Keno guide
Titan Keno: measuring the swing in a top-heavy table
Titan is a label for this page and nothing in the game depends on it. The focus is variance: how wildly a keno ticket's result swings when most of its return sits in rare prizes.
The table under study
Here is an invented schedule for an 8-number ticket in the 80/20 format. It pays nothing below 4 hits, 2x at 4 hits, and then climbs to 25,000x for all 8. Its return is 93.20%.
| Hits | Probability | As odds | Illustrative pay |
|---|---|---|---|
| 8 | 0.0004% | 1 in 230,114.6 | 25,000.00x |
| 7 | 0.0160% | 1 in 6,232.3 | 1,500.00x |
| 6 | 0.2367% | 1 in 422.5 | 100.00x |
| 5 | 1.8303% | 1 in 54.6 | 10.00x |
| 4 | 8.1504% | 1 in 12.3 | 2.00x |
| 3 | 21.4786% | 1 in 4.7 | 0 |
| 2 | 32.8146% | 1 in 3.0 | 0 |
| 1 | 26.6464% | 1 in 3.8 | 0 |
| 0 | 8.8266% | 1 in 11.3 | 0 |
Return to player of this illustrative pay table: 93.20% (EV per unit -6.80%). The table is a teaching example, not any operator's schedule.
Computing the standard deviation
Variance is the average squared distance from the mean. For a pay table it is easiest in two steps. First compute the mean of the squared pays: the sum of chance x pay^2. For this table that is 3,102.9. Then subtract the square of the return. The square root of the result is the standard deviation of one ticket: 55.70 units. The expected loss per ticket is only 0.068 units, so a single ticket's swing is many hundreds of times larger than its average cost.
Where the variance comes from
Squaring the pays makes the top row dominate. The 8-hit row alone contributes 2,716.0 to the mean of squared pays, about 88% of it. That row comes up 1 in 230,115 tickets. Most people who buy this ticket will never see it, and the 7-hit row, at 1 in 6,232, is not much more common.
The return without the rare rows
Drop the 7- and 8-hit rows and the return falls to 58.27%. That is a fair picture of what a typical long session delivers. Over 1,000 tickets, the chance of seeing at least one 7- or 8-hit result is 15.19%. So most 1,000-ticket sessions lose more than the headline edge suggests, while a small minority win large amounts. The average over everyone matches the headline; the typical person does worse.
A calmer table for contrast
A flatter hypothetical schedule for the same ticket might pay 0.5x for 2 hits, 1.5x for 3, 3x for 4, and top out at 250x. Its return is 95.03%, close to the first table, but its standard deviation per ticket is only 2.08 units. Over 1,000 tickets the spread of the total is about 66 units for the flat table, against 1,761 for the top-heavy one.
Why the normal approximation fails here
For flat tables, the total over many tickets is roughly bell-shaped, and "mean plus or minus two standard deviations" is a fair guide. For a table like this one it is not. The total is a long, low line of losses with a few rare spikes. A side-by-side of two schedules with similar returns is on Ninja Keno.