Gonnosuke is right about the fidemove-extension and rollback, which i mentioned also - from my prev post "(it was for a few years extended by FIDE to 100 moves for troitski endings)" [they restored everthing to 50-move when tablebases + John Nunn et al discovered that K+B+B v K+N+N in the generic case (ie out of positions likely to arise in a chess game, not a simple tablebase %age) is very often a win  but in 100+ moves.

KBNK OTOH just needs 3 days re-practice --- even I can do it [ive played too few games to see where i stand here - i guess it would stabilise to 1775 or 1800 and then fluctuate around it. I hope i meet Gonnosuke in a K+B+N v K ending - but that'll never happen! :( how am i ever to reach said ending against 2000+ player? :(    :) ] 

normajeanyates  said: "...another way to see weekness of programs - set up a troitski ending (K+N+N v K+P, P behind  the troitski line) where tablebase says it is a 30-move win. remove the K+N+N v K+P tablebase, and no program will be able to win this." 

One of the many reasons that engines have a very hard time with KNNKP endgames is that in some instances they require more than 50 moves to win.  That kind of depth is too much for programs to handle; they end up in a futile search, chasing rainbows and white whales that always seem to remain just over the horizon and out of reach.

I remember FIDE temporarily changing the rules and exempting certain endgames from the 50 move rule and instead allowing 75 or 100 moves (I can't remember which it was) for KBBKN, KNNKP, KQKBB, KQKNN, KRBKR and KQPKQ endgames.  Last I heard, they had changed the rules yet again and reinstated the 50 move rule for all endgames which is a potential problem, even for tablebases, because they don't factor in the 50 move rule nor do they account for positions where castling is possible.

On a more practical level, I suspect that a lot of chess players, and possibly even some titled players, would experience difficulties in finding the mate under time pressure in a "basic" KBNk endgame simply because it occurs so rarely -- 1 in every 5000 games - or something ridiculous like that.  This subject has been on my mind recently because I have a game here at chess.com that I thought for a time was looking like a KBNk endgame.  Fortunately, it turned into a KBNPK endgame somewhere around move 90!  Talk about a sigh of relief...

In correspondence chess, it's not a matter of "if" you'll be able to pull off the mate, it's a matter of "when".  The prospect of the game taking another 2 months to complete is what I was most worried about! 

-Roy aka 'Gonnosuke'

About the game Self-Proclaimed Best Computer Player? White play 38. f4! as I expected.  I am not in the team, for I find it's interesting to watch too, and I need not to join now at the end. (I wrote my negative opinion about vote chess at my week report No4, and  normajeanyates mention it here in the comment.) I need two minutes to find this move (more of course for a definite conclusion): first I watch 38. Bb6?, and concluded  white lose for they will need to sacrifice rook for f-pawn after black gxf3. And then, a picture apeared in my mind: after f3-f4, black g-pawn is blocked, and if he play g4-g3, white take Rxg3 and it's a check and white will take Bh3 too! So now, after 38.  f4! we can nave a situation with five black pawns agaist rook - but, as I can see, they all will be blocked and taken.

I  can't belive programs didnt see it. Maybe it's still some uncovered posibility...

another way to see weekness of programs - set up a troitski ending (K+N+N v K+P, P behind  the troitski line) where tablebase says it is a 30-move win. remove the K+N+N v K+P tablebase, and no program will be able to win this. But most GMs can win this:) many non-masters can win troitski endings specially if the 50-move draw rule is extended to 100-moves  (it was for a few years extended by FIDE to 100 moves for troitski endings) - maybe lydia also can if she has practised this!

lydia wrote in the article: I remember a similar move in the first game of the match Spaski-Fischer 1972, when Fischer as black in a simple drow endgame toke a h2 pawn with his bishop, and lost.

For general readers, I'll inform that Jonathan speelman in his book "analysing the endgame" chapter 8 is entirely an analysis of the position after 29..Bxh2. Speelman gives 18 pages of small-print analysis to make a convincing case that fischer could have drawn after losing his bishop; Fischer made 3 more mistakes otherwise he could possibly have drawn. Of course Speelman admits in the beginning that it was very difficult over the board at 2 hrs 30 mins for 40 movesto find the "correct" moves instead of the 3 mistakes - over the board at 2 hrs 30 mins for 40 moves, but he says least one of fischer's 3 moves was clearly weak (at GM level).

Also Olafson's book [speelman doesnt mention which book], and Byrne/Nei " Both sides of the chessboard, 1974".

Programs are useless in the Spaski-Fischer 1972 29..Bxh2 position - endgames with so many pawns - programs can only "calculate" - there is no way to tell them  difficult endgame concepts.

[i posted this  in "how to analyse a game" also] 

lydia,

I intend to analyse that position in your game, using programs as tools not masters :) -  in fact i noted it down as soon as you posted it.

The delay is because at present my computer is busy with vote-chess "positionals" vs. "tacticals" [i am in the tacticals: i convinced people to go for gruenfeld - white transposed into KID Pertosian system.] and a few games - i am playing correspondence chess after about 30 years!

In the team forum of that game I  have often quoted your statement about vote-chess in your last weekly blog. You can see it after the game is over. So much political-type campaigning is needed! I am mainly doing the political campaigning :) But i think i will not play vote chess after this.