Spunk Odds Reference

Hi-lo guide

Wizard Hi-lo: how it compares with dice and a coin flip

Wizard is just this page's name; nothing magical sits behind it, and the game is regular hi-lo. The question here is comparative: if hi-lo, dice and a coin flip all carry the same 1% edge, what, if anything, is different about them?

Three bets with the same edge

Line up three bets that each win close to half the time. In hi-lo, call higher on a 7, which wins with chance 6/13. In dice, bet that a roll from 0.00 to 99.99 lands under 46.15. And flip a coin. Each pays (1 - edge) / chance with a 1% edge.

Three bets with a 1% edge, one-unit stake
BetWin chancePayoutEV per unitSD per bet
Hi-lo: 7, call higher46.1538%2.1450x-1.00%1.0693
Dice: under 46.1546.1500%2.1452x-1.00%1.0694
Coin flip50.0000%1.9800x-1.00%0.9900

All three follow the site's generic models: 13 equally likely ranks, 10,000 equally likely dice rolls, a fair coin.

The expected value column is identical. The standard deviation column is almost identical, because the win chances are close. On average, and in their spread, these three bets are the same bet wearing different clothes.

Where hi-lo really differs: the grid of chances

Dice lets the player set the win chance in steps of 0.01%, so almost any chance from 0.01% up to just under 99% is on offer. A coin offers exactly one. Hi-lo sits in between: a single call can only win with probability k/13 for k from 1 to 12, twelve distinct values. The smallest single-call chance is 7.69% and the largest is 92.31%.

Chains fill in the gaps. Two calls in a row reach products such as (6/13) x (6/13) = 21.30%, and three calls on a King-lower, Ace-higher, 7-higher path come to 39.33%. But a chain pays the edge at every step: three steps at 1% each keep 97.03% of the fair value, against 99.00% for a single dice roll at the same overall chance.

Where they are the same: verification

All three can be built on the same hash construction. A float u from HMAC-SHA256 becomes a hi-lo rank by floor(u x 13) + 1, a dice roll by floor(u x 10,000) / 100, and a coin side by u < 0.5. The checking procedure is the same; only the last line differs.

What the comparison shows