Which dedicated was best at rook endings?

[IanO]

Greetings,

After seeing Star Diamond lose a drawn KR-KRP ending against Fidelity Elite in the 11th Active championship, I wonder if any of the high-end models had specific endgame knowledge for this common ending. It would have made an even better marketing feature than mating with KNB-K, since this is such a crucial ending for master-level play.

Of those with hash tables, which obviously make ending play stronger, which were considered the best models for endgame training?

Ian

[Larry]

Greetings,

After seeing Star Diamond lose a drawn KR-KRP ending against Fidelity Elite in the 11th Active championship, I wonder if any of the high-end models had specific endgame knowledge for this common ending. It would have made an even better marketing feature than mating with KNB-K, since this is such a crucial ending for master-level play.

Of those with hash tables, which obviously make ending play stronger, which were considered the best models for endgame training?

Ian

I'm on my way out the door for my daily coffee/crossword fix, so I'll be
brief. The Lang programs seem to be the most knowledgeable in this phase
of the game, followed perhaps by the de Koning programs in the risc/montreux machines.
L

[ricard60]

Greetings,

After seeing Star Diamond lose a drawn KR-KRP ending against Fidelity Elite in the 11th Active championship, I wonder if any of the high-end models had specific endgame knowledge for this common ending. It would have made an even better marketing feature than mating with KNB-K, since this is such a crucial ending for master-level play.

Of those with hash tables, which obviously make ending play stronger, which were considered the best models for endgame training?

Ian

Probably you are talking about this position that came up between Star Diamond and fidelity V11

[fen]8/8/7p/4ppk1/2p3P1/1rP2PK1/5P2/2R5 w - - 0 1[/fen]

In this position at master level, chances are for black its rook is more active so probably even the V11 did not make it right and got in to that KR-KRP draw position that Star Diamond did not know how to draw.

[IanO]

Greetings,

After seeing Star Diamond lose a drawn KR-KRP ending against Fidelity Elite in the 11th Active championship, I wonder if any of the high-end models had specific endgame knowledge for this common ending. It would have made an even better marketing feature than mating with KNB-K, since this is such a crucial ending for master-level play.

Of those with hash tables, which obviously make ending play stronger, which were considered the best models for endgame training?

Ian

Probably you are talking about this position that came up between Star Diamond and fidelity V11

[fen]8/8/7p/4ppk1/2p3P1/1rP2PK1/5P2/2R5 w - - 0 1[/fen]

In this position at master level, chances are for black its rook is more active so probably even the V11 did not make it right and got in to that KR-KRP draw position that Star Diamond did not know how to draw.

No, I'm referring to KR-KRP specifically. These endings are the bread and butter of the working chess master, so I was wondering if any dedicated machines had special evaluation that recognized them.

Here is the first position in question from later in the game, where black (Fidelity Elite 68k) needs to make a crucial decision:

[fen]8/8/3r4/8/2p1p3/2P2k2/5P2/1R2K3 b - - 15 53[/fen]

A master would immediately recognize that black's advantage evaporates to a drawn endgame if you try to directly win material with 53... e3 54. fxe3 Kxe3 55. Rb8 Kd3 56. Kd1 Kxc3 57. Kc1

Do any dedicated engines have endgame evaluation smart enough to realize you can't let the defender get his king in front of the pawn like this?

The losing error came later:

[fen]8/8/8/8/6R1/K1p5/3kr3/8 w - - 2 65[/fen]

65. Rg1 is forced, but Star Diamond played 65. Rd4+ losing.

Did any of the old computer chess journals do rook and pawn ending tests on the best computers?

[ricard60]

Do not know from the old dedicated chess machines which can solve a 5 piece endgame or at least with a good knowledge endgame and a strong cpu that can see the only move that makes a draw with 65.Rg1 at least Mysticum can see it but this is because it has the EGTB for 3,4 and 5 pieces. Maybe this is a good test position for strong old dedicated chess machines.

[vonderlasa]

Radio Shack Master [not 2200X] was unable to win a win R+ 4P [h,g,f,a pawns] vs R + 3P [h,g,f] against me. It couldn't figure out how to bring the king to support the a pawn.

[Larry]

Maybe this is a good test position for strong old dedicated chess machines.

For the record:
Montreux plays Rg1 instantly, holding a white score of -191 for
several minutes. -191 = -1.91 pawns, meaning that white is evaluated as not
quite toast, but is probably lost.
Berlin Pro (non London) plays Rg6 for about 30 seconds then switches
to Rg1, with an even bigger minus for white of -2.48.
L

[Steve B]

The losing error came later:

[fen]8/8/8/8/6R1/K1p5/3kr3/8 w - - 2 65[/fen]

65. Rg1 is forced, but Star Diamond played 65. Rd4+ losing.

Did any of the old computer chess journals do rook and pawn ending tests on the best computers?

interestingly no R+P endgame positions appear in any of the well known dedicated chess computer test suites (Colditz,BT 2450-2650 or BS 2830)

I fed the position to the Novag Chess Champion Super System III(with LCD and printer units attached)
the computer is rated about 1400 running on a 6502 at 2 Mhz
after 1 Hr and 27 minutes it wants to play 1.Rd4+(like the Star Diamond)

The SS III can solve up to mates in 5 although it can take days
ill let it run for another few Hrs to see if it stumbles upon the drawing resource 1.Rg1

Patient Regards
Steve

[Steve B]

After several days of pondering The SSIII never switched from 1.Rd4+

Lets see how the Novag Sensor Dynamic handles the position
Rated about 1150 it could choose the correct 1.Rg1 for all the wrong reasons
will let it ponder for several weeks....

Hey Ya Never Know Regards
Steve

[Larry]

Lets see how the Novag Sensor Dynamic handles the position

Novag Sensor Dynamic... is'nt that the one that runs on compressed air?
L

[Steve B]

Lets see how the Novag Sensor Dynamic handles the position

Novag Sensor Dynamic... is'nt that the one that runs on compressed air?
L

that would be the one
not to be confused with the Novag Super Sensor IV or the Dynamic S
strangely there is no mention of the Dynamic S on the wiki
even more strangely ..while I own the SD and the SSIV I seem to have missed the Dynamic S in my collecting endeavors
I must have been out sick the day it was released in the US?

Surprised At Myself Regards
Steve

[Cornersquare]

The losing error came later:

[fen]8/8/8/8/6R1/K1p5/3kr3/8 w - - 2 65[/fen]

65. Rg1 is forced, but Star Diamond played 65. Rd4+ losing.

Did any of the old computer chess journals do rook and pawn ending tests on the best computers?

The Mephisto Atlanta is known to be not as good at rook endings as the Lang machines, but at least he manages to find 65. Rg1 after around a quarter of an hour (using the two-player mode) displaying -1.6 which means no draw in sight. Before that he also suggested 65. Rd4+.