Most quantitative research is optimized on a very small portion of the search space. In practice, less than 5% of possible configurations are explored, and even that is often done after resampling the data (1m → 5m → 15m), which introduces structural bias. At the same time, common methods like correlation, covariance, and cointegration only give you estimates of relationships. They don’t tell you how those relationships actually behave in time. We built two tools around this gap. PermuCheck collapses data onto a one-minute canonical axis and runs systematic parameter sweeps without resampling. Patterns across 1, 5, 10, 15 minute structures emerge naturally while preserving temporal order. This lets you explore a much larger, unbiased portion of the statistical landscape under your own thresholds. Gemina works on two time series. It aligns them (even across different markets), then measures exactly: • when they diverge • the magnitude of divergence • whether they converge • how long convergence takes The output is simple: empirical statistics of divergence and convergence under defined conditions. Both are early pilot products. The larger goal is to build quantitative systems where the statistical landscape is not optimized or curve-fit into local optima, but systematically explored — so that multiple viable structures are visible, not just one.

DOJO by Student One

Run large scale permutations of high frequency statistics

Run large scale permutations of high frequency statistics