Keno guide
Wolf Keno: check the numbers yourself
Wolf is a label on this page and has nothing to do with the rules. This is an exercise. Work through it with a calculator or a few lines of code, and you will be able to check any keno table you meet.
What you need
- A calculator with a combinations key (nCr), or any programming language with a binomial coefficient function.
- The format: 80 numbers, 20 drawn.
- A 4-spot ticket and an illustrative pay column: 2 catches pays 1x, 3 pays 8x, 4 pays 120x.
Task 1: count all possible draws
Compute C(80, 20). You should get 3,535,316,142,212,174,320. That is the number of different sets of 20 the game can draw. Every one is equally likely in a fair draw.
Task 2: count draws for each catch count
For h catches, choose which h of your 4 spots are drawn, C(4, h), and fill the rest of the draw from the 76 numbers you did not pick, C(76, 20 - h). Multiply.
| Catches h | C(4, h) x C(76, 20 - h) | Probability |
|---|---|---|
| 0 | 1,090,013,712,241,956,540 | 30.8321% |
| 1 | 1,529,843,806,655,377,600 | 43.2732% |
| 2 | 751,733,594,649,625,200 | 21.2635% |
| 3 | 152,894,968,403,313,600 | 4.3248% |
| 4 | 10,830,060,261,901,380 | 0.3063% |
Task 3: check that the counts add up
Add the five counts. The sum must equal C(80, 20) exactly; this is Vandermonde's identity. If it does not, a term is wrong. It is a quick way to catch typing errors before you go further. The probabilities, likewise, must sum to 100%.
Task 4: compute the return
Multiply each probability by its pay and add. With the illustrative pays you should get 92.62%. The edge is then 7.38%. As a cross-check, the expected catch count is 4 x 20 / 80 = 1, and the sum of h x probability over your table should give the same.
Task 5: verify one hashed draw
The table below derives a full 20-number draw from a seed pair and nonce, using one common construction. Pick any 4 numbers from 1 to 80 before you look at the result, then count your catches against the sorted list.
| Server seed (revealed later) | wolf-exercise-seed-3e8d |
|---|---|
| Commitment shown before play: SHA-256(server seed) | 709d9b8935676cc37ce3d548f7e40a816c291e470f6ba946898151dffee31cda |
| Client seed : nonce : round | wolf-student:3:0 |
| HMAC-SHA256(server seed, message) | e952e5b78d423f051eb7cbf625e141d3650035c3482fd3b8bf8fa1135bfeb3bd |
| First 4 bytes (8 hex digits) | e952e5b7 = 3,914,524,087 |
| Float = value / 2^32 | 0.9114211628 |
| 20 floats, each picks from the numbers not yet used | 73, 44, 10, 13, 32, 24, 61, 30, 78, 35, 22, 38, 17, 39, 57, 20, 19, 56, 54, 16 |
| Drawn numbers, sorted | 10, 13, 16, 17, 19, 20, 22, 24, 30, 32, 35, 38, 39, 44, 54, 56, 57, 61, 73, 78 |
One common construction, shown for teaching. Real services document their own message format; check theirs before verifying.
To repeat it yourself: compute SHA-256 of the server seed and compare with the commitment row; compute HMAC-SHA256 with the server seed as key and the message shown; take 4 bytes at a time and divide by 2^32; then use each float to pick from the numbers not yet drawn, as explained on Royal Keno.
Task 6: repeat it many times
If you can program, draw 20 of 80 a large number of times and count catches on a fixed 4-spot ticket. With 100,000 draws, the share of 4-catch results should be near 0.306%, give or take about 0.017%. Your own results will differ a little from run to run. That is expected.