IIRC, the board changed randomly between five or so sequences, so in a single spin, if you waited a while, you'd see, say, sequence 2, and then when that ran it's course it would jump to sequence 4 and execute that, then do sequence 1, then back to 2, etc. Larsen figured out how to tell when a "right" sequence was running early enough to watch for a certain light and hit the button, by which time it jumped to one of the safe (as in, all additional spins or Big Bucks) squares.
The fix was to expand that five or so sequences to a whole bunch more. I think Chris is right in that the board, while electronically-driven, was not PC-driven.