There were some extremely non-obvious moves in this game that deserve explanation.
|10.m21 is a 1-4 dropback defense that I sometimes use. It feels weak; on the other hand, a more typical 0-4 or 1-2 dropback didn't look strong enough on both sides.
|16.e17 was on account of judging the bottom path to be more promising than the top. For example, if 17.k13, then my plan was the completely non-obvious 18.j21, which fortifies my m21 dropback and also threatens to the left across the bottom, so that if White closed the gap 19.i1620.h1421.j1522.j1723.j14, then I could make my way across: 24.f2125.f2226.h2027.e2028.f19.
Instead, Csaba chose to continue the windmill with |17.u14, but he was out of room and had no windmill response to |18.m5. That's a good sign for Black.
I thought I had the game well in hand by |24.h7, with a path across the top. However, Csaba is a talented endgame player. I knew to stay on guard. It didn't become evident until later what he intended with |25.o6 and |27.j6... those pegs didn't look like they were good for anything. |29.s3 was the big surprise. It doesn't look natural. I knew something was up, and I had to figure out what.
And so I played |30.p8 with confidence, thinking there was nothing Csaba could do.
Wrong again. The unlinked |31.t6 peg seems to work, even though it is right between the linked s6 and u6 options, which don't. There I was, with a large group of pegs and links bearing down on the right side, and Csaba was putting up a net of pegs preventing it from connecting to the right! Some lines that don't work:
However, Csaba was being just as careful with me as I was with him. It seems that he figured out my plan, knew that |33.q20 would not work, and found the much more awkward |33.n19, which I hadn't considered. There seemed to be no way to get to my magic p17 spot: |34.p1735.q1636.o1937.o1738.o2039.l2040.k2041.j19 - not good. Also,
Rethinking the problem, I realized that p17 was not the only spot that worked as a connector to r13. There's 5 spots in a little "w" shape: o16, p17, q16, r17, s16. My task was to connect one of these to the right side, while threatening to connect to L17 (and the left side) if the o15-s15 gap closed.
It was looking pretty grim until I stumbled upon 34.p18.
The nice thing about |34.p18 is that it connects to 3 of those "w" spots in a single link, and it's a tilt away from my m21 dropback peg. Maybe there is a better proof that it works, but all I can say is that it sure seems to:
I think |34.p18 was the move that sealed the game. It's probably the only reply to |33.n19 that works, and there is probably no reply that works against it. That's my story, and I'm sticking to it! ;-)
|37.l13 was also not terribly obvious, but it underscores a key Twixt way of thinking. Its purpose is to kill my path across the bottom: 38.n1739.i1640.h1441.j14 or 38.i1639.n17. He could have killed that path in other, more obvious ways, such as |37.n17, but |37.l13 raises the question of influence on the path across the top. Now I have no n12 threat, for example. It turns out that threat has no value, but there are a lot of things that can happen in a Twixt game. Should Csaba consider them all? No, he doesn't have to: all he needs to know is that |37.l13 is the most influential move on the top half of the board that kills my line across the bottom, and is therefore the best move.
Finally, |43.d9 was a little disappointing. I was expecting |43.d8 for a little extra battle, which might have gone something like this: 44.e1045.c1146.g947.d948.e1349.e1150.f11. On the other hand, it was over anyway, and it was a relief to put this already long game to a quicker end.
|10.m21 is a 1-4 dropback defense that I sometimes use. It feels weak; on the other hand, a more typical 0-4 or 1-2 dropback didn't look strong enough on both sides.
|16.e17 was on account of judging the bottom path to be more promising than the top. For example, if 17.k13, then my plan was the completely non-obvious 18.j21, which fortifies my m21 dropback and also threatens to the left across the bottom, so that if White closed the gap 19.i16 20.h14 21.j15 22.j17 23.j14, then I could make my way across: 24.f21 25.f22 26.h20 27.e20 28.f19.
Instead, Csaba chose to continue the windmill with |17.u14, but he was out of room and had no windmill response to |18.m5. That's a good sign for Black.
I thought I had the game well in hand by |24.h7, with a path across the top. However, Csaba is a talented endgame player. I knew to stay on guard. It didn't become evident until later what he intended with |25.o6 and |27.j6... those pegs didn't look like they were good for anything. |29.s3 was the big surprise. It doesn't look natural. I knew something was up, and I had to figure out what.
|30.r2 31.t5 32.t10 was my first reaction, but then 33.u10 34.v11 35.s10 36.r11 37.q11 38.p12 39.l7 40.i5 41.n8 - that j6 peg was good for something!
Ah, but then surely |30.r6? No, then 31.q4, and he has a double threat to t8 (32.p8 33.r7) or to make good on the j6 threat (32.q8 33.l7 34.i5 35.n8).
