⚠ A replay is a single historical instance — an anecdote, not proof. The Signals tab is the statistical view. Don't generalize from one replay, and remember past reads carried the same low confidence.
Time Machine
Pick a past moment — see what the read was, then reveal what price actually did next
—
—
Price around selected moment
What happened next
⚠ CVD measures taker (aggressor) flow only — the resting maker side is invisible. Read it as pressure, not net positioning.
⚠ TWAP & AF flows are programmatic (protocol/scheduled), not crowd conviction — and are warming until the native indexer lands.
Net taker flow (amber, left axis) vs HYPE price (blue, right axis) — CVD rising with price = confirmation; diverging = exhaustion signal.
Perp vs Spot CVD 24h
Derivatives (perp) vs spot net taker flow — divergence flags who is driving: leverage or real cash.
Programmatic Flow — Net TWAP & AF
Scheduled net TWAP bars + AF protocol buy rate line — warming until native indexer lands; empty state is expected.
⚠ L/S ratios count heads, not capital — laggy and gameable; treat as folklore until Signals scorecard proves IC. Retail ratios = contrarian (fade); top-trader position = smart money (follow).
⚠ CEX OI is a separate pool (~⅓ of HL perp OI) — never add it to HL OI.
Leverage & Positioning
Funding · premium · basis · open interest (HL vs CEX) · long/short ratios
Loading…
Charts
Funding Rate & Premium
Bars = funding (red when longs pay = bearish lean). Line = premium. Persistent positive = crowded longs.
HL OI vs CEX OI — Separate Pools
HL OI (blue) and CEX OI (amber) — distinct pools, never summed.
Crowd L/S Ratios
Dashed = 1.0 balance. >1 = crowd net long (fade). Agg Long % on 0–1 scale (pivot 0.5).
⚠ Staked supply can exceed circulating — staking includes locked/team allocations, so staked-vs-circulating isn't a simple float ratio.
⚠ Unstake pipeline (queue + maturing) is warming — it needs the native indexer (same stage-3b gap as TWAP).
Supply
Staking, validators, the unlock pipeline, and the assistance-fund balance
Loading…
Charts
Staked vs Circulating
Locked (staked) vs circulating — more staked = less sell-side float.
Net Staking Flow
Net staking — locking (+, bullish) vs unlocking (−, bearish).
Unstake Pipeline
Coins lining up to unlock — rising = forward sell-pressure overhang. Warming until the indexer lands.
⚠ Holder concentration excludes flagged system/protocol wallets (bridge, assistance fund). If that list is incomplete, top-N share is overstated — the full exclusion set is still being populated.
⚠ CEX flows are address-heuristic from a watched wallet set — coverage is partial; an unlabeled exchange wallet is invisible. The wallet set is still being populated (cards warming).
On-chain
Exchange deposit/withdrawal flows and holder concentration — HYPE only
Loading…
Charts
Holder Concentration
Whale ownership by tier — rising = fewer hands. Excludes system wallets (overstated if the list is incomplete).
CEX Net Flow
Net HYPE to (+, bearish) / from (−, bullish) exchanges — warming until the wallet set is populated.
CEX Balance
HYPE parked on exchanges — rising = more sell-ready supply.
⚠ Order-book depth & imbalance are gameable — resting walls can be spoofed and pulled. Read imbalance as a hint, not conviction.
⚠ Spread & depth are point-in-time L2 snapshots (one read per poll), not averages — a momentary liquidity picture.
Microstructure
Bid-ask spread, book depth, and order-book imbalance — perp + spot
Loading…
Charts
Spread (Perp + Spot)
Bid-ask gap in bps — tighter = more liquid; spikes = stress/illiquidity.
Book Imbalance (Perp + Spot)
Above zero = bid-heavy (buy-side walls); gameable — read as a hint, not conviction.
Book Depth — Bid vs Ask (Perp)
Resting buy-side vs sell-side size — the raw walls behind the imbalance ratio.
⚠ CEX and HL liquidations are separate pools — never added. A combined "total" would be meaningless; they are shown in separate blocks for this reason.
