Expected value

Expected value (EV) is the probability-weighted average outcome of a decision:

EV = Σ  p_i · value_i

It's the first gate. If EV is negative, nothing downstream saves you — not clever Kelly criterion sizing, not a tighter stop, not "discipline". You're just losing slowly and feeling responsible about it.

The trap is that EV is only as honest as the probabilities you feed it, and humans are spectacularly good at nudging p toward the answer they want. The antidote is to estimate from data and settlement, not vibes — which is the whole argument of validate before automating.

Positive EV plus disciplined sizing is the entire game. Everything else is decoration.