[wdrgufod] Many deviation bases

Extend the idea of searching for a flawless chess game by permitting many Deviation Bases.  Any position in any previous game for which the only known continuation went on to lose (or white draw) may serve as the start position for the current game.  This system makes sense according to game theory.  Essentially the same idea as as previously described for a different scenario.

Run a match of a computer chess program playing itself according to this system.  The purpose is automatic exploration of opening theory.  Encourage earlier deviations in order to widely explore theory.  Most of the time, the computer plays the best move it finds.  Only when it needs to deviate into unknown theory does it play a suboptimal move (perhaps calculated with MultiPV, selected with probability decreasing according to the loss from the best move).