Week 11 Problem Session: Kelly criterion workshop

Computing Kelly fractions and growth rates for bets not covered in lecture, then locating the actual peak of the growth-rate curve by hand

3 May 2027

Before the session#

Read the lecture page. Bring the f* = 0.2 and g(0.2) ≈ +2.01% figures — the session recomputes both for a bet with different win probability and payout odds.

In the session#

The group computes f* and g(f*) for a new (p, b) pair, then evaluates g(f) at several multiples of that f* (0.5x, 1x, 1.5x, 2x) to locate by hand where growth turns from positive to negative — checking that it isn’t exactly at f*×2 in general, since the growth curve isn’t symmetric around its peak.

Second half: starting from a blank page, the group re-derives why expected log growth is the right objective for a repeated, bankroll-compounding bet, rather than expected bankroll itself — connecting back to week 4’s variance content on the same page.

Afterwards#

Nothing is marked this week. Going into week 12: the course’s own list of what’s been formalized so far, and — just as importantly — what’s been explicitly flagged as a gap along the way (multi-way MDF, real Kuhn poker equilibria, estimating p and b from actual play) rather than quietly assumed solved.