← Research logStat-Arb & Pairs
PARTIAL · sizing law
#

Sizing each trade by a fraction of the Kelly-optimal bet maximizes long-run log-growth without the ruinous variance of full Kelly.

Optimal fraction from edge $b$ and win prob $p$:

$$ f^\* = \frac{bp - (1-p)}{b} $$

We deploy a de-rated fraction $\kappa f^\*$ with $\kappa\approx 0.25$–$0.5$.

Apply quarter-to-half Kelly using empirically estimated (not assumed) edge/odds per strategy; compare growth and drawdown vs fixed-fractional.

Full Kellyhighest growth, brutal DD
Quarter–half Kellynear-max growth, sane DD
Sensitivity to edge mis-estimatehigh
PARTIAL EDGE
Adopted as the sizing discipline, not a signal: fractional Kelly captures most of the growth at a fraction of the variance. Critical caveat — Kelly is brutally sensitive to over-estimated edge, so we always de-rate hard and size off conservative, realized estimates.
Kelly assumes you know your edge exactly. Since you never do, the only safe Kelly is a small fraction of it — over-betting a mis-estimated edge is how good strategies go to zero.

We publish the failures too.

This is one of 100+ documented hypotheses. Browse the full lab notebook, or see the strategies that survived into production.