3 ms·

I remember looking at some early Draw Poker machines in Las Vegas back in the late 80's/early 90's, and thinking about pseudo-random number generation (as it ex

I remember looking at some early Draw Poker machines in Las Vegas back in the late 80's/early 90's, and thinking about pseudo-random number generation (as it existed at that time). If it was using a standard linear-congruential RNG with a 16 bit value, there would be only 65536 possible seeds, and hence only that many distinct sequences of random numbers, and hence only that many possible shuffles. Even a 64-bit RNG would have only 1.844674407370955e+19 possible shuffles, far lower than 52!.

zaik 2y ago
ln(52!)/ln(2)≈226 bits. You don't have to depend on a single random 64-bit number: https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle
- Straw 2y ago
  If the RNG has < 226 bits of state, such as most non cryptographic PRNGs, you cannot reach all shuffles even if you use multiple numbers.
  - IanCal 2y ago
    I'm not totally sure that's true. I get where you are coming from, an 8 bit rng can only output 256 distinct values at most. 256 would be the longest cycle. However if you combine the outputs of two, but don't step them at the same time, you can have more outputs. Imagine you had one that outputs just 0 and 1 and then loops. You could have two of these updating at a different frequency and have four distinct outputs. I think that makes sense.
    - wduquette 2y ago
      Oh, yeah, there are ways to finesse it. I’m just wondering whether anyone involved in those early poker machines had made the effort.
      - cardiffspaceman 2y ago
        In Nevada one was required to “prove the randomness” of random outcomes in video poker machines as well as other electronic gambling machines. So effort was made.
  - kcnj 2y ago
    Hey, sorry to contact you via this method. I meant to reply to your comment under the thread on SCALE but it's locked. I'm expecting an offer from that company and would like to ask you about your experience there. If you don't mind, could you accept my connection request on LinkedIn and we could chat there? Thank you :)
DerekL 2y ago
When playing draw poker, you only need to select 10 cards from the deck, because the other cards will never be played. So that's 52⋅51⋅…⋅43 = 57,407,703,889,536,000, which needs 56 bits.

zaik 2y ago

ln(52!)/ln(2)≈226 bits. You don't have to depend on a single random 64-bit number: https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle

Straw 2y ago
If the RNG has < 226 bits of state, such as most non cryptographic PRNGs, you cannot reach all shuffles even if you use multiple numbers.
- IanCal 2y ago
  I'm not totally sure that's true. I get where you are coming from, an 8 bit rng can only output 256 distinct values at most. 256 would be the longest cycle. However if you combine the outputs of two, but don't step them at the same time, you can have more outputs. Imagine you had one that outputs just 0 and 1 and then loops. You could have two of these updating at a different frequency and have four distinct outputs. I think that makes sense.
  - wduquette 2y ago
    Oh, yeah, there are ways to finesse it. I’m just wondering whether anyone involved in those early poker machines had made the effort.
    - cardiffspaceman 2y ago
      In Nevada one was required to “prove the randomness” of random outcomes in video poker machines as well as other electronic gambling machines. So effort was made.
- kcnj 2y ago
  Hey, sorry to contact you via this method. I meant to reply to your comment under the thread on SCALE but it's locked. I'm expecting an offer from that company and would like to ask you about your experience there. If you don't mind, could you accept my connection request on LinkedIn and we could chat there? Thank you :)

DerekL 2y ago

When playing draw poker, you only need to select 10 cards from the deck, because the other cards will never be played. So that's 52⋅51⋅…⋅43 = 57,407,703,889,536,000, which needs 56 bits.