← Research

CRYPTO / MARKET STRUCTURE · Active Research System · Preview

On-Chain Market Intelligence

Investigating liquidity, capital flows, and market structure through on-chain data.

The question

“How can on-chain holder behavior, exchange capital flows, macro liquidity, and narrative rotation be synthesized into an explainable market-regime classification?”

Context

Digital asset markets are notoriously reflexive, yet public blockchains expose granular structural data—such as realized profit and loss, cohort holding duration, and exchange net flows—that rarely exist in traditional asset classes.

Taken individually, single on-chain indicators frequently fail when market structure shifts. This project explores how combining on-chain behavior with macro liquidity and capital flow layers can classify the prevailing market regime more transparently.

Approach

Instead of treating on-chain metrics as standalone price predictors, the pipeline organizes signals into four distinct layers: Macro Liquidity, Capital Flow, On-Chain Holder Behavior, and Narrative Strength.

Each indicator casts an explicit, auditable directional vote. The system aggregates these votes into a seven-state regime ontology while reporting source coverage and confidence.

Method

How signals are organized and read.

Layer 01 · Macro Liquidity Context

External monetary and risk conditions affecting digital assets

Rates trajectory, dollar liquidity proxies, cross-asset risk appetite

Contextualizes whether crypto market activity is supported by broader macro tailwinds or fighting liquidity headwinds.

Layer 02 · Capital Flow & Exchange Pressure

Net capital entering or exiting trading venues

Exchange net flows, liquidity concentration, stablecoin and wrapper flow proxies

Tracks immediate supply-demand imbalances on spot and derivative venues.

Layer 03 · On-Chain Cohort Behavior

Realized profit/loss and holder rotation

SOPR (7D MA), RHODL Ratio, Long-Term vs Short-Term Holder (LTH/STH) balance, Puell Multiple

Identifies whether long-term holders are accumulating, distributing into strength, or realizing losses.

Layer 04 · Narrative & Network Activity

Internal rotation and network fee pressure

ETH gas regime, relative asset rotation, narrative dispersion

Measures whether risk appetite is broadening across the ecosystem or compressing into Bitcoin.

System

Pipeline.

A Python and Dune SQL analytical pipeline that queries blockchain and market series, evaluates transparent rule logic across four composite layers, and renders an HTML regime report with probability distributions.

  1. 01Fetch on-chain indicator series and verify data freshness / source coverage
  2. 02Score individual indicators (SOPR, RHODL, Exchange Net Flow, LTH/STH, Puell, ETH Gas)
  3. 03Compute composite layer scores and map to a 7-state regime probability distribution
  4. 04Export explainable HTML regime report detailing every vote and underlying value
Inputs
Dune SQL query executions for on-chain cohort and flow metrics · Cross-asset market series via yfinance · Rule-weight configuration and regime ontology definitions
Stack
Python · Pandas · NumPy · Dune SQL · yfinance · Static HTML
Output
Explainable Regime Classification Report & 7-State Ontology Matrix

Output

What the system produces.

ON-CHAIN MARKET STRUCTURE ENGINE // REGIME ONTOLOGY

Regime StatesTransition · Compression · BTC-Led Risk-On · Early Rotation · Liquidity Expansion · Narrative Expansion
Composite LayersMacro Liquidity · Capital Flow · On-Chain Behavior · Narrative Strength
Tracked IndicatorsSOPR (7D MA) · RHODL Ratio · Exchange Net Flow · LTH/STH · Puell Multiple · ETH Gas
Transparency RuleEvery composite score exposes its individual indicator votes and data source state
Uncertainty HandlingHeuristic probability distribution across candidate regimes rather than a single deterministic label

Explainability is treated as a core design requirement: weak or missing data feeds are surfaced directly in the report rather than hidden behind a black-box score. This shows the structure of the output, not a historical backtest.

Limits

What the model does not claim to know.

  1. 01Live execution of full on-chain layers depends on active Dune API credentials and query execution quotas; some inputs fall back to proxy series when offline.
  2. 02Regime probability distributions are heuristic rule mappings designed for structural clarity, not statistically calibrated probability forecasts.
  3. 03Rapid off-chain leverage flushes in perpetual futures markets can move prices faster than daily on-chain settlement data reflects.

Related

Back to the collection

Research