The following arose in a blitz game this weekend.
Earlier black played ...a3-a2, and after the game commented that he should have left it on a3 and played for ...Kb2-a2. I'm not so sure. I haven't looked at those positions, but with the pawn only on a3 white might have time for Rxg6 (or Rxg7 as the case may be). Anyway, while black was playing ...a3-a2 white played Kg2-f3-g4?!, black replied ...g7-g6, and white followed up with the dreadful f2-f4?? (my punctuation). But this may have been a winning attempt since white was ahead on the clock. If so then white is an evil genius, as he did win in the end.
1...Kf8 was played, and in the game black sent his king on a long journey to b2, over to g2, then back to the center. Finally with his king on e6 he played ...h6-h5+ Kg4-g5 ...Ra1-g1 Ra7xa2 ...Rg1xg3+, quickly reaching a drawn position with a lone h-pawn for black. Then with his time running out black hung his rook ...
2. Ra6 Kf7 3. Ra7+ Ke6 4. Ra6+ Kd5 5. Ra5+ This is my analysis. It seems better for white to check the black king away from the kingside before black gets in his big idea.
The counterattack 5. Rxg6 is way too slow. 5... Rb1 6. Ra6 a1=Q 7. Rxa1 Rxa1 -+.
5... Kc6 6. Ra6+ Kb5 7. Ra8 h5+! Cleanest, but I only found this idea after analyzing the zugzwang idea.
Having witnessed black's king march, initially I was interested to see if there would be a possible zugzwang with the black king on g2/h2.
7... Kc4 8. Ra6 Kd3 9. Ra4 Ke3 10. Ra3+ Ke2!
(Black has to be careful crossing the f-file. Incidentally, black king to the first rank would allow white to retreat Kg4-h3-h2. In the game black gave white this opportunity, but white did not take it.
10... Kf2? 11. f5! gxf5+ 12. Kxf5 h5 13. Kf4 Ke2 14. Ke4 Kd2 15. Kd4 Kc2 16. Kc4 Kb2 17. Rb3+ Kc1 18. Ra3 Kc2 19. Ra5 I have to admit that Fritz is finding the white defense here.
19... Kb2 20. Rb5+ Kc1 21. Ra5 Kd2 22. Kb3 Rh1 23. Rxa2+ =. Fritz thinks white is drawing, and the tablebase confirms it.)
11. Ra4 Kf2 12. Ra3 Kg2
zugzwang
13. Ra4 h5+ 14. Kg5 Kxg3 15. Kxg6
(or
15. Ra3+ first)
15... h4 16. f5 h3 17. f6 h2 18. Ra3+ Kg2 19. f7 h1=Q 20. f8=Q
So far this is all my analysis.
20... Qb1+ Fritz gives this winning move.
(I analyzed the immediate
20... Rg1 21. Qf4! Qh3! 22. Rxh3 Kxh3+. But according to Fritz after
23. Kf5! a1=Q it's only a draw. Ugh.)
21. Kg5 Qb5+ 22. Kg4 Qc4+ 23. Kg5 Rg1! -+ Fritz. Now the ...Rg1 idea works.
8. Kg5 h4! The big idea.
9. Kxh4 9. gxh4 Rg1+ 10. Kf6 a1=Q+ 11. Rxa1 Rxa1 12. Kxg6 Rg1+ 13. Kh6
(
13. Kf6 Rh1 14. Kg5 Kc6 -+)
13... Rf1 14. Kg5 Kc6 15. h5 Kd7 16. h6 Ke7 17. Kg6 Kf8 -+
is how I analyzed it. I double-checked it against the tablebase, and my line does win, but it is far from best play for either color.
9... Rh1+ 10. Kg5 a1=Q 11. Rxa1 Rxa1 Well, this is already in the online 6-piece tablebase. It seems that in my analysis I did not make any critical mistakes, but I did make some suboptimal moves:
12. Kxg6 12. g4 Rg1 13. f5 gxf5 14. Kxf5 Kc6 15. g5 Kd7 16. Kf6 Ke8 17. g6 Kf8 -+.
12... Rg1 13. f5 Rxg3+ 14. Kh7 Kc6 15. f6 Kd7 16. f7 Ke7 -+. Returning to the initial position, the direct 1... h5+ 2. Kg5 h4 does not work: 3. Kxg6 Kf8 4. Kf6 Ke8 (a little safer is 4... Kg8) [b]5. gxh4 =.
I thought it was interesting, I hope you agree.