Skip to content
← The Scroll
Canonical · Framework

The Casino's Four Gears: Edge, Volume, Sizing, and Bankroll

John Kelly Jr.; Edward Thorp · 1956

"A tiny, repeatable statistical edge becomes a near-certain profit only when combined with three other things — enormous volume, bets sized to the edge rather than to conviction, and a bankroll large enough to survive a bad stretch. Miss any one of the four and having an edge stops mattering."

The idea

American roulette pays 35-to-1 on a bet that wins 1-in-38 times, which works out to a 5.26% house edge — invisible on any single spin, but the entire casino industry runs on it.

Why it works

The edge alone does nothing; it needs the other three gears to turn into a certainty. Jacob Bernoulli's 1713 Law of Large Numbers is why the edge becomes reliable at all: repeated enough times, the average outcome converges on the true expected value, which is why casinos want you playing longer, not winning more on any one spin. John Kelly's 1956 Bell Labs paper solved the second problem — how much to bet — proving that betting a fraction proportional to your edge (roughly twice your edge for an even-money bet) grows a bankroll fastest without risking ruin; betting more than that, even with a real edge, leads to eventual blowup. The fourth gear is bankroll: in the classic gambler's-ruin problem, two players flipping a fair coin until one goes broke — the player with less money is overwhelmingly likely to bust first, even with zero edge on either side, simply because a small bankroll runs out of room to survive variance before a large one does.

The takeaway — recall it first
Check your understanding

According to the Kelly Criterion, what happens if someone with a real, small statistical edge bets a large fraction of their bankroll on it, rather than a fraction proportional to the edge?

Further reading

Read more about the topic

The explanation above is written with AI assistance. These are the originals — go to them to check it.

  • A New Interpretation of Information RateJohn L. Kelly Jr., Bell System Technical Journal 35 (1956), via Internet Archive
  • Edward O. ThorpWikipedia
  • Renaissance TechnologiesWikipedia
Up NextSuggested: Continues the theme of Decision-Making

Expected Value

"Choose the option whose average outcome, weighted by probability, is highest."

Blaise Pascal · ModelContinue→
Listen
0 / 5