🔬 Signal Methodology
How every signal on Algo Tick is computed, what data feeds it, where it works, and where it fails. Transparency is not optional — if you're going to trade on a signal, you should know exactly how it's built.
Jump to Signal
General Principles
- Same math, live and historical: The computations you see on the live dashboard use the same formulas applied to historical data in backtests
- Dual timestamps everywhere: Every data point has
time_chain(event time) andtime_local(ingestion time) for latency auditing - Unsupported = unpublished: Signals are only shown for assets with sufficient underlying data. No fake pages for unsupported combinations
- Failure modes disclosed: Every signal lists known limitations, edge cases, and scenarios where it degrades
📈 Market Regime
Every 60 secondsA Hidden Markov Model (HMM) classifier that labels the current market state as risk-on, risk-off, or neutral based on multi-factor inputs.
P(regime | observations) via Baum–Welch algorithm; regime = argmax P(state)Three states: risk-on (trending up), risk-off (trending down), neutral (range-bound)
Risk-on regimes favor momentum strategies; risk-off regimes favor mean-reversion or hedging; neutral regimes favor range-bound plays.
- Price momentum (multi-timeframe)
- Volatility regime
- Funding rate
- Orderbook imbalance
- Cross-asset correlation
Regime transitions can lag fast market moves by 1–5 minutes. The model can oscillate between states during choppy markets.
/v3/signals/regime?coin=BTC/v1/history?coin=BTC&metric=alpha_regime_prob&hours=24🧭 Composite Signal
Every 60 secondsA multi-factor directional signal that combines orderbook imbalance, funding rate, volatility regime, correlation, and whale flow into a single score.
composite = Σ(wᵢ × normalized_signalᵢ) where weights are regime-adaptive−1.0 (strong sell) to +1.0 (strong buy)
Readings above +0.5 are bullish. Below −0.5 are bearish. The signal is designed to have low false-positive rate at extremes.
- Orderbook imbalance
- Funding rate
- Regime probability
- Whale flow
- Correlation breaks
During black swan events, multiple input signals can fail simultaneously. The composite degrades gracefully to fewer factors.
/v1/signals/composite?coin=BTC/v1/history?coin=BTC&metric=alpha_regime_prob&hours=24📊 Orderbook Imbalance
~1–5 secondsMeasures the relative strength of buy vs sell pressure in the L2 orderbook within 1% of the mid-price.
imbalance = (bid_depth − ask_depth) / (bid_depth + ask_depth)−1.0 (all asks) to +1.0 (all bids)
Values above +0.3 suggest aggressive buying pressure; below −0.3 suggest selling pressure. Extreme readings (>0.6) often precede short-term price moves in the same direction.
- L2 orderbook snapshots (20 levels, ~100ms)
- HyperLiquid exchange data
Low volume periods can produce noisy readings. Spoofing (placing/cancelling large orders) can temporarily distort the signal.
/v1/signals/imbalance?coin=BTC/v1/history?coin=BTC&metric=hlp_orderbook_imbalance_pct&hours=24💰 Funding Rate
~5 seconds per coinThe periodic payment between long and short perpetual futures holders, reflecting the cost of leverage and market directional bias.
funding_velocity = Δ(funding_rate) / Δt; z-score = (rate − μ₁ₕ) / σ₁ₕTypically −0.01% to +0.01% per 8h cycle. Extremes can reach ±0.1%+.
Positive rates mean longs pay shorts (bullish bias). Extreme positive rates often precede local tops. Negative rates often precede local bottoms or short squeezes.
- HyperLiquid funding rate stream
- Mark price and open interest
During low open interest, funding can be volatile without directional meaning. Rates normalize quickly after liquidation cascades.
/v1/signals/spreads?coin=BTC/v1/history?coin=BTC&metric=alpha_funding_zscore&hours=24⚡ Volatility Analysis
Every 30 secondsRealized volatility computed from tick-level trade data, compared against implied volatility from Deribit options to identify volatility premium or discount.
RV = σ(log returns) × √(252×24×60); IV from Black-Scholes inversion of ATM optionsAnnualized: typically 30–120% for BTC. IV − RV spread can be −20% to +40%.
When IV > RV (positive premium), options are expensive — volatility sellers profit. When IV < RV, the market is underpricing risk.
- Tick-level trade data
- Deribit options chain
- Gamma exposure profile
Short lookback windows can miss slow regime shifts. IV data is only available for BTC and ETH.
