One thesis started it all
The research we do today becomes the systems we deploy tomorrow. Black Quantum itself began with a doctoral thesis, submitted at Oxford in 2016, on the correlation of markets that never trade at the same instant. Everything the lab does now, the principle, the problem pages, the code that runs live, grows out of it.
The problem gets a name
Thomas Epps, studying stock data, documents something strange: correlations between related stocks shrink as the measurement interval shrinks. The finer you look, the less the market seems to hang together. For decades the effect is treated as a nuisance to be worked around, usually by sampling coarsely and quietly throwing most of the data away. The cause sits in plain sight. No two assets trade at the same instants, and every method of forcing them onto a shared clock injects bias of its own.
Take the clocks seriously
The doctorate begins at the University of Oxford in 2008, supervised by Peter Clifford and Brian Ripley, in the same months the financial crisis is making the question of how markets move together anything but academic. Yang Azzollini takes the problem on directly. The doctorate asks a plain question: what happens if you stop fighting the observation times and condition on them instead, the way regression conditions on its design? The answer is unusually clean. For a whole class of estimators built on the raw asynchronous clocks, variances that the literature knew only in the limit become exact formulas, valid at any sample size, on any given day.
The thesis, Correlation methods in the statistical analysis of financial trading data, was submitted in 2016. The idea at its centre never let go.
The estimators meet real order books
Then a decade on the other side: production trading systems, real order-by-order data, live capital. Industry is a harsh referee. A method that merely looks good in a paper does not survive contact with a matching engine, and the years taught a discipline that now defines the lab: a backtest is a claim, not a fact. Positive results are treated as bugs until they survive walk-forward testing, leakage audits, and replication, because most of them do not.
The same years confirmed the doctorate's premise from the practical side. The engineers' fixes, resample, interpolate, carry forward, are exactly the operations that corrupt the statistics. The theory and the practice point at the same culprit: the artificial shared clock.
Black Quantum
Black Quantum is the return to research, carrying the industry discipline back into statistics. The founding idea now has a name, the observed-clock principle: condition on the observation times you actually saw, and inference on how two processes move together becomes exact rather than asymptotic. The lab runs it as a programme. State each problem on real data, one page and one figure at a time. Prove what can be proved, and machine-check the proofs that matter most. Carry every method through to something that runs live.
The same problem, everywhere
The surprise of the past year is how far the problem travels. Glucose monitors and heart-rate sensors, telescopes and ice cores, networked machines with drifting clocks: almost everything measured in the wild is measured asynchronously, and almost every field still forces a shared clock and hopes. Finance happens to be where the rigorous tools were built first. Our work now is to state the problem beautifully in each new field, and to carry the tools across.
The chart from 1979 still opens our first problem page, drawn from real trades. It is still the best one-picture argument we have, and the passion for answering it exactly has stayed with us, from the thesis, through the trading floor, to this very day.