Chess Computers are they IM? or GM? or Super GM?

[smirobth]

[d]8/4k3/5p2/6p1/4P1Bp/5PP1/1r5P/6K1 b - - 0 24
Anand komments 34...Re2!!
and says: 34...Kd6? 35. gxh4 gxh4 36. f4 Rd2 37. Bh3 Kc5 38. e5 Rd4
39. exf6 Rxf4 40. Kg2 Rxf6 41. Bc8 =
He says "when the white king can leave the 1. rank the position is a draw.
Anand won this. He is a GM. Engines seam to have no idea about this -

Fritz 11 switches from 34...Kd6 to 34...Re2 after 1.5 minutes on my machine and Shredder 10 is even faster needing only 27 seconds to find and stick with 34...Re2 (both using all 3,4,5 plus KRPKBP tablebases).

[BBauer]

That looks great!
I assume that the 6-piece tablebase is very important. If we remove 4 pieces from the initial 11 we get 7 and than the engine may find the right continuation.
Thank you
Bernhard

[smirobth]

I agree, not many humans will "understand" this position, but an educated chess player will at least know the Saavedra position.
When I give the following study from Platow to Rybka 1.0Betax64 with a CEGT-rating of 2864 and a CCRL-rating of 2921 the engine gives

Code: Select all

  18	00:04	   4.276.673	19.651	+4,03	Kc3b4
  19	00:08	   7.786.554	446.523	+0,09	Kc3b4
  20	00:10	  10.135.652	183.446	+0,03	Kc3b4

So the Elo may be >2800 but the understanding is 0.
Here the position
[d]8/8/8/2P4P/8/2K5/7B/k4r2 w - - 0 1

kind regards
Bernhard

With 3,4,5 man tablebases most programs find Kb4 with a winning score in 0 seconds. And humans will only "solve" this one quickly if they have seen it before.

There certainly are positions that computers don't understand, but I'm not sure the positions you have presented are the best examples of this.

[Dirt]

Here a better example from the same game
[d]r1b1r1k1/1p1nnpp1/pq1Bp2p/8/4N3/6Q1/2PRB1PP/5R1K w KQkq - 20 23

enjoy!
kind regards
Bernhard

Actually, I think I recognize these middle game positions from somewhere. A Najdorf if I'm right.

Toga does find the right move in about the same time as the other position, but it only sees the position as equal.

Your endgame position was indeed a nice example of where a GM can best a computer, as far as I could tell.

[smirobth]

That looks great!
I assume that the 6-piece tablebase is very important.

Yes I think so. If you remove the KRPKBP tablebase the programs have trouble. There are many drawn 6 man positions where without tablebases programs see a large advantage due to being up a full exchange and this will throw them off track.

[BBauer]

Here is the game. It shows what some clever guys with help of computers can do. And this is much more than an unatended computer.
Note that the humans were *no* GMs.

[Event "6th Freestyle Tournament, Final, 1st round"]
[Site "playchess.com #101513"]
[Date "2007.06.22"]
[Round "1.1"]
[White "Flyingfatman"]
[Black "Mission control, Rybka 2.1d3 mp"]
[Result "1-0"]
[WhiteElo "2682"]
[BlackElo "2468"]
[ECO "B97"]
[EventDate "2007.06.22"]
[PlyCount "52"]
[EventType "match (rapid)"]

1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6 6. Bg5 e6
7. f4 Qb6 8.Qd2 Qxb2 9. Rb1 Qa3 10. e5 dxe5 11. fxe5 Nfd7
12. Ne4 Qxa2 13. Rd1 h6 14.Bh4 Qd5 15. Qe3 Qxe5 16. Be2 Bc5
17. Bg3 Bxd4 18. Rxd4 Qa5 19. Rd2 O-O 20.Bd6 Nc6 21. O-O Re8
22. Qg3 Qb6 23. Kh1 Ne7 24. Bc7 Qc6 25. Rxf7 Kxf7 26.Nd6 Kg8
1-0

The difficult part is to find 24.Bc7
The rest is more or less forced.
kind regards
Bernhard

[BBauer]

This position was not about computers in general, but only about one single engine. What I wanted to say is: The engine may have ~2800 Elo
but doesn't even know about underpromotion.

In general engines are lost when dealing with fortresses.

Here again the position from Topalov-Shirov
[d]8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 47
Do you know of any engine that will find Shirov's continuation?
May be 6-piece table bases can help.
kind regards
Bernhard

[Ovyron]

And this is much more than an unatended computer.

I think that Mission Control was the 32core unattended computer of Suj.

[Uri Blass]

This position was not about computers in general, but only about one single engine. What I wanted to say is: The engine may have ~2800 Elo
but doesn't even know about underpromotion.

In general engines are lost when dealing with fortresses.

Here again the position from Topalov-Shirov
[d]8/8/4kpp1/3p1b2/p6P/2B5/6P1/6K1 b - - 0 47
Do you know of any engine that will find Shirov's continuation?
May be 6-piece table bases can help.
kind regards
Bernhard

I remember that
Rybka with no tablebases(32 bits single processor) could find it after some hours of search.

Uri

[Father]

...Top hradawres copmuters diong waht they kow how to do(tacicts,dtaabsaes and momerozing etc) , are betetr tahn a Top G.M.; but, cmoptuers donig what they dnot kow how to do, tehn, tehy are porbably not more tahn an amaterur.

......... but cmoptuers all, are mnius than Hmuan bineg rcae intllegneec in cmopresion of gneral concepts.

Wiht best rgeards,

Pbalo
Fhater