Longest Forced Mate Announcement In Computer vs. Computer

[hgm]

It is only a draw if you claim it.

[Bo Persson]

about 12 years ago Crafty announced a mate in 140-something against GM Roman Dzhindi in a 4-program vs 4-gm round robin played on chess.net. Crafty found a way to trade into a KNN vs KP ending. However, Roman didn't go that way and got mated much quicker... There are probably examples of deeper mates when I was using 6 piece EGTBs on ICC but I honestly do not remember any specific examples, it became rather passe' after a while...

Funny how a mate-in-140 announcement against a GM who was at one time near the top 10 is no big deal



I can still remember the "will a program ever play consistent master-level chess?" debate.

i wonder what are the rules regarding huge mate announcements?
generally ..if no pawn has been moved or piece captured a game would end in a draw after 50 moves
so a 150+ mate announcement really has no meaning at all

are the Fide rules different for rated games in which engines are involved(either human v Engine or Engine v Engine)?

i seem to remember that Fide has modified the 50 move rule to include positions known to take more then 50 moves to deliver mate although i cant recall any of them now

Steve

They have reverted that decision again, as it turned out to be too hard to remember. And hardly ever occuring in real games.

[Steve B]

about 12 years ago Crafty announced a mate in 140-something against GM Roman Dzhindi in a 4-program vs 4-gm round robin played on chess.net. Crafty found a way to trade into a KNN vs KP ending. However, Roman didn't go that way and got mated much quicker... There are probably examples of deeper mates when I was using 6 piece EGTBs on ICC but I honestly do not remember any specific examples, it became rather passe' after a while...

Funny how a mate-in-140 announcement against a GM who was at one time near the top 10 is no big deal



I can still remember the "will a program ever play consistent master-level chess?" debate.

i wonder what are the rules regarding huge mate announcements?
generally ..if no pawn has been moved or piece captured a game would end in a draw after 50 moves
so a 150+ mate announcement really has no meaning at all

are the Fide rules different for rated games in which engines are involved(either human v Engine or Engine v Engine)?

i seem to remember that Fide has modified the 50 move rule to include positions known to take more then 50 moves to deliver mate although i cant recall any of them now

Steve

They have reverted that decision again, as it turned out to be too hard to remember. And hardly ever occuring in real games.

I see
Thanks Bo
Regards
Steve

[bob]

about 12 years ago Crafty announced a mate in 140-something against GM Roman Dzhindi in a 4-program vs 4-gm round robin played on chess.net. Crafty found a way to trade into a KNN vs KP ending. However, Roman didn't go that way and got mated much quicker... There are probably examples of deeper mates when I was using 6 piece EGTBs on ICC but I honestly do not remember any specific examples, it became rather passe' after a while...

Funny how a mate-in-140 announcement against a GM who was at one time near the top 10 is no big deal



I can still remember the "will a program ever play consistent master-level chess?" debate.

i wonder what are the rules regarding huge mate announcements?
generally ..if no pawn has been moved or piece captured a game would end in a draw after 50 moves
so a 150+ mate announcement really has no meaning at all

are the Fide rules different for rated games in which engines are involved(either human v Engine or Engine v Engine)?

i seem to remember that Fide has modified the 50 move rule to include positions known to take more then 50 moves to deliver mate although i cant recall any of them now

Steve

KNNKP is a tricky one. There are mates > 50 moves with no pawn move (can't be a capture or the mate disappears). But the problem for a human is to find the very narrow path that gets to A 50 move draw before getting mated or being forced to push his pawn. When I started using the 6 piece egtbs I saw even deeper mate announcements, but things went sort of "mate in 190, mate in 140, mate in 62, mate in 21, mate in 11, etc. Human was not able to play optimally. I don't even remember any computers playing those very well at first since the 6-piece files were first available on my ftp box and nobody had 'em...

[Steve B]

about 12 years ago Crafty announced a mate in 140-something against GM Roman Dzhindi in a 4-program vs 4-gm round robin played on chess.net. Crafty found a way to trade into a KNN vs KP ending. However, Roman didn't go that way and got mated much quicker... There are probably examples of deeper mates when I was using 6 piece EGTBs on ICC but I honestly do not remember any specific examples, it became rather passe' after a while...

Funny how a mate-in-140 announcement against a GM who was at one time near the top 10 is no big deal



I can still remember the "will a program ever play consistent master-level chess?" debate.

i wonder what are the rules regarding huge mate announcements?
generally ..if no pawn has been moved or piece captured a game would end in a draw after 50 moves
so a 150+ mate announcement really has no meaning at all

are the Fide rules different for rated games in which engines are involved(either human v Engine or Engine v Engine)?

i seem to remember that Fide has modified the 50 move rule to include positions known to take more then 50 moves to deliver mate although i cant recall any of them now

Steve

KNNKP is a tricky one. There are mates > 50 moves with no pawn move (can't be a capture or the mate disappears). But the problem for a human is to find the very narrow path that gets to A 50 move draw before getting mated or being forced to push his pawn. When I started using the 6 piece egtbs I saw even deeper mate announcements, but things went sort of "mate in 190, mate in 140, mate in 62, mate in 21, mate in 11, etc. Human was not able to play optimally. I don't even remember any computers playing those very well at first since the 6-piece files were first available on my ftp box and nobody had 'em...

