thin.
history10long-memory structure from each coin's full daily price history: trend strength, momentum, volatility regime, and drawdown position
Honest coverage. Coins without a feed get the weight redistributed across the components they do have. Options and ETF flows are BTC/ETH only. Polymarket covers XRP, BTC, ETH, SOL, DOGE, and LINK where real markets exist. A component that does not exist for a coin is absent from its score, never invented.
02 / SCORE BANDS

Reading the number

The same cut points everywhere: the site, the video scripts, and the paper trading portfolios.

ScoreLabelWhat it means
60 to 100Strong BullComposite leaning hard into the move
20 to 60BullGreen, not screaming
-20 to 20NeutralThe score does not trust the move
-60 to -20BearFade the move
-100 to -60Strong BearThe composite wants out
The site's market table collapses this to three tags: BULL (Bull or Strong Bull), NEUTRAL, BEAR (Bear or Strong Bear). A score is evidence, not a call: the hourly briefing's verdict is a separate, falsifiable claim.
03 / BRIEFING ENGINES

The feeds behind the briefings

The Pulse Score is one input. The hourly briefing also reads these feeds directly. Each row lists exactly which coins it covers; anything not listed is not covered.

FeedWhat it watchesCoin coverage
Whale trackingLarge wallet transfers, labeled to_exchange or off_exchange. Direction labels are behavioral proxies and are still being validated.All tracked coins except XMR (amounts are hidden by protocol design)
CME positioningWeekly regulated-futures positioningBTC, ETH, SOL
CME gapsWeekend futures gaps against spotBTC, ETH
Rug watchTop-holder concentration and large wallet moves on the meme coinsSHIB, PEPE
GasNetwork fee activity as a chain-usage readETH
SOL DEXDecentralized-exchange buy vs sell volume on SolanaSOL
HavenPrivacy-coin rotation measured against BitcoinXMR, BTC
XRP escrowScheduled Ripple escrow unlocks and supply releasesXRP
News sweepHourly research sweep of crypto news wires; breaking headlines carry a timestampAll tracked coins
Coverage is a fact, not a promise. If a feed is down for a cycle, the briefing says so and the missing read is marked as carried from the last run, never silently filled in.
04 / SOURCE TIERS

Ranked by reliability

Not all sources earn the same trust. Higher tiers can overrule lower ones.

Tier 1

Live exchange data and exchange-published records

Kraken spot prices and order book, Coinbase premium, Deribit options positioning, CME positioning and gap data, ETF flow prints from the flow publisher. This is the tape itself or records published by the venue. It is the hardest to fake, so it outranks everything else.

Tier 2

On-chain and analytics feeds

Whale transfer labels, XRP escrow schedules, SOL DEX volume, gas fees, and network metrics from established analytics providers. The data is verifiable on-chain; the interpretation is the desk's, and some of it (whale direction labels) is still being validated.

Tier 3

Prediction-market odds

Polymarket crowd odds. Real money sits behind the quotes, which makes them a useful second opinion, but thin markets can be pushed and the odds are a crowd's guess, never the desk's view.

Tier 4

News and social

News wires and the hourly research sweep, plus crowd sentiment from public social posts. Fast but noisy: headlines get facts wrong and social chatter is checked against the live print before it counts. Anything unverified is labeled rumor, never fact.

05 / FRESHNESS

How fresh each input must be

Stale data counts only if it is labeled as stale. The briefing is composed hourly; anything older than its window below is marked as carried from the last run.

InputFreshness standard
Live pricesFresh snapshot at each hourly compose. Call grading is priced against live Kraken spot.
Order book, funding, momentumMinutely snapshots. Must come from the current hourly cycle or the read is marked as carried from the last run.
OptionsDaily publisher prints. Used with the run they came from labeled.
ETF flowsDaily publisher prints. Streaks count consecutive daily prints only.
CME positioningWeekly report. It describes the week, not the hour.
CME gapsChecked each session. Only weekend gaps are tradable.
Polymarket oddsHourly. Only markets with real liquidity count.
Social sentimentHourly scan. Dropped entirely when chatter is stale or thin, never filled in.
NewsHourly research sweep. Every breaking headline carries a timestamp.
06 / CONFLICTS

When sources disagree

Disagreement is disclosed and lowers conviction. It is never averaged away or hidden.

The tape wins

Live exchange data outranks everything. If social chatter quotes a price the live print contradicts, the chatter is labeled unconfirmed rumor, not fact.

