I actually enjoyed this. Both took slightly different approaches to the opening than I would have taken. Started out very even so the idea that it was heading towards a tie is not implausible.
Are either of you interested in a game vs. Pie? Same general idea though I may make some adjustments to the numbers.
can we rematch soon? I know how to beat Oromagi's strategy, even if he alters his earlier moves. I know how he approaches the game now and know a fatal flaw in it. I also know what I did wrong with my 9, not just 8 and 7. I also know a better variant to start with 10 even if I'm going aggressive.
If this counts as a loss one my record here, I want a rematch vs him.
I know what I did wrong now and know that I was not wrong to avoid aiming for a forced tie. What I did wrong was to deny the natural luck that comes fromt he extra 5 in the 55 that each of us have. I didn't properly force it naturally, i artificially assumed he'd attack me and made +6's for no reason.
I will vs pie right now if he wants, I know how to win every time almost, assuming I properly understand how my opponent thinks if they haven't screwed up by turn 6-7
I’m not as interested in playing myself as I am in writing an AI to play the game…. Though saying that, I’m not 100% sure I will end up having the motivation
I designed an AI that was about 95% random number generator (basically just ignored zones impossible to win or lose then randomly played the rest of the zones). Gave it a few extra troops to make up for the fact that it was so dumb but was still disappointingly easy to beat.
This min-max algorithm with some smart Alpha/beta pruning, with some smart optimization maybe able to churn through all the possibilities as the search tree is wide but shallow.
Is your bot specifically programmed for the starting board we have been using? If it is capable of playing other boards then there are some changes I would like to make to the starting board for the you vs. RM game.
Interestingly, I started by trying to generate a vector containing all the possible R1 moves; of which there are 92378, it’s a non-trivial algorithm to generate - as there are millions of potential duplicate moves. Even though I didn’t end up using it for a min-max algorithm: I was very happy with myself that it could generate the full set of combinations in less than 20 seconds.