The most famous study cooked by computer analysis?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
mwyoung
Posts: 2062
Joined: Wed May 12, 2010 8:00 pm

Re: The most famous study cooked by computer analysis?

Post by mwyoung » Tue Feb 11, 2020 7:02 pm

jp wrote:
Tue Feb 11, 2020 5:25 pm
mwyoung wrote:
Tue Feb 11, 2020 12:57 am
5B2/1n6/1k6/3K4/B7/8/8/8 w - - 0 1

This position was thought to be drawn for over a century. It is a mate in 57 moves!
Yes, it was, but do you know of any analysis of BB vs N that claimed to show it?

Ovyron wrote:
Tue Feb 11, 2020 5:03 pm
I sometimes wonder about TB positions, people keep referencing them as the ultimate solution of chess, because for any position with n pieces you can know what moves win, draw or lose. But, can you conceive that there's an undiscovered bug that has one of those positions tagged wrongly? That when you follow the lines the draw evaporated or the win evaporates??
Clearly, the more independent TBs the better, so e.g. we can be quite comfortable with 5-man positions, where we have many different TBs (though not so much if we have access to only one of the many).

Jouni wrote:
Tue Feb 11, 2020 5:10 pm
Speaking of endgames here's Alekhine - Capablanca position

And analysis from 2005! "..."
But I noticed soon in testgames, that surprisingly 1.Rd5! is the best move.
Whose 2005 analysis is that?

What were the results after 1.Rd5 compared with 1.Ra4?
e.g. if they are both winning, 1.Ra4 would be the human way. (M&L also give it an exclamation mark.)
There is analysis showing it is a win. And that is why the 75 move rule was put into place. Since it mates in over 50 moves without a pawn move.
Professing themselves to be wise, they became fools,
take on me. Foes 0.

jp
Posts: 1255
Joined: Mon Apr 23, 2018 5:54 am

Re: The most famous study cooked by computer analysis?

Post by jp » Tue Feb 11, 2020 7:09 pm

mwyoung wrote:
Tue Feb 11, 2020 7:02 pm
There is analysis showing it is a win. And that is why the 75 move rule was put into place. Since it mates in over 50 moves without a pawn move.
What are you referring to? The BB vs N? Well, clearly we have the TB results showing BB vs N is a win. Do we have written, earlier human analysis claiming to show it is a draw (or just belief with zero analysis)? That's the question.


According to TBs, the standard draw rule is not a problem here. It claims DTZ is 89, so White captures the Black N comfortably before 50 moves.

todd
Posts: 56
Joined: Thu Apr 19, 2018 7:09 pm

Re: The most famous study cooked by computer analysis?

Post by todd » Tue Feb 11, 2020 9:21 pm

mwyoung wrote:
Tue Feb 11, 2020 7:02 pm
There is analysis showing it is a win. And that is why the 75 move rule was put into place. Since it mates in over 50 moves without a pawn move.
The new 75-move rule is a rule for arbiters to step in and declare a game drawn if the players don't do so themselves. Players can still claim a draw after 50.

There is also a 5-time repetition rule that allows arbiters to step in (but players can still claim after 3).

In the past, the 50 move rule was extended for certain endgames, but then the rule was changed back to a universal 50 move rule, and that's still what we have today.

syzygy
Posts: 4477
Joined: Tue Feb 28, 2012 10:56 pm

Re: The most famous study cooked by computer analysis?

Post by syzygy » Wed Feb 12, 2020 12:07 am

mwyoung wrote:
Tue Feb 11, 2020 12:57 am
Jouni wrote:
Mon Feb 10, 2020 8:24 am
Who has good candidates?
Here is my candidate.



This position was thought to be drawn for over a century. It is a mate in 57 moves!

Special rule in chess.

Chess rules 3: The new 75-move rule
2 states: “any series of at least 75 moves have been made by each player without the movement of any pawn and without any capture. If the last move resulted in checkmate, that shall take precedence.”Aug 23, 2018
The 75-move rule isn't going to help much since there is still the 50-move rule that allows the "losing" side to claim a draw after 50 moves.

However, in this particular position white can force a capture before the 50-move counter runs out:
https://syzygy-tables.info/?fen=5B2/1n6 ... _w_-_-_0_1

mwyoung
Posts: 2062
Joined: Wed May 12, 2010 8:00 pm

Re: The most famous study cooked by computer analysis?

Post by mwyoung » Wed Feb 12, 2020 1:37 am

jp wrote:
Tue Feb 11, 2020 7:09 pm
mwyoung wrote:
Tue Feb 11, 2020 7:02 pm
There is analysis showing it is a win. And that is why the 75 move rule was put into place. Since it mates in over 50 moves without a pawn move.
What are you referring to? The BB vs N? Well, clearly we have the TB results showing BB vs N is a win. Do we have written, earlier human analysis claiming to show it is a draw (or just belief with zero analysis)? That's the question.

This position was thought to be drawn (Kling and Horwitz, 1851)
Professing themselves to be wise, they became fools,
take on me. Foes 0.

jp
Posts: 1255
Joined: Mon Apr 23, 2018 5:54 am

Re: The most famous study cooked by computer analysis?

Post by jp » Wed Feb 12, 2020 6:36 am

mwyoung wrote:
Wed Feb 12, 2020 1:37 am
This position was thought to be drawn (Kling and Horwitz, 1851)
Thanks for the reference.

Here's what they say:
Kt. and B. against B. cannot win, if the weaker party can obtain a position similar to the above, but they win in most cases. For example, see diagram No. 2.

Diagram No. 2:


1. Bb4 Ng(8) 2. Bd5+ Kg7 3. Kg5 Kh7 4. Bf8 Kh8 5. Be4

jp
Posts: 1255
Joined: Mon Apr 23, 2018 5:54 am

Re: The most famous study cooked by computer analysis?

Post by jp » Wed Feb 12, 2020 8:28 am

zullil wrote:
Tue Feb 11, 2020 11:47 am
jdart wrote:
Tue Feb 11, 2020 11:07 am
The thing that impresses me is how much of pre-computer and pre-TB endgame analysis by humans is flawed, or wrong.
A lot of chess analysis by humans is flawed or wrong. It's just more obvious with endgame positions
Engine analysis too. e.g. myoung's candidate position above.

This is DTZ 89, DTM 112.

SF10, depth 91 (eval. +3.65) fails to make any progress and clearly can't beat the 50-move rule.


Once the depth exceeds the DTZ, I think it's fair to say SF10 has failed completely.
Maybe someone with a fast machine can see if SF11 or dev gets anywhere at depths >112.

Post Reply