Tic-Tac-Terminator is not yet winning, but its ability to play a game that results in a tie is clearly improving. Of the last 25 games played, 12 ended in a draw; and 11 of those draws occurred in the 18 most recent games.
It is becoming more apparent in some of the common openings that moves that previously led to a loss are dying out as T3 is penalized a bead for each loss and awarded a bead for each draw. Box pattern #30, for example, began with 10 beads: 2 of each color for its five possible moves. That same box currently has 9 beads, 6 of those are orange and that is the one move that blocks a certain loss on the next move.
However, it is also clear that T3 is not yet trained for obvious wins. Several times in recent games it played a successful strategy that gave it a fork with two X's on two different lines. When its human opponent blocked one line, T3 unfortunately failed to choose the alternative winning move. Had it selected the winning square, it would have been rewarded 3 beads of that color, increasing the likelihood that it would play for victory the next time. By ending in a draw instead, and getting awarded a single bead, it reduced slightly the probability it will make the winning move the next time.
It should be interesting to watch how long it takes Tic-Tac-Terminator to discover winning moves and to begin playing them with regularity.
