Thanks, definitely software-controlled display. No LCD driver chip indeed. I didn't realize at first that it's actually 2*2 ribbon cables on top of each other.

Strange that they didn't solder the LCD panel.

The Supra only has 1 LCD compared to Royal, so I guess it only tracks time for white? Or does it switch to black's total time when it's the computer's turn?

LCD panel is 34 pins. 2 outputs from ribbon cable go to 2 LCD corner pins, possibly VCC/GND.
Then 16 outputs each go to 2 LCD pins, very odd.

The LCD has 4 7-segment digits, or does it have more segments? Like a blinking colon (I didn't see any on photos online), or DP (display point, at bottom-right of a digit).

Yes, you are right that it is basically a swapoff algorithm.

Because of the RAM limitation I came up with the idea of evaluating each square on the board sequentially, and when the calculations for a square were completed the net score for that square was added to a cumulative value. From what I remember there were three main components for the square score - one was a positional evaluation (each piece gets a certain score for the square it is on); one is a mobility count - 1 point for each piece that attacks this square; and the most significant was the swapoff value. I don't recall if there was anything sophisticated in the swapoff, for example a rook acting through a queen is different to a queen acting through a rook, and I can't recall if there was an upper limit on the number of men on each side which could participate in the swapoff on a particular square, but I do recall that it worked reasonably well. If, as I suspect, you have one or more of the products using my algorithm, you could conduct some tests in order to work out the level of sophistication of the swapoff function.

There were also some bonus scores but I can't recall the details, though I believe one of them was a "push passed pawn" bonus; one was a penalty for losing castling rights. Until a few years ago I still had the source code, but then a pipe burst in my cellar and the paper was destroyed.

I also devised an algorithm for mating with KRvK and with KQvK - both were rule based I think, and both were in the code.

...

The algorithm starts by looking at square a1. It cycles through every piece on the board and when it encounters one which attacks a1 it adds to the "score for the current square" variable whatever is the appropriate score for that piece.

Was haltet ihr davon ?

Was haltet ihr davon ?