Eternity Puzzle - Solution

Solution

As soon as the puzzle was launched, an online community emerged devoted to solving it, centred on a mailing list on which many ideas and techniques were discussed. It was soon realised that it was trivial to fill the board almost completely, to an "end-game position" where an irregularly-shaped void had to be filled with only a few pieces, at which point the pieces left would be the "wrong shapes" to fill the remaining space. The hope of solving the end-game depended vitally on having pieces that were easy to tile together in a variety of shapes. Computer searches were carried out to find which pieces tiled well or badly, and these data used to alter otherwise-standard backtracking search programs to use the bad pieces first, in the hope of being left with only good pieces in the hard final part of the search.

The puzzle was solved on May 15, 2000, before the first deadline, by two Cambridge mathematicians, Alex Selby and Oliver Riordan. Key to their success was the mathematical rigour with which they approached the problem of determining the tileability of individual pieces and of empty regions within the board. These provided measures of the probability that a given piece could help to fill or 'tile' a given region, and the probability that a given region could be tiled by some combination of pieces. In the search for a solution, these probabilities were used to identify which partial tilings, out of a vast number explored by the computer program, were most likely to lead to a solution. A complete solution was obtained within seven months of brute-force search on two domestic PCs.