Endgame test

[smirobth]

I think it would make more sense to design tablebases that follow all the rules of the game rather than to modify the rules of the game to accommodate computers.

But what's the reason of the 50 move rule? Because they thought that 50 moves without pawn move or capture should mean no more progress and a draw. But they were wrong.

Of course, setting it longer is useless for humans, so I just say that we should have different rules for computers. If they say it's mate in 360 moves, let them play them.

So when there is a game between a computer program and a human, which game do they play?
And when someone is analysing a human game using a computer program, how does the analyst avoid making analysis mistakes?

I still think the best solution is to redesign tablebases with the 50 move rule in mind. Otherwise tablebase induced blunders will continue. Take the following position from Gurevich-Rowley, 1990:
[d]8/8/1k2N3/1P1p2p1/8/3NK3/8/8 b - - 0 70
70...Kxb5! (Programs using only 3,4,5-man tablebases will avoid this move, for fear of the resulting tablebase win for White. Yet this is the only conceivable chance to draw!)
71.Nd4+! (Programs using 5-man tablebases want to play 71.Nxg5+, which is a tablebase mate in 68. Trouble is 68 moves is too many and the position is now a draw.)
71...Kc4 72.Ne5+ (Programs with 5-man tablebases again blunder with 72.Nf3??, since it leads to another 5-man tablebase "win" that isn't)

[Terry McCracken]

I think it would make more sense to design tablebases that follow all the rules of the game rather than to modify the rules of the game to accommodate computers.

But what's the reason of the 50 move rule? Because they thought that 50 moves without pawn move or capture should mean no more progress and a draw. But they were wrong.

Of course, setting it longer is useless for humans, so I just say that we should have different rules for computers. If they say it's mate in 360 moves, let them play them.

So when there is a game between a computer program and a human, which game do they play?
And when someone is analysing a human game using a computer program, how does the analyst avoid making analysis mistakes?

I still think the best solution is to redesign tablebases with the 50 move rule in mind. Otherwise tablebase induced blunders will continue. Take the following position from Gurevich-Rowley, 1990:
[d]8/8/1k2N3/1P1p2p1/8/3NK3/8/8 b - - 0 70
70...Kxb5! (Programs using only 3,4,5-man tablebases will avoid this move, for fear of the resulting tablebase win for White. Yet this is the only conceivable chance to draw!)
71.Nd4+! (Programs using 5-man tablebases want to play 71.Nxg5+, which is a tablebase mate in 68. Trouble is 68 moves is too many and the position is now a draw.)
71...Kc4 72.Ne5+ (Programs with 5-man tablebases again blunder with 72.Nf3??, since it leads to another 5-man tablebase "win" that isn't)

As I said earlier, it's time to amend these rules. A mate is a mate, it's won.

So the 50 move rule shouldn't apply in such cases.

[smirobth]

As I said earlier, it's time to amend these rules. A mate is a mate, it's won.

So the 50 move rule shouldn't apply in such cases.

Yes, a mate is a mate. But a rule is also a rule, and there is a very good practical reason for the 50 move rule. Do you think people should be forced to play out QN vs RBN endings for 517 moves? At a 40/2 time control and 8 hours per day, one single game could take a week.

[Terry McCracken]

As I said earlier, it's time to amend these rules. A mate is a mate, it's won.

So the 50 move rule shouldn't apply in such cases.

Yes, a mate is a mate. But a rule is also a rule, and there is a very good practical reason for the 50 move rule. Do you think people should be forced to play out QN vs RBN endings for 517 moves? At a 40/2 time control and 8 hours per day, one single game could take a week.

I think if you can demonstrate mate in reasonable time then there should be an exception to the 50 move rule or admenments. There are cases of course that will make playing on impractical.

Computers are going to rewrite the rule someday themselves