Research & insights
Our own exploratory analysis of public data.
We publish what we can measure and demonstrate — and we're careful to separate what is a built case study from what is still a research direction. Everything here is exploratory analysis of public data, not investment advice.
Cross-domain synthesisOne instrument, eleven studies - the honest verdicts side by side
Sharp reference, deviation, lead-lag, reversion vs surrogate, verdict for every study here - structure where it exists, none where it doesn't.
WhitepaperCross-source price integrity - the method, in full
Independent, reproducible measurement of whether reported cross-venue crypto prices are genuine - free public data, method disclosed, nulls published. Downloadable PDF.
WhitepaperDay-ahead vs real-time price integrity - the method, in full
Independent, reproducible measurement of the day-ahead vs real-time electricity price relationship - the day-ahead premium and its hour-to-hour serial dependence, on free public ISO data (NYISO and CAISO), method disclosed, every claim tested against a null. Downloadable PDF.
WhitepaperCross-market coupling integrity - the method, in full
Independent, reproducible measurement of how US Treasury and euro-area AAA 10-year sovereign yields move together - dominant same-day co-movement, and a smaller replicated one-day-lag reported as coupling not leadership, across three independent seven-year blocks on free public first-party data, every null reported. Downloadable PDF.
Cross-exchange price discovery in BTC
Whether the gap between crypto exchanges is a real dislocation or a structural artifact — now tested across six independent windows spanning eight months, the binding force is real but provably too small to trade.
Sovereign rates — cross-market · Sept 2026 Case studyUS–euro-area AAA sovereign 10Y yield coupling
Whether daily changes in US Treasury and euro-area AAA sovereign 10-year yields are coupled — measured across three independent seven-year blocks from 2007 to 2026, and honest that the dominant signal is same-day co-movement while the one-day-lag piece cannot be separated from the mechanical effect of two markets closing four hours apart.
Cloud infrastructure — spot pricing · Sept 2026 MethodologyCross-AZ spot coupling: when the honest answer is 'this design can't tell'
We set out to measure whether one availability zone's spot price reverts because it diverged from the others — genuine cross-AZ coupling — or merely because all zones drift toward a shared regional factor. Before touching a single day of the analysis window, the engine's own controls proved the two are indistinguishable by this design: a reflection problem. We closed it there, rather than run a test that could only manufacture an answer.
Retail — fuel price dynamics · Aug 2026 MethodologyLocal price convergence in French retail fuel — not supported by a geography placebo
Diesel prices across ~9,600 French stations appear to converge toward local consensus — material deviations that mean-revert, replicated across four windows. A geography-scramble placebo finds the convergence is not specifically local: stations revert toward a random département's median as strongly as their own. The honest finding is a downgrade — generic price stationarity, not local competition — plus a real, reference-independent chain-pricing result the engine surfaced along the way.
Crypto — perpetual basis · Aug 2026 Case studyPerpetual–spot funding basis: a six-window replication
We pre-committed to replicating this study, then ran it across six windows. The funding-tethering holds across regimes and the majors stay efficient — while the one apparent structural edge, SOL's mean-reversion, turns out to be specific to a single week.
Crypto — triangular no-arb · Aug 2026 MethodologyTriangular no-arbitrage in ETH/BTC: a six-window replication
We pre-registered a stricter test of our own earlier finding — a second null the first pass lacked — and ran it across six windows. The identity holds to within one quote tick and stays untradable, but the mean-reversion we had read as a binding force turns out to be tick quantization, not a force beyond it.
Energy — day-ahead vs real-time · Aug 2026 Case studyThe day-ahead premium, reproduced on an independent grid (NYISO)
We ran the CAISO day-ahead-premium instrument, unchanged, on a structurally independent grid — NYISO — across six recent months. Both the premium and its hour-to-hour persistence reproduce, turning a single-grid result into a market-structure one.
Crypto — DEX vs CEX lead-lag · Aug 2026 MethodologyDEX versus CEX: does the on-chain price lag the exchange?
The same engine on a different venue type — an on-chain pool against an exchange order book — testing whether the DEX lags the CEX by about one block, and finding the adjustment completes faster than the block clock can resolve.
Energy — day-ahead vs real-time · Aug 2026 Case studyDay-ahead vs real-time: the CAISO power basis
The same engine on a third market — measuring the day-ahead premium in wholesale power and testing whether the real-time-minus-day-ahead basis is predictable, using a null we had to correct for a stationary settlement series.
Crypto — alt cross-exchange lead-lag · Aug 2026 MethodologyCross-exchange lead-lag in a liquid altcoin (SOL)
Testing whether a liquid altcoin shows the cross-exchange lead that BTC lacks — and finding none, down to the fraction-of-a-second resolution free public tick data allows.
Crypto — stablecoin peg · Aug 2026 MethodologyStablecoin peg integrity: USDT, USDC and the $1 anchor
How far USDT drifts from $1, whether independent venues agree on that drift, and whether the discount is a fixed self-correcting level or a moving one — a pre-registered test finds the venues agree closely while the discount itself drifts.