Hi-lo guide
Blaze Hi-lo: what the edge does to a 5
Blaze is a name for this address, nothing more. The page takes one face-up card, the 5, and shows how its two payouts move as the house edge changes.
The 5 in this model
Ranks run from ace (1) to king (13), and the next card is any of the 13 with equal chance. A 5 has four ranks below it (A to 4) and eight above it (6 to K). So P(lower) = 4/13 = 30.77% and P(higher) = 8/13 = 61.54%, and a tie on another 5 loses either call under the strict rule.
The 5 has a neat property. Higher has exactly twice as many winning ranks as lower, so at any edge the lower call pays exactly twice what the higher call pays. Fair prices are 13/8 = 1.625x and 13/4 = 3.250x.
Five edges side by side
| House edge | Higher pays | Lower pays | EV per unit (either side) | Expected loss per 100 calls |
|---|---|---|---|---|
| 0.0% | 1.6250x | 3.2500x | 0.00% | 0.00 |
| 1.0% | 1.6087x | 3.2175x | -1.00% | 1.00 |
| 2.0% | 1.5925x | 3.1850x | -2.00% | 2.00 |
| 3.0% | 1.5762x | 3.1525x | -3.00% | 3.00 |
| 5.0% | 1.5437x | 3.0875x | -5.00% | 5.00 |
Computed in code from P = 8/13 (higher) and 4/13 (lower). At 0% the bet is exactly fair.
Reading the table
Each row cuts both payouts by the same factor. At a 3% edge the higher call pays 1.5762x, which is 0.0488 below fair. The lower call pays 3.1525x, a gap of 0.0975. The gap in units is bigger on the lower side, but it is the same 3% in proportion, and the expected loss per call is identical.
The last column is the useful one for budgeting. At a 3% edge, a hundred one-unit calls from a 5 cost 3 units on average. At 1% they cost one unit. Results will scatter widely around those figures, but the figures are the centre the scatter moves around.
Spotting the edge in a posted payout
Suppose a pay table shows 3.15x for lower from a 5. Multiply by 4/13: the result is 0.9700. Subtract that from 1 and you get the edge, about 3%. The same test works on any rank once you know whether ties lose or win. The tie rules guide explains that check.
Spread at a 3% edge
A higher call from a 5 has a standard deviation of 0.767 units, and a lower call 1.455. Over 100 lower calls the standard deviation is ten times the single-call figure, about 14.6 units, so a 3-unit expected loss is easily hidden in any one session.
Is a 5 a good card to call on?
No card is good or bad in this model. The edge sets the average loss, and it is the same for every rank.
Why is the 0% row there?
It is the fair benchmark. Every real pay table sits below it, and the distance from it is the edge.
For the general idea, see the house edge guide.