Mines guide
Tsunami Mines: a seeded simulation of a high-risk setting
Tsunami is merely the name filed against this URL. It does not describe a variant. The page simulates one mines setting on a computer and compares the output with the exact maths.
The setting simulated
The board has 25 tiles and 10 mines. The rule is fixed: reveal two tiles, then cash out. The exact chance of getting through both is C(15, 2) / C(25, 2) = 105 / 300 = 35.00%. The offered multiplier with a 1% edge is 2.8286x, a profit of 1.83 units on a win.
Seven simulated sessions
Each session is 120 rounds at one unit per round. Mine positions and pick order are drawn by Python's pseudo-random generator from seed 4242. This is a computer simulation for teaching. It is not real play and no service produced these results.
| Session | Rounds won | Net (units) | Lowest point | Longest losing run |
|---|---|---|---|---|
| 1 | 38 | -12.51 | -12.51 | 6 |
| 2 | 42 | -1.20 | -2.03 | 12 |
| 3 | 52 | 27.09 | -1.34 | 5 |
| 4 | 44 | 4.46 | -1.66 | 7 |
| 5 | 39 | -9.69 | -9.69 | 9 |
| 6 | 53 | 29.91 | -6.69 | 9 |
| 7 | 53 | 29.91 | -5.17 | 9 |
Seeded pseudo-random simulation (Python random.Random, seed 4242), 7 sessions of 120 one-unit rounds. Average session result 9.71 units. This is a simulation, not a record of real play; a different seed gives different numbers.
Checking the output
- Expected wins per session: 120 x 0.35 = 42.0, with a standard deviation of about 5.2.
- Expected net per session: -1.20 units, with a standard deviation near 14.8. Session results that differ by tens of units are normal.
- Losing runs: with a 65% chance of losing each round, a run of 8 or more losses within 120 rounds has a probability of 75.77%.
A longer run
Short sessions scatter widely. Running the same setting for many more rounds with the same seed brings the average close to the exact figure. Over 2,000 rounds the simulated mean net per round is 0.0607; over 40,000 rounds it is -0.0067. The exact value is -0.0100. With a high-variance setting, convergence is slow, which is why short-term results say little about value.
What simulation is good for
It makes drawdowns and streaks visible in a way a single expected value cannot. It is not a forecast. With a different seed every row in the table would change, and any one session could finish well ahead or well behind. The expected loss per round stays at 1% throughout.