Every strategy we run starts as a research question and ends inside the same risk framework. Arbitrage, statistical arbitrage, and a systematic multi-strategy program are three different answers to that process: one closes gaps fast, one closes them on a model's timeline, and one runs trend and mean-reversion side by side.
We don't believe we've found a permanent advantage, only a temporary one. Our job is to find the next edge before this one gets arbitraged away by everyone else looking for the same thing.
A strategy that avoids catastrophic loss tends to outlast one that occasionally has a spectacular year. We size every position for the scenario where we're wrong, not the one where we're right.
We don't try to predict what a market will do next. We build rules designed to hold up across many possible futures, and let position sizing, not conviction, do the deciding.
Every signal has a shelf life we can't know in advance. We plan for ours to decay, and budget accordingly for the research required to replace them before they do.
We look for two prices that should agree and don't: the same instrument trading across multiple venues at once, or a derivative priced away from what it's derived from. We hold the position with the aim of capturing that convergence.
Cross-exchange and cross-venue arbitrage, and funding-rate and basis trades between an instrument and its derivative.
Holding periods vary with market conditions rather than following a fixed rule, and the edge decays the moment the rest of the market notices it, so the research pipeline never really stops.
We model the historical relationship between correlated instruments and trade the moments that relationship breaks down: long the one that's cheap, short the one that's rich, market-neutral in aggregate.
Pairs and basket trades, sized and exited by the model's confidence in the relationship, not by how the position feels.
Holding periods are set by how quickly the model expects the relationship to reassert itself: long enough for the edge to play out, short enough that we're rarely exposed to a single instrument's direction.
A systematic program built from two complementary signal sets: trend-following, which adds to a move once it's confirmed and rides it while it persists, and mean-reversion, which fades short-term overextensions and bets on price reverting back toward its recent average.
The model weighs both signal sets continuously rather than committing to one regime, and positions scale with the strength of whichever signal is active.
Turnover is high and holding periods are set by the model rather than a fixed schedule. Trend positions can run for weeks, mean-reversion positions typically resolve much faster.
Same risk framework, same execution and data platform, applied at three different speeds. Read more about how the firm is built on Who We Are →