Our research team publishes notes on systematic strategies, market microstructure, and the intersection of machine learning and finance. Shared openly because rigor compounds.
A practical note on reducing noisy reinforcement-learning rewards by separating execution quality, market drift, and risk-adjusted outcome signals before policy updates.
A framework for splitting trade decisions into hierarchy levels so signal selection, sizing, TP/SL placement, and exit timing can be optimized without collapsing into one brittle rule.
How volatility, funding, and liquidity regimes change the usefulness of common signal families, and why fixed thresholds decay quickly in perpetual futures.
A research note on making backtests less optimistic by charging spread, slippage, latency, and order-management costs before comparing policy variants.
A portfolio construction note on limiting aggregate risk when multiple models express related momentum, carry, or liquidity-premium exposures.
A compact checklist for graduating a promising research signal into a live candidate with data coverage, risk controls, and monitoring in place.