I suggested something like 7.t11 in a previous game, but it doesn't seem to wprk now. Maybe |7.m14 would have created more threats, for example 8.n79.t11 I don't really know.
On 2026-06-01 at 23:53, Alan Hensel said:
Twixtbot likes |17.e12 for move 17, which is interesting because it's just one hole away from what White actually did, e13, and because Twixtbot plays it out as a win for White:
... scraping by with no space to spare on the left.
As for move 7, Twixtbot responds to both |7.m14 and |7.t11 with 8.n7, so that seems to be the threat that needs to be addressed. The bot prefers |7.g10.
If |17.e1218.f2019.h1220.j20 then 21.g18 threatening either f21 or i19... interesting... and then 22.h1923.d19 which you might not think of because usually 24.d17 works, but not this time because 25.e19. After 23.d19 Twixtbot flails deciding which corner to attack, and in this case settled on 24.f10, which is of course easily answered by 25.d10.
On 2026-06-03 at 04:12, Peyrol said:
That is wild. Thank you, What would you say was black's losing move? Would |14.i14 have been better?
I don't trust long lines, even with MCTS set to 1600, but it does suggest that perhaps Black made a suboptimal move 12, giving White an edge until White gave it back on move 17.
I suggested something like 7.t11 in a previous game, but it doesn't seem to wprk now. Maybe |7.m14 would have created more threats, for example 8.n7 9.t11 I don't really know.
18.e5 19.g6 20.e10 21.f9 22.g4 23.i5 24.j5 25.j3 26.g14 27.d14 28.f16 29.c16 30.d18 31.d17 32.e16 33.e15
... scraping by with no space to spare on the left.
As for move 7, Twixtbot responds to both |7.m14 and |7.t11 with 8.n7, so that seems to be the threat that needs to be addressed. The bot prefers |7.g10.
Your 19.f17 variation also works. 20.d19 21.d18 22.c17 23.k19 24.i18 25.j17 26.l21 27.m18 28.j20 29.n20 30.o22 31.o23 32.p20 33.m22 34.g12 35.g13 36.e9 37.f10.
When I set it up to choose its own move 14, it seemed absolutely certain that |14.h9 is the best move. It didn't really consider much else. But the result is another long line, with White prevailing again. 15.e12 16.e9 17.f10 18.f7 19.h11 20.i15 21.d18 22.g14 23.d14 24.k8 25.g6 26.c8 27.r6 28.l12 29.p15 30.q10 31.p9 32.q6 33.t7 34.s5 35.p7 36.o5 37.u5 38.u4 39.v3 40.x3 41.u1 42.o19 43.m12 44.n11 45.k13 46.j13 47.j15
It also thinks h9 was the best move for the previous turn. Another long line, but Black wins it. |12.h9 13.q8 14.n15 15.g10 16.e5 17.s18 18.u18 19.e11 20.j5 21.m12 22.k14 23.n16 24.r12 25.d5 26.c4 27.g6 28.h4 29.d3 30.d2 31.b4 32.a3 33.p11 34.m17 35.k13 36.i13 37.e15 38.j11 39.k8 40.o5 41.f4 42.f3 43.l6 44.m4 45.r6 46.l10 47.p15 48.o18 49.q19 50.p16
I don't trust long lines, even with MCTS set to 1600, but it does suggest that perhaps Black made a suboptimal move 12, giving White an edge until White gave it back on move 17.