Register
to add
comments.
Game number: Main Page
White: Flemming Jensen    Black: David J Bush
twixt.DEFAULT.mc.2022.jul.1.1 (LG) | This game (LG) | Download JTwixt file
On 2022-07-11 at 09:34, Peyrol (info) said:
The bot tells me that |12.s8 would have been better.

The line it gives is 13.q7 14.n8 15.r9 16.k13 17.u15 18.s12 19.t10 20.t14 21.u20 22.v14 (This is typical bot weirdness, but black maintains the connection) 23.w14 24.u12 25.v13 26.w15 27.s16 28.x13 29.k14 30.n11 31.e6 32.e4 33.g11 34.h8 35.n10 36.j15 37.m15 38.m14 39.f4 40.d8 and black threatens J11 or H16.

Earlier, the bot likes |7.m11 for white. 8.k19 9.m15 10.l13 11.j14 12.g19 13.n17 14.n6 15.r7

Earlier still, |6.m12 7.u11 8.o17 9.g10 10.g5 11.i9 12.k5 13.e5 14.f7 15.d12 16.e9 17.c9 18.d11 19.u16 20.s18

Earlier still, |5.n14 6.p7 7.s7 8.r6 9.t5 10.s8 11.t14 12.s4 13.n8 14.m6 15.m5 16.r20 17.o10 18.m16 19.k15 20.j17 21.h16 22.g18 (This is the longest 3:1 slope ladder chase I have seen.) 23.g17 24.f16 25.h19 26.n13 27.q17 28.o15 29.t18.

Of course, the bot is not always perfect. I set both the "trials" sliders to 8700. Thinking time was roughly 3 to 6 minutes per move.

For me, these variations illustrate the importance of playing in the central region early in the game. On moves 5, 6, and 7, we chose to play away from the center, and our reward was (apparently) a lost position.

Any questions, ask Alan. :-)

On 2022-07-12 at 11:18, Peyrol (info) said:
I discovered that the "visualize" option for the bot slows the process down by about 58%, so the search takes more than twice as long. With that option and the heat map turned off, but the evaluation left on, I set the bot against itself after move 3.Q11. The sliders were maxed out at 50000. The time taken per move varied from 7 to 14 minutes, averaging about 9:48. The resulting game was very interesting, but in this post, I want to talk about how white was winning but blundered into a draw. Of course, I will try to answer any questions. I was just joking about Alan.

|4.m8 5.f8 6.q8 7.f13 8.g17 9.t14 10.m4 11.i16 12.j18 13.l13 14.k16 15.k15 16.m15 17.g15 18.f5 19.j6 20.e18 21.s8 22.i4 23.l5 24.k5 25.e6 26.r4 27.g5 28.h4 29.n4 30.o3 31.d4 32.e3 33.e2 34.s20 35.t18 36.u19 37.o18 38.q16 39.q17 40.s12 41.r13 42.n13 43.s16 44.l12 45.m10 46.o7 47.n12

At this point, white threatens to connect at K11 with a simple win. Let's look at how white can force a win in the upper left area. This is not what the bots played. 48.k10 49.l8 50.j8 51.i6 52.j7 53.k7 54.l6 55.k6 56.h6 57.i5 58.j3 59.g6 or 52.k7 53.j7 54.h7 55.i5 is the same.

It's understandable that black might want to try to confuse the issue with superfluous threats, but what I don't understand is why white would cooperate. Maybe the bot does not look at a winning path that involves link crossing? Here's what the bots played.

48.p19 49.r19 50.n18 51.v19 52.v21 53.v23 54.q21 55.m17 56.x22 57.n19 58.m21 59.m22 60.o21 (threatening black L23) 61.n20 62.l19

After all of this faffing around, white has to recognize black is threatening to force a draw on the bottom, which white should not have allowed in the first place. White can still play M18 and hold the win, which now involves link crossing moves on the bottom as well as the top. Does white see the threat? Of course not.

63.l8 64.n9 65.o14 66.m18 67.h19 68.i18 69.n21 70.m19 71.o19 72.m20

White is completely blocked. I stopped the game.

On 2022-07-12 at 13:24, Peyrol (info) said:
I missed a relatively simpler win for white.


1.d22 2.i10 3.q11 4.p15 5.g15 6.i6 7.u11 8.q13 9.o12 10.m13 11.l11 12.l10 13.l15 14.o14 15.j10 16.m12 17.h9 18.k7 19.e6 20.g9 21.f8 22.k11 23.e10 24.resign
|4.m8 5.f8 6.q8 7.f13 8.g17 9.t14 10.m4 11.i16 12.j18 13.l13 14.k16 15.k15 16.m15 17.g15 18.f5 19.j6 20.e18 21.s8 22.i4 23.l5 24.k5 25.e6 26.r4 27.g5 28.h4 29.n4 30.o3 31.d4 32.e3 33.e2 34.s20 35.t18 36.u19 37.o18 38.q16 39.q17 40.s12 41.r13 42.n13 43.s16 44.l12 45.m10 46.o7 47.n12 48.k10 49.l8 50.j8 51.h7 52.k7 53.j7 54.h9 55.i5

You must be logged in to add comments.