When an engine announces a mate in 190(for example) ..is it taking into account the 50 move rule and say ..pushing a pawn somewhere in the 190 moves to reset the count and avoid the draw rule ...or ignoring the rule totally?

Steve

[Brunetti]

When an engine announces a mate in 190(for example) ..is it taking into account the 50 move rule and say ..pushing a pawn somewhere in the 190 moves to reset the count and avoid the draw rule ...or ignoring the rule totally?

If the announcement is based on tablebases, 50-move rule is not taken into account. If it was based on search, yes, if it's implemented correctly. But that's not the normal case, since a normal search can't go so deeply. And when the hash is involved, the answer depends on how it's implemented. Usually, the 50-move count is not stored into the hash.

Alex

[bob]

about 12 years ago Crafty announced a mate in 140-something against GM Roman Dzhindi in a 4-program vs 4-gm round robin played on chess.net. Crafty found a way to trade into a KNN vs KP ending. However, Roman didn't go that way and got mated much quicker... There are probably examples of deeper mates when I was using 6 piece EGTBs on ICC but I honestly do not remember any specific examples, it became rather passe' after a while...

Funny how a mate-in-140 announcement against a GM who was at one time near the top 10 is no big deal



I can still remember the "will a program ever play consistent master-level chess?" debate.

i wonder what are the rules regarding huge mate announcements?
generally ..if no pawn has been moved or piece captured a game would end in a draw after 50 moves
so a 150+ mate announcement really has no meaning at all

are the Fide rules different for rated games in which engines are involved(either human v Engine or Engine v Engine)?

i seem to remember that Fide has modified the 50 move rule to include positions known to take more then 50 moves to deliver mate although i cant recall any of them now

Steve

KNNKP is a tricky one. There are mates > 50 moves with no pawn move (can't be a capture or the mate disappears). But the problem for a human is to find the very narrow path that gets to A 50 move draw before getting mated or being forced to push his pawn. When I started using the 6 piece egtbs I saw even deeper mate announcements, but things went sort of "mate in 190, mate in 140, mate in 62, mate in 21, mate in 11, etc. Human was not able to play optimally. I don't even remember any computers playing those very well at first since the 6-piece files were first available on my ftp box and nobody had 'em...

When an engine announces a mate in 190(for example) ..is it taking into account the 50 move rule and say ..pushing a pawn somewhere in the 190 moves to reset the count and avoid the draw rule ...or ignoring the rule totally?

Steve

Depends. If you use thompson type DTC databases, you don't get mate distances at all, but if they say "win" then you will win without violating the 50 move rule, you just don't know how long it will take. The DTM databases (Nalimov among others) give you the distance to mate metric, but nalimov's do not factor in the 50 move rule, so you might get mate in 200 when it is really a draw by FIDE rules (50 move rule).

It is possible to do a combination so that you pick the shortest DTM that does not violate the 50-move DTZ (distance to zero the 50 move counter) requirement but again, you won't know exactly how deep the mate is.

[ernest]

It is possible to do a combination so that you pick the shortest DTM that does not violate the 50-move DTZ

So there is a way to find the shortest mate path, that does not violate the 50-move rule?
How can this be done???
(a friend of mine tried this for knnkp, but he had to create DTM tables for 100-halfmove rule, 99-halfmove rule, 98-halfmove rule, ...)

[bob]

It is possible to do a combination so that you pick the shortest DTM that does not violate the 50-move DTZ

So there is a way to find the shortest mate path, that does not violate the 50-move rule?
How can this be done???
(a friend of mine tried this for knnkp, but he had to create DTM tables for 100-halfmove rule, 99-halfmove rule, 98-halfmove rule, ...)

I did not say what you thought. I said you can pick the "shortest" path to a mate, while living within the 50-move rule. That doesn't mean you can find the shortest valid path to mate at all. Just that you won't end up in a 50 move draw case so long as there is some way to actually mate without running into the 50 move case. I'm not quite sure how to find the shortest mate that doesn't violate the 50 move rule, and have not given any thought to the problem at all...

[LiquidNitrogenOverclocker]

It is possible to do a combination so that you pick the shortest DTM that does not violate the 50-move DTZ (distance to zero the 50 move counter) requirement but again, you won't know exactly how deep the mate is.

I believe the Namilov Database Builder gives you the option to solve for DTM as well as DTZ, where you specify the number of moves (50 of course) for the count to zero when an irreversible move is made (pawn move, capture).

So, technically speaking, there might exist a "Mate in 51" that would be declared a draw because of the 50 Move Rule, if the defender played optimally each of the 50 moves along the way.

This same hypothetical mate could conceivably become a Mate in 60, or possibly even longer, if the database solver were tasked to interject a pawn move within the DTZ horizon, in order to preserve the overall win.

I don't believe a collection of "DTZ Positions of Interest" exist, possibly because a fair amount of data mining would be required to produce them, plus you'd have to solve the 6-piece tablebase set twice, once with DTW and once with DTZ.