Top Movers

Pearson correlation of daily returns on the days both legs traded over the last 92 days; a pair needs 30+ common days and must be negatively correlated in both halves of the window with a Fisher z of 5 or more; only vehicles that trade on 70%+ of their sessions and move at least 0.2% a day are candidates (two stablecoins hedge nothing). The long and short book of one index are excluded; each vehicle appears once. Box 30D = the long leg's 30-day return minus the short leg's, per unit of one leg's notional (hover for the figure on both legs' capital), before funding and borrow costs; the direction shown is the one that paid, in hindsight. Past co-movement is not a guarantee. API: /api/v1/universe/movers/anticorrelated

A screen over stored closes, not a recommendation. The hit rate is measured after the fact on picks recorded before their outcome was known; most picks do not double and past results never guarantee future ones. Live picks are recorded the day they are made and scored a week later; the backtest reconstructs the same rule on stored closes and is reported separately.

The same crash-bounce screen restricted to vehicles a $100k book could actually buy. Measured after the fact on picks recorded before their outcome was known; most picks do not double and past results never guarantee future ones.

A scoreboard, not a recommendation. Entries lock the stored close on the round day and are scored once on the real 7-day return. No prizes are attached and nothing moves money.