Hi-lo guide
Tiger Hi-lo: how likely a bankroll is to run out
The Tiger in this page's name is decoration; the rules discussed are those of ordinary hi-lo. This page answers a practical question exactly: starting with a given number of units and staking one per call, how likely is the bankroll to hit zero within a set number of calls?
Why this example uses whole-number payouts
The exact method used here tracks every possible balance as a whole number of units. That only works if a win adds a whole number of units, so the total return on a win must be an integer such as 2x or 4x. Payouts built as (1 - edge) / chance usually are not whole numbers, so this page uses a hypothetical simplified pay display with rounded figures.
Scenario A: face card 7, call higher, which wins 46.15% of the time and pays a flat 2x. Scenario B: face card 10, call higher, which wins 23.08% of the time and pays a flat 4x. Both return 12/13 of the stake on average, a house edge of 7.69%. That edge is much larger than the 1% used elsewhere on this site, a direct cost of rounding payouts down to whole numbers.
Scenario A: 2x on a 6-in-13 call
| Starting bankroll (units) | P(bust within 300 one-unit bets) |
|---|---|
| 10 | 90.65% |
| 30 | 45.67% |
| 60 | 2.44% |
Scenario B: 4x on a 3-in-13 call
| Starting bankroll (units) | P(bust within 300 one-unit bets) |
|---|---|
| 10 | 89.54% |
| 30 | 58.38% |
| 60 | 15.81% |
Reading the two tables
The expected loss over 300 calls is the same in both scenarios: 300 x 1/13, or about 23.1 units. What differs is variance. Scenario B wins rarely and in bigger lumps, so its standard deviation per call is 1.685 units against 0.997 for scenario A. Bigger swings bring small bankrolls to zero sooner, which is why the 30-unit and 60-unit rows are clearly worse in B. With only 10 units, both scenarios bust in most 300-call stretches, because the expected loss alone is more than twice the bankroll.
For a large bankroll the picture changes. Ruin within 300 calls then requires losing far more than the expected 23.1 units, and that takes an unusually bad run in either scenario. Over a longer horizon, though, a negative edge drives the chance of eventual ruin toward 100% for any finite bankroll that keeps playing.
How the exact calculation works
Start with all probability on the starting balance. For each call, move a share p of every balance up by (payout - 1) units and a share 1 - p down by one unit. Any probability that reaches zero is stored and removed. After 300 steps, the stored total is the exact chance of ruin, with no simulation involved.