Position sizing that survives a bad month

Sizing kills more accounts than bad picks. A trader with a genuine three percent edge who stakes twenty percent of the bankroll per position will very likely go broke before the edge pays out.

The Kelly criterion gives the growth-optimal stake, but full Kelly assumes your probability estimate is exactly right. It never is. Practitioners use a fraction, typically a quarter or a half of Kelly, which sacrifices a little growth for a large reduction in drawdown.

Flat staking, where every position risks the same one to two percent, is not sophisticated but it is close to unbreakable. If you are early, use it. Move to fractional Kelly only after you have a few hundred resolved positions and a calibration record you trust.

Watch correlation. Five positions on the same team, the same sector or the same macro thesis are one position wearing five hats. Add up the exposure before you decide whether you are diversified.

Finally, define your stop before you need it. A pre-committed monthly loss limit that pauses trading is worth more than any model, because it protects the one input everything else depends on: the bankroll.