/v1/signals/volatility?coin=BTC/v1/history?coin=BTC&metric=alpha_funding_zscore&hours=24💥 Liquidation Cascades
Every 30 secondsDetects zones where concentrated liquidation risk exists by analyzing open interest distribution, leverage patterns, and recent liquidation events.
cascade_risk = f(OI_concentration, leverage_distribution, recent_liq_volume, price_proximity_to_liq_zones)0.0 (no risk) to 1.0 (imminent cascade)
Readings above 0.5 indicate elevated cascade risk. Above 0.8, expect significant forced selling or buying within hours.
- Open interest by price level
- Recent liquidation events
- Funding rate extremes
- Price proximity to high-OI zones
Cannot detect OTC liquidations or cross-exchange arbitrage unwinds. Post-cascade, risk readings drop rapidly.
/v1/signals/liquidations?coin=BTC/v1/history?coin=BTC&metric=hlp_liquidation_alerts&hours=24🐳 Whale Flow
Every 10 secondsTracks large order placement and execution from the L2 orderbook. A 'whale' event is any order or cluster of orders exceeding $100K notional within a 10-second window.
whale_score = volume_above_threshold / total_volume (rolling 5min)0.0 (no whale activity) to 1.0 (dominated by whales)
High whale flow on the bid side is structurally bullish. High whale flow on the ask side signals distribution.
- L2 orderbook depth changes
- Trade-by-trade execution data
- Order size clustering
Iceberg orders (hidden liquidity) can cause false negatives. Wash trading can inflate whale scores.
/v1/signals/whale-flow?coin=BTC/v1/history?coin=BTC&metric=hlp_orderbook_imbalance_pct&hours=24📐 Gamma Exposure (GEX)
Every 30 secondsDealer gamma exposure computed from Deribit options open interest. Positive gamma means dealers are long gamma (dampening moves); negative gamma means they amplify moves.
GEX = Σ(OI × gamma × contract_multiplier) across all strikes; gamma_flip = price where GEX crosses zeroLarge negative = amplified moves; large positive = suppressed moves
Below the gamma flip level, price moves are amplified (dealer hedging creates positive feedback). Above it, moves are dampened.
- Deribit options chain
- Open interest by strike
- Implied volatility surface
Only available for BTC and ETH. Model assumes all option sellers are dealers (overestimates hedging pressure). Does not capture OTC options.
/v3/signals/gamma-exposure?coin=BTC/v1/history?coin=BTC&metric=alpha_regime_prob&hours=24🔗 Cross-Asset Correlation
Every 60 secondsRolling Pearson correlation between crypto assets and between crypto and macro indicators, computed from 1-minute OHLCV data.
ρ(X,Y) = cov(X,Y) / (σ_X × σ_Y) over rolling windows (1h, 4h, 24h)−1.0 (perfect inverse) to +1.0 (perfect positive)
Correlation breakdowns (sudden decorrelation from ρ > 0.8 to ρ < 0.3) often precede large moves. BTC-ETH decorrelation is a key regime-change signal.
- 1-minute OHLCV from HyperLiquid
- BTC/ETH/SOL price feeds
Short windows (1h) can produce noisy correlation estimates. Lag between correlation breakdown and price impact varies.
/v1/signals/correlation?coin=BTC/v1/history?coin=BTC&metric=alpha_geo_coherence&hours=24🛡️ Signal Safety (Coherence)
Every 60 secondsMeta-signal that measures agreement across all other signals. When signals agree on direction, conviction is high. When they disagree, the market is uncertain.
coherence = 1 − entropy(signal_directions) / max_entropy0.0 (all signals disagree) to 1.0 (all signals agree)
High coherence (>0.7) means signals are aligned — higher confidence in directional bets. Low coherence (<0.3) means conflicting signals — reduce position size.
- All signal outputs
- Regime state
- Cross-asset correlation
Maximum coherence during strong trends can mask reversal risk. Coherence is lagging — it confirms trends rather than predicting them.
/v3/signals/safety?coin=BTC/v1/history?coin=BTC&metric=alpha_geo_coherence&hours=24Data Collection Architecture
Algo Tick runs a dual-citadel architecture with independent data collection nodes in Frankfurt (EU) and Canada (NA). Each node independently connects to data sources and writes to the shared R2 data lake. This provides:
- Redundancy: If one node fails, the other continues collecting
- Cross-validation: Compare timestamps and values between nodes to detect anomalies
- Geographic diversity: Different network paths catch different market microstructure effects
Related Pages
📋 Coverage
Full inventory of metrics and datasets
🟢 Freshness
Live health and latency monitoring
📁 Data Library
Browse and download historical datasets
Don't just stare at the dashboard. Automate it.
Every metric on this page is available via our sub-millisecond API.
Build trading bots, backtest strategies, and power AI agents with institutional-grade data.