Hacker News Reader

The loophole is that the state of the pseudo random number generator is observed (indirectly) and used for predictions. What if slot machines would use a real source of randomness? I assume that this would cause difficulties with the licensing process for these machines because the "fairness" of the machines would be harder to prove.

View on HN

I wouldn't think so. Whatever random result you get, you can bound it to make sure the probability works in your favor a fixed proportion of the time, right?

`if (pseudoRandom() < 0.57) houseWins()` gives the same payout as `if (trueRandom() < 0.57) houseWins()` for all intents and purposes; the only difference is the former is more predictable to people trying to game the machine.

I think the licensing requirements for these machines are strict enough that they want reproducibility in the testing, which would be one reason to require pseudo-random numbers instead of real random numbers.

I imagine they might just simulate a million transactions and test the outcome. If it falls outside of expected probabilities, they can dive deeper, and also have a pseudo-random seed to fall back on to figure out if it was just randomness or something is broken or configured to do the wrong thing.

That's exactly the reason why they use PRNGs.

Reply on news.ycombinator.com