A Drawn Game

This game between Castro and Ring-of-Fire

looks like an honest draw (I don't see any blunders where one of them would have missed a win.)

23. G5 instead of F2?

How does Black reply to 23.G5 24.E2 ?

R-o-F was considering 23. G5 for some time

but it thought it was refuted by 24. C1.

That's probably just a re-ordering of moves (with F2 played later).

The evaluation wasn't terribly low there,

so could still be a drawing line.

The continuation is then

25. B1 (forced by bridge threat)

26. F2

27. H1 (forced by another bridge threat)

but at some point Black can A4 (destroy white's fork threat)

and then it could be white's ring in the center

but that's far down the tree.

Why must LG use 3 different coordinate systems?!?

Has anyone done any work on draw detection? This is provably a draw as of move 53 even if we assume the two players are working together to allow one of them to win. Before that there were some virtual connections that could still needed to be probed that would have allowed a win if they'd both been working towards that goal. If I force the rollouts to reply to virtual connection probes, I get 0 wins given 200,000 simulations as of move 46. Before that there are potential rings left. Without early detection, this tree is far too big for me to be able to solve it as a draw to prove blunders.

So far, we have seen a few artificial draws on

4-board, and a natural one on size 5.

My prediction: earlier or later we will also

see draws on 6, 7, 8, and likely bots will be involved...

Ingo.