Just reading that Tim Krabbe (in 2000) wrote that K+NN vs K+NR is lost.
(In Dutch, on his site under /admag -> god.htm)
When I put a random position on with NN vs NR, Lichess tablebases claim : draw.
https://lichess.org/analysis/4k3/3nn3/8/8/8/8/3RN3/4K3_w_-_-_0_1
When I put the position on, which Tim Krabbe mentions in his article, and calls it "the most difficult position to win" (he mentions that using the tablebases by Ken Thompson, it claims a checkmate in 262 moves) then Lichess tablebases show 1 winning move, and also mentions :
"Win prevented by 50-moves rule"
https://lichess.org/analysis/1N6/1RK5/5n2/8/8/8/5n2/6k1_w_-_-_0_1
So, what is going on ? Are some positions drawn for some reason, despite the 50 moves rule, or is this a bug ?
It simply means that position would have been a win if you disregard the 50 move rule, as to win that position more than 50 moves will be played without a capture or a pawn move.
#1
The 50 moves rule has been changed a few times, there once was a 75 moves rule, there was a time when there was a 50 moves rule with exceptions. In ICCF they accept wins beyond the 50 moves rule by table base adjudication.
I don't know Dutch but I strongly doubt Krabbe claimed that RR v RN is always a win. Like in RB v R: Some positions are winning, others drawn. I think that particular position with a win in 262 moves has often been used as an example of the "Godly" chess represented in tablebases, because many times there is only one correct move that wins, and for humans it's simply impossible to see why. I remember someone asked Garry Kasparov whether the winning process makes any sense to him and he laughed and said no, the moves just seem absurd and random.
But yes, in a practical game these days the 50 moves rule would be valid anyway. Chessgames.com's endgame database gives white 40 % wins, 60 % draws: https://www.chessgames.com/perl/chessending?s=RR&i=RN&result=*
@RoikkuvaUpseeri
Yes, I started to wonder whether I had misread the things that Tim Krabbe wrote, and you are right !
Thank you. Appreciated.
The article was about the most difficult checkmate in those cases it is won. "The endgame is won in 76 % of the cases" was a sentence I overlooked. And the word God was used as well.
I'll quote it here, for those who'd like to use some translation software :
from his site timkr.home.xs4all dot nl /admag/kddkdd.htm :
Ik schreef hier een tijdje geleden (zie 'Schaken met God') over het eindspel KTPKPP; koning, toren en paard tegen koning en twee paarden. In het moeilijkste geval waarin dat gewonnen is, kan de sterkste partij op z'n snelst in 262 zetten mat geven. De 'eindspeldatabase'-techniek waarmee die zettenreeks wordt gevonden, knoopt achterwaarts vanuit de matstellingen alle stellingen van een eindspel zodanig aaneen dat men steeds kan zien of mat bereikbaar is en zo ja, hoe ver weg het is.
from his site /admag/god.htm :
Onlangs echter heeft Ken Thompson, na ongeveer een maand rekentijd op een supercomputer, een database gemaakt (zes miljard stellingen, 93 Gigabyte) die de mat-maximin voor KTPKPP bevat. Die bleek 262 zetten; de langste langste kortste winstgang ooit gevonden. Dit eindspel is in 76 % van de gevallen gewonnen.
The 93 GB number gives a great perspective on the advances in tablebase formats. Nowadays Syzygy tablebases store the same (and additional information regarding the 50-move rule) in less than 2 GB. Histogram and additional stats can be found on https://syzygy-tables.info/?fen=1N6/1RK5/5n2/8/8/8/5n2/6k1_w_-_-_0_1.
I think hardest win Rook opposite color bishop vs Knight opposite color bishop Rook and Bishop vs Two knights
Two knight vs pawns Rook Knight vs Two knights generally draw
This topic has been archived and can no longer be replied to.