Onchain

Completed UTC daily observations only. Flows compare equal adjacent totals; stocks compare endpoint values. All intervening days must be present. Flow windows use start-inclusive/end-exclusive UTC boundaries, with exact midnight token prices at both endpoints. Stocks retain their observed daily endpoints. Prices must have a verified provider identity.

7D / 30D / 90D: fundamentals leading requires growth ≥25% and divergence ≥20 percentage points. Price leading requires a price rise ≥20% and fundamentals lagging by ≥20 points. Acceleration requires growth ≥40%; deterioration ≤-25%. 1D uses stricter 50% / 30 point / 75% / -35% thresholds.

Growth requires economically meaningful starting and current values. Deterioration uses baseline scale and absolute loss; a valid reported zero can qualify. Minimum starting scales: fees $50K/day; revenue and token-holder revenue $10K/day; DEX volume $10M/day; TVL and stablecoin supply $5M. Freshness follows completed UTC reporting dates, with a 12-hour publication grace for the preceding day. A new fetch never makes an old observation fresh. Local daily collection runs when this page is opened or revisited, with one normal confirmation phase and one conditional noon retry; discovery reserves at most eight fundamental requests normally and four for incomplete noon evidence, plus six missing token histories and two noon retries. If the normal phase never reserved prices, its first noon price confirmation is limited to tokens from the frozen fundamental plan. Valid cached observations are reused before spending price-target allowance; incomplete cached history keeps its missing periods explicit. Hosted reads never collect providers. Viewers check published evidence after the minute-17 collector, with one later check only if evidence remains incomplete. 90D needs 180 days of flow history or 91 stock observations. A stale/missing price withholds divergence categories while qualifying fundamentals can still appear.

Automatic confirmation prioritises meaningful changes with verified token identities, revenue and token-holder revenue, fees and DEX activity, same-direction multi-day comparisons, then daily-equivalent dollar change. At most one large protocol anchor follows the first seven meaningful candidates. Confirmed price targets also exclude dominant daily spikes from persistence. Signal priority: corroboration across multi-day windows, multiple metrics, valuation comparison, absence of a dominant daily spike, percentage magnitude, daily-equivalent dollar change, then period and stable identity. Overview prioritises current reporting periods and reliable evidence, then absolute USD change (flow changes per day), percentage or verified divergence magnitude, corroboration and stable identity. At most one finding per exact entity or known protocol family appears, with no category quotas. Capital findings retain the separate Rotation and Stablecoin thresholds; this presentation order creates no combined financial score. Stale or fallback capital observations stay in the analytical views. A day accounting for over half the latest 7D flow total is marked concentrated. Overlapping windows are corroboration, not independent statistical evidence. Heuristic flags are not validated alpha; USD values can reflect asset prices and protocol revenue may not accrue to holders.

Price-led screening uses exact completed-day Hyperliquid perpetual endpoints: ±10% over 24H or ±20% over 7D. Spot-token evidence is checked separately. Missing 7D baselines remain unavailable. Saved Anchor screening makes no additional provider requests; selected Battle history is bounded and is not continuously collected. Provider definitions and child/parent coverage differ. Families are grouped without summing metrics. No persistent alert history or “new” labels; no notification service. The homepage is outside this iteration.