⚠ Magnitudes undercount — Binance & OKX throttle liquidation reporting to ~1/sec; Bybit is more complete; native-HL is a reconstructed lower bound. Read direction and spikes, not absolute notional. Never sum CEX with HL.
⚠ Forced selling reads bearish and forced buying bullish — but extreme flushes often mark capitulation/reversal. Read big spikes as possible turning points, not just continuation.
CEX — Liquidations
Forced selling and forced buying on CEX venues — partial coverage, never added to HL
Loading…
CEX Forced Flow
Forced selling (red) vs forced buying (green) on CEX venues — a separate pool, never added to HL.
Hyperliquid — Liquidations (separate pool)
Forced selling and forced buying on Hyperliquid — reconstructed lower bound, warming
Loading…
HL Forced Flow
Forced selling (red) vs forced buying (green) on Hyperliquid — warming until native indexer lands.
Blow-up Events — HL vs CEX
Liquidation event counts by venue — counts only; magnitudes are not comparable across venues.
⚠ IC is warming — clean hourly pairs only started accumulating recently; treat magnitudes as provisional and nothing as proven until n is large and |IC| ≥ 0.10.
⚠ Weak or zero IC means the signal doesn't predict price here — that's a finding, not a bug. Only a strong IC pointing the opposite way from a signal's mapping is a red flag.
Signal Scorecard
Spearman IC per scored signal × 4 horizons — grouped by expected direction
Signal
Dir
4h
24h
72h
168h
Check
Loading…
⚠ HYPE only · execution is testnet but market data is read from mainnet · hourly snapshots. Per-metric caveats (what's gameable, what's a separate pool, what undercounts) live in the Glossary.
Every metric defined — plain English, source, directional reading, and methodology
Concepts
Long / Short — betting price goes up (long) or down (short).
Perp — perpetual futures contract; tracks an asset's price without expiry.
Spot — the actual asset (real coins), not a derivative.
Leverage — using borrowed capital to amplify a position.
Liquidation — forced close of a leveraged position that hit its margin limit.
Funding rate — periodic payment between longs and shorts to keep perp price near spot.
Open interest (OI) — total value of outstanding leveraged positions.
CVD — Cumulative Volume Delta; net aggressive (taker) buying minus selling.
TWAP — Time-Weighted Average Price; a scheduled order that executes evenly over time.
Assistance Fund (AF) — Hyperliquid's protocol fund that auto-buys HYPE from trading fees.
Unstaking queue — coins waiting to unlock from staking (forward sell-pressure overhang).
Basis — the gap between spot and futures prices; signals crowding.
Book imbalance — ratio of resting buy vs sell orders; hints at near-term pressure.
Long/short ratio — proportion of accounts positioned long vs short (crowd positioning).
IC (Information Coefficient) — Spearman rank correlation between a signal's value and future price return. Measures predictive power.
Hit rate — fraction of times the signal's direction matched the subsequent price move.
BPS — basis points; 1 bps = 0.01%.
PP — percentage points; the arithmetic difference between two percentages.
Confidence — stays LOW until signals are IC-proven (|IC| ≥ 0.10 over enough hourly pairs). HIGH means the signals have demonstrated predictive power.
Reading the data
The bias
The bias is an IC-weighted vote of the signals for a chosen horizon, shown with a confidence flag that stays LOW until signals are IC-proven (|IC| ≥ 0.10 over enough hourly pairs). Weak IC means the signal doesn't predict price here — that's a finding, not a bug.
The contrarian convention
Crowd/retail positioning and exchange-deposit signals read contrarian (high = bearish, fade). Order flow, smart-money positioning, accumulation, and withdrawals read with the move (high = bullish).
Cross-cutting data caveats
CEX OI and HL OI are separate pools — never summed.
CEX and HL liquidations are separate pools — never summed; magnitudes undercount.
L/S ratios count heads, not capital — laggy and gameable.
Order-book depth/imbalance is gameable (spoofable walls).
CVD is taker (aggressor) flow only; the maker side is invisible.
Holder concentration excludes flagged system wallets (overstated if that list is incomplete).
TWAP / AF / unstake-pipeline metrics are warming until the native indexer lands.
Execution is testnet; all market data is read from mainnet. On-chain layer is HYPE-only.