Research

Working papers and studies.

We publish once an idea is finished with: a design we are ready to discuss, or a study whose outcome is settled. Live strategies stay private.

By area

Agentic trading

Can self-learning AI agents develop trading strategies of their own, and how would anyone know? We study the question at the level of a firm: a population of agents with careers, scored on forecasts rather than luck.

Prediction markets

Short-dated event markets are quoted by a handful of makers and move when faster markets move. We study how they are quoted, where resting orders sit in the queue, and how quickly quotes react.

In review

Resting deep: maker orders in five-minute crypto prediction markets

How deep resting orders behave when large orders sweep the book, studied with a pre-registered forward test and paper trading only.
Soon

Solana latency

On Solana a block is produced every 400 milliseconds by a rotating leader. Where a transaction lands inside that window decides what it is worth. We measure it directly.

In review

Inside the slot: measuring where our transactions land on Solana

Landing rate and position in the block, measured with transactions that land on chain but never trade.
Soon

Emerging venues

New chains and sequencer designs change who gets to trade first. We look at ordering rules, data feeds and what a new entrant can and cannot reach. Notes in this area are in preparation.

Method

How we run research: what gets written down before a test, how results are recorded, and what it takes for an idea to reach capital.

In review

Pre-register, test, kill: how we decide what gets capital

The research loop behind every study on this page, and the rules that stop us trading a backtest.
Soon