Statistical arbitrage and market making, grounded in probability and causality.
We are a lean team of researchers and engineers who go all the way down: causal models, owned infrastructure, and strategies that only trade when the statistics say to.
What we focus on
LynxLinkage is built around one conviction: crypto's structural inefficiencies are still large enough to reward deep research — but not large enough to forgive sloppy thinking. We work in two disciplines where probability theory and causal inference determine whether you make money or give it back: statistical arbitrage and market making. Every strategy starts as a falsifiable claim about how a market behaves, and every claim must survive statistical scrutiny before it sees real capital.
We are deliberately lean. Every person on the team owns research, code, and live risk simultaneously. There are no hand-offs, no approval layers, and no organisational distance between the person who writes the model and the person who answers for the PnL — because they are the same person.
Statistical arbitrage and market making across crypto.
Each strategy family has its own data requirements, its own causal model, and its own execution profile. We allocate research effort based on capacity, regime, and edge — and we only run strategies that have a theory we can defend.
Cross-venue statistical arbitrage
The same instrument trades on dozens of venues with different microstructures, fee schedules, and latency profiles. We model the spread as a stationary process, test for mean-reversion with rigorous statistical criteria, and size positions from the full residual distribution — not a point estimate.
Basis & funding arbitrage
Spot, perpetuals, and dated futures are linked by hard arbitrage constraints. When funding pressure or basis dislocations break those constraints, there is a clear causal mechanism behind the trade — not just a correlation. We model the reversion speed and hold time probabilistically.
Cointegrated pairs & baskets
Related tokens move in statistically measurable clusters. We apply cointegration tests and error-correction models to isolate pairs and baskets where mean-reversion has a structural cause, a measurable half-life, and a residual distribution that is stable enough to trade.
Market making
We provide liquidity across crypto spot and derivatives venues. Our quoting models are built on arrival-rate distributions, adverse-selection metrics, and real-time inventory risk — not fixed spreads. Execution is a first-class research problem, not an afterthought.
How we work
Five principles that are non-negotiable regardless of market conditions, team size, or how good the last backtest looked.
Statistics, not stories
Every edge is a statistical hypothesis with a null. We write the null before we run the test, report the full distribution of outcomes across regimes, and only go live when the evidence clears a high bar. A great Sharpe on one window is not evidence.
Causality over correlation
A correlated pair is a starting point, never a trade. We require a structural reason — funding mechanics, index pressure, cross-venue constraints — before committing research time. If we cannot write a causal story for why the spread should mean-revert, we do not build the model.
Theory precedes data
Each strategy begins as a written conjecture about how a market must behave given its microstructure. The backtest is the disconfirmation step, not the research itself. If the theory breaks in the data, the theory was wrong, and we discard it without ceremony.
Own the stack
From tick-data capture to risk management to execution — we build what we need. Off-the-shelf tools optimised for equities miss the microstructure details that determine whether a crypto strategy actually works at scale.
Lean by design
Researchers and engineers are the same people. There are no hand-offs, no approval layers, no strategy decks that go nowhere. Everyone owns a strategy end-to-end — from conjecture to live risk. The team stays small on purpose because that is how full ownership works.
- Founded
- 2024
- Markets traded
- Crypto only
- Strategy style
- Stat arb & MM
- Operating hours
- 24 / 7
Where we are
The team is fully remote and based in Taiwan. We organise around overlapping working hours rather than a single office, and we hire wherever the right person is — provided they are comfortable working on Taiwan time and are happy with a flat team where your ideas go straight into production without three levels of sign-off.