[detofuuu] Avoiding recent repetitions

A computer chess program serves as an opponent for a human practicing how to defend a drawn position.  The computer should avoid repeating positions, avoid positions in which the human can force repetition, and (this is new) when it has no choice but to allow repetition, prefer repeating positions that the human last saw a long time ago (does he or she still remember how to play it correctly?) over those seen recently.  There are likely many devilish details.

The computer might permit repeating a recently seen position if it can then diverge from the recently seen play and force play toward unseen states, or perhaps states seen a very long time ago.

Probably, the computer computes minimax over drawn positions using an evaluation function based on when, if ever, the leaf node was last seen.  Tricky is that the value of a position depends on the path taken to get there, not just the position itself.