I had to kill the j6 threat, but would |30.p8 be strong enough? It looked just barely:
31.s6 32.r2 33.t5 34.t10 35.t12 36.u8
31.v7 32.t10 33.t12 34.s8 35.u10 36.u3 37.q4 38.t6
31.u6 32.t10 33.t12 34.s8 35.u10 36.u3 37.q4 38.t6
31.r9 32.t6
31.v9 32.t6
And so I played |30.p8 with confidence, thinking there was nothing Csaba could do.
Wrong again. The unlinked |31.t6 peg seems to work, even though it is right between the linked s6 and u6 options, which don't. There I was, with a large group of pegs and links bearing down on the right side, and Csaba was putting up a net of pegs preventing it from connecting to the right! Some lines that don't work:
|32.t10 33.t12 34.u8 35.s10 36.t2 37.r5 38.r3 39.q3
|32.t12 33.t13 34.u10 35.s11 36.r7 37.r9 38.u5 39.r5
Thus my predicament. I was looking at |32.t12 33.t13 when I remembered my previous analysis across the bottom: |32.s20 33.q20 34.q19 35.p18 36.p17 37.n17, which is not good for Black, but could those two be combined? Yes! |32.t12 33.t13 34.r13 35.u11 36.s20 37.q20 38.q19 39.p18 40.p17 41.n17 42.q15 works! This made me nervous, though: after |32.t12 33.t13 34.r13 35.u11 36.s20, my plan was too obvious, and it's not obvious that there isn't some way for White to still connect to the bottom. I decided to go for |32.s20 first, to make it seem like just a desperate endgame move or stalling tactic, in hopes of the |33.q20 response.
However, Csaba was being just as careful with me as I was with him. It seems that he figured out my plan, knew that |33.q20 would not work, and found the much more awkward |33.n19, which I hadn't considered. There seemed to be no way to get to my magic p17 spot: |34.p17 35.q16 36.o19 37.o17 38.o20 39.l20 40.k20 41.j19 - not good. Also,
34.r18 35.o17 36.p17 37.q16 38.o20 39.l20 40.k20 41.j19
34.q19 35.o17 36.p17 37.q16 38.o20 39.l20 40.k20 41.j19
Rethinking the problem, I realized that p17 was not the only spot that worked as a connector to r13. There's 5 spots in a little "w" shape: o16, p17, q16, r17, s16. My task was to connect one of these to the right side, while threatening to connect to L17 (and the left side) if the o15-s15 gap closed.
34.n18 35.p18 36.q19 37.q16 38.o20 39.l20 40.k20 41.j19
34.q16 35.q14 36.r18 37.p18 38.o20 39.l20 40.k20 41.j19
34.r17 35.q16 36.p19 37.p18 38.o22 39.l20 40.k20 41.j19
34.o20 35.p20 36.m19 37.r21 38.r22 39.q23 40.r18 41.q16 42.p17 43.o17
It was looking pretty grim until I stumbled upon 34.p18.
The nice thing about |34.p18 is that it connects to 3 of those "w" spots in a single link, and it's a tilt away from my m21 dropback peg. Maybe there is a better proof that it works, but all I can say is that it sure seems to:
35.p20 36.r17 37.t18 38.u18 39.v17 40.s19 41.u19 42.u21
35.o17 36.o16 37.n15 38.t12 39.t13 40.r13
35.q16 36.o20
35.r17 36.q20
35.n17 36.o16 37.l18 38.t12 39.t13 40.r13
35.q18 36.r17 37.o19 38.t18 39.p17 40.t12 41.t13 42.r13
35.l18 36.o16 37.n17 38.t12 39.t13 40.r13
35.p17 36.q16 37.q15 38.o20
35.s19 36.t18 37.t17 38.r18 39.n17 40.o16
35.r18 36.o20
35.r19 36.r18
I think |34.p18 was the move that sealed the game. It's probably the only reply to |33.n19 that works, and there is probably no reply that works against it. That's my story, and I'm sticking to it! ;-)
|37.l13 was also not terribly obvious, but it underscores a key Twixt way of thinking. Its purpose is to kill my path across the bottom: 38.n17 39.i16 40.h14 41.j14 or 38.i16 39.n17. He could have killed that path in other, more obvious ways, such as |37.n17, but |37.l13 raises the question of influence on the path across the top. Now I have no n12 threat, for example. It turns out that threat has no value, but there are a lot of things that can happen in a Twixt game. Should Csaba consider them all? No, he doesn't have to: all he needs to know is that |37.l13 is the most influential move on the top half of the board that kills my line across the bottom, and is therefore the best move.
Finally, |43.d9 was a little disappointing. I was expecting |43.d8 for a little extra battle, which might have gone something like this: 44.e10 45.c11 46.g9 47.d9 48.e13 49.e11 50.f11. On the other hand, it was over anyway, and it was a relief to put this already long game to a quicker end.