Request: modern Tetris randomization logic

[JobLeonard]

Currently any piece can follow any piece. In modern Tetris implementations we instead get pieces in shuffled sequences, where each sequence contains every tetronimo piece exactly once. See also:

https://tetris.fandom.com/wiki/Random_Generator

Would make the game a lot more "fair" to play :)

[sneakernets]

Here's what I did for my terrible Commodore version: I created a "bag" and put each piece in twice, then I made sure you didn't get the same piece twice. It ended up being a good chunk of the program, though.

It's probably not the best way to do it but it sure kept the program from generating a deluge of S and Z pieces.

[JobLeonard]

Why two though? Wouldn't the simplest way be to have a sequence of seven and shuffle that sequence every seven steps?

Currently, the logic for a new piece is:

// create a new piece, don't remove old one (it has landed and should stick)
void new_piece() {
  y = py = 0;
  p = rand() % 7;
  r = pr = rand() % 4;
  x = px = rand() % (10 - NUM(r, 16));
}

The key line being p = rand() % 7;

If I'm correct, something like this should work: (just wrote this of the top of my head, haven't really written C++ in a while so probably contains a mistake somewhere)

// index into bag sequence and bag of seven tetronimos
int n = 0, bag[7] = {0, 1, 2, 3, 4, 5, 6}

void new_piece() {
  y = py = 0;
  // Fisher-Yates shuffle the bag every time
  // one random sequence is finished
  if (n == 0) { 
    for(int i = 1; i < 7; i++) {
      int j = rand() % i;
      p = bag[j];
      bag[j] = bag[i];
      bag[i] = p;
    }
  }
  p = bag[n];
  n = (n + 1) % 7;
  r = pr = rand() % 4;
  x = px = rand() % (10 - NUM(r, 16));
}