Notes on Oracle-Settled Markets

lagproof@newsletter.paragraph.com (Lagproof) — Tue, 21 Apr 2026 21:12:28 GMT

Oracle-settled derivatives — perps, options, parimutuel binaries, and lending liquidations — now clear billions of dollars per week against a handful of on-chain reference prices. The economic properties of these markets depend on two things almost no one measures directly: how fast the price actually arrives, and who can act on it before everyone else can.

This publication collects my working notes on those two questions.

Research focus

Publication latency. The interval between off-chain price formation (CEX matching engine) and on-chain availability to consuming contracts. Across Data Streams, Price Feeds, and proprietary DEX oracles.
Cross-venue basis. How aggregated oracle prices diverge from the CEX order books they're built from, and what that divergence implies for liquidations and settlement finality.
Microstructure of on-chain binaries. Order-flow, queue priority, and adverse selection on Polymarket-style CLOB markets that settle against Chainlink feeds.
Preconfirmation economics. Base flashblocks and Arbitrum Stylus preconfirmation as a source of directional information ahead of block finalization.

Methodology

Empirical, reproducible, and open where possible. Working stack:

Rust + Python. Alloy + reth archive nodes on Arbitrum, Base, and Polygon.
Full L2 orderbook logger (~30M records / day across BTC, ETH, SOL 5- and 15-minute markets).
Base L2 flashblocks observer with per-pool price series.
Historical Binance, Coinbase, and Deribit tick data for cross-venue reference.
Cached gamma-api / Polymarket settlement data for outcome labelling.

Results are published as short research notes, with code and data links where licensing permits.

What's in progress

A public latency dataset for Data Streams publication → settlement consumption across Arbitrum and Base, benchmarked against CEX tick data.
A falsification-first series: testing claims that get repeated about on-chain microstructure (e.g. "flashblock DEXes lead CEXes on BTC") against real data. Early results: most fail.
Microstructure decomposition of a single high-performing wallet class on Polymarket 5-minute binaries — what's edge, what's execution, what's noise.

Why this matters

Published settlement-latency analysis is almost entirely produced by the oracle vendors themselves or by protocols that consume their feeds. Independent measurement matters for the same reason independent security research matters: the biases are different, and the failure modes show up in different places.

Contact

Open to collaboration with protocols integrating Data Streams, researchers working on oracle security, and anyone publishing on-chain data openly.

Reach me at jay.kichodhan@gmail.com