With Game Boy Tetris, it's possible to know if the (bitwise or) 3 in a row re-rolling is impossible. If 2 pieces already have clashing bit patterns then definitely the first piece roll will be accepted by the algorithm -- the next piece will be random. Likewise, if you know that 2 pieces have a matching bit pattern, then you know that certain pieces will be aggressively (up to 3 rolls) avoided. In TGM1 and TAP, even with just 1 piece preview, you develop a sixth sense for the "probable 2nd preview". When bitwise matches occur in GB Tetris, the same potential manifests itself. Perhaps players are already subconsciously catching on to the patterns. But even if not, they are orderly enough to be learned without too much trouble. Here is a diagram that illustrates the bitwise relationships: https://www.dropbox.com/s/67ipbq2aqypy70v/GB-Tetris-Circle.png?dl=0 To use this diagram: -look at the white shape that contains your active piece -follow the extension of this shape towards the middle -if your next piece is contained in this extended shape, you have a match -if you have a match, then pieces outside the extended shape are more probable while pieces inside the shape are less probable Ex 1: O piece with J coming next -O's white shape extends into a circle, encompassing O,I,L,J -this is a match with the J coming next -you will very likely receive one of T,S,Z afterwards Ex 2: I piece with L coming next -I's white shape extends into a almond, encompassing I,L -this is a match with the L -you will almost definitely not get one of I or L afterwards Ex 3: L piece with Z coming next -L's shape is the middle already, and includes only itself -Z clashes with the L -the next piece will be unpredictable Now call them what you will, circle/2-bit matches, almond/1-bit matches, triangle/0-bit matches, and clashes. It should be clear that all possible active+next piece combinations fall into one of those 4 categories. And the particular probabilities of the likely/unlikely pieces of these specific scenarios are straightforward to calculate. I'll leave that as an exercise to the reader. Now the way I calculated the theoretical piece distributions earlier in the thread gives some other cool info for free. You might be wondering, Is this all worth it? How often do you get clashes and how often do you get matches? Matches: 36211/107442 ~= 33.7% Clashes: 71231/107442 ~= 66.3% So events become predictable roughly a third of the time. Probably not game changing and perhaps even borderline useless. But cool to know nevertheless. Now, we know that the randomness feeding the algorithm is biased. So the specifics of what I've just outlined don't apply. But even with the bias, the algorithm is pushing things in this direction, and so the patterns should still hold to an extent.