Risk is not a control layer. It is part of the research.

StatMind evaluates every signal through uncertainty, liquidity, sizing, exposure, drawdown, and attribution.

Late, revised, or point-in-time-incorrect data poisons everything downstream. We treat data integrity as the foundation of the stack.

A model can be wrong, miscalibrated, or overfit. We assume it will decay and design monitoring around that assumption.

Even a correct view carries exposure to moves in the underlying event. We measure it before we hold it.

The price we get is rarely the price we saw. The gap between intended and realised fills is measured and budgeted.

A good edge sized wrong is a bad strategy. Size follows calibrated confidence and real liquidity, not conviction.

Individually acceptable positions can compound into unacceptable exposure. Correlation is managed across the whole book.

The final risk is mistaking luck for skill. We decompose results to their source so we know why we won or lost.