Published records beat crowd odds

Options, CME, and ETF-flow prints outrank prediction-market odds and social sentiment when they point in different directions.

Disagreement is disclosed, not averaged

When two publishers report the same number differently, the briefing says so and treats the read as unsettled. Example: an ETF inflow streak against a conflicting redemption print means the institutional bid is steady but the last-day print is unsettled.

Conflict lowers conviction

Conflicting feeds do not cancel out into a quiet average. The verdict's stated conviction drops and the conflict is named in the briefing text.

Unresolvable before publish goes out flagged. If a conflict cannot be settled before the hourly briefing ships, it ships flagged and is confirmed or corrected in the next briefing.
07 / CORRECTIONS

How errors are corrected

The record is append-only. Corrections point forward.

The ledger is append-only

Every verdict is logged at publish time. Nothing is backfilled. Wins and losses both stay on the call ledger.

Grades never change retroactively

Grading follows fixed rules (a 24-hour window and fixed thresholds), so a corrected fact does not rewrite a past grade.

Corrections are published, not silent

When the desk gets a fact wrong, the next hourly briefing carries the correction explicitly, and this page's updated date moves when the copy changes.

Past briefings are not rewritten

The archive is the record. Corrections are made in the next briefing, not by editing the old one.

08 / GRADING

How calls are scored

Every hourly briefing publishes one verdict per coin (bullish, bearish, or neutral) with a stance, a conviction level, and an entry price. Each is logged before the outcome is known.

The window

Each call is graded 24 hours after it is made, against live Kraken spot. The 24-hour grading delay is fixed and automatic.

Win / loss / push

Bullish wins if the 24h forward return exceeds +1.0%, loses below -1.0%, otherwise pushes. Bearish is mirrored. Neutral calls never win or lose: they are reported as pushes.

Kill levels

Each call may carry a kill price. A long dies if the 24h low touches it, a short if the 24h high does. Killed calls count as losses and are tracked in their own bucket.

Conviction weighting

Win rates are also reported conviction-weighted: high counts triple, medium double, low single. Sizing the score to the stated confidence keeps big claims honest.

Anti-cherry-picking. Every verdict goes into the ledger at publish time. Nothing is backfilled. Wins and losses both stay on the call ledger.
09 / VALIDATION STATUS

What is proven, what is not

Every claim below is labeled: proven, or still being validated. The desk does not blur the line.

Proven: the ledger exists

Calls are logged before outcomes, graded on a fixed rule, and published with wins and losses. That is a process claim, and the process is running.

Being validated: the score predicts

The Pulse Score's predictive value is unproven at any sample size. Paper trading two strategies against the published scores is the live test, shown on the track record page.

Being validated: whale direction labels

to_exchange / off_exchange labels are behavioral proxies from a small verified wallet list. Their accuracy against subsequent price action has not been audited yet.

Being validated: small samples

A 60% win rate on 25 calls can easily be luck. Early numbers are a baseline being built, not a verdict. The longer the ledger runs, the harder it is to fake.

10 / HISTORICAL INTELLIGENCE

It remembers every day.

The tenth signal. Most services read the last few candles. This desk studies the full history of every coin it covers, cross-checked against what was happening in the world at the time, and grades every call it makes so the record keeps it honest.

The archive

34,668 daily candles across 12 tracked coins, extended every day with each newly closed candle. Bitcoin runs from September 7, 2011; Ethereum from May 18, 2016. Every candle comes from Coinbase, Kraken, Binance.US, and Bitstamp spot markets.

The world timeline

Price history is cross-checked against a sourced timeline of 561 major events in crypto and world affairs, 2008 to 2026: halvings, ETF approvals and denials, exchange collapses, hacks, elections, and central bank rate decisions. Every event carries its source. A move is never explained by an event the desk cannot cite.

The tenth signal

Each day the engine reads trend strength, momentum, volatility regime, and drawdown position from a coin's full history and folds it into the Pulse Score as its own weighted input. It is one voice in the composite, not the verdict.

The calibration loop

Every published call is graded 24 hours later against the live print, and the grades feed a daily calibration log that tracks where the desk runs overconfident. The ledger is public on the track record page, wins and losses both.

Honest limit. A long memory does not prove predictive power. The track record page shows what the graded record actually says, and the predictive value of the score is listed above as still being validated.

Methodology updated Nov 21, 2026. Weights sum to 115, verified against the scoring engine.