Since Google's Deep Mind conquered Go earlier this year, and perfect information games are becoming unbeatable, I was wondering if it were possible to give chess more of a lifespan by introducing randomness and expanding the game tree from the beginning.
People who are familiar with chess AIs and AI development in general, do you think the following would work, or would it be just as easy to solve as regular chess with enough computing power?
The game would play as follows:
Board has four 3x3 grids: red, blue, green and yellow. Before the game starts, one square from each color is randomly chosen. That is 9^4=6561 combinations. The game has also four dice, correspondingly red, blue, green and yellow.
Each randomly chosen square will act as a "nuke square" during the game. After every move by either black or white, the four dice are rolled, and if a 6 is rolled, then any piece that currently was on that square gets taken off the board. If the square is empty, nothing happens. (E.g. red and yellow dice roll 2 and 5, but blue and green dice both roll 6, so any pieces in blue and green nuke squares are removed from the board. The only exception being that black pawns cannot be nuked after white's first move. After black has taken it's first turn, pawns can be nuked.)
Two examples of boards with different nuke pieces (b2, d5, e7, f4 in the first & c6, d4, f2, f5 in the second.)