SWISSTEST: 1.RobboLito - 2.Rybka - 3.Stockfish

[Glarean]

After this I admit I have no desire to pursue any discussion.

Yeah, of course...

Being rude to Kurt for no reason, and calling CCC, a site for trolls was uncalled for.

You didn't read correctly: He attacked without any justification.

You are correct. I didn't read it. He attacked? Where? Was the post deleted?

No, you are right, I'm sorry. ----> http://www.talkchess.com/forum/viewtopi ... 79&t=30944

.

[Albert Silver]

I'm not a fan of these types of positions, as if your program is strong enough, it may well see that everything loses, and then choosing the best move is really irrelevant, if you can "see".

I understand this point, but at the same time I would prefer to have the program play to extend the game as long as possible when losing, or to win as quickly as possible when winning. What to do when the position is equal is more interesting.

Here's why: I have personally witnessed hundreds of games where the final result was altered by one side's flag falling. This doesn't just occur at fast time controls either-- it's a real possibility anytime there's a zero increment game.

I should add that a significant percentage of those are not really due to time management issues, but to unexpectedly long games where the result is obvious but altered anyway.

I don't disagree. Obviously, actually, since I originated the "swindle mode" idea years ago which does exactly this (if a position is a forced draw, keep the draw as complicated as possible to give the opponent a chance to make a mistake.) But for dead lost positions, there is always a chance that if all moves are lost, with enough analysis they are all lost in about the same number of moves. Proving that is hard, but a very smart program might see that what was considered an inferior move is actually about as good as the rest...

I don't like the lose in 50 vs lose in 100 move type positions, as they are too deep for me to measure whether they are correct or not.

Yes. I think his position would make a great addition to one of those endgame playing suites to see how the program handles it, and how often it can convert or defend. I just think that as an Avoid Move position it is inappropriate, since in my view an Avoid Move position isn't about trying to swindle the opponent later, but avoiding an attractive move that leads to an undesired position as opposed to others.

As to the depth, well, I'm not a bad player, but nor am I a GM. In my view, if I can handily beat an engine of Rybka 3's calibre in this position, in the conditions provided, it is a compelling argument about its assessment. On the other hand, the truth-seeker in me forces me to admit that Rybka showed some unexpected weaknesses in its analysis that suggests that there is plenty of room for progress yet.

Here is an example:

[D]8/3bkppp/p7/3p4/5P2/2KB1P2/PP5P/8 b - - 0 1

Analysis by Rybka 3 x64:

3...Bb5 4.Bxb5 axb5 5.Kb4 d4 6.b3 Kd6 7.a4 bxa4 8.bxa4 Kd5 9.f5 h5 10.h4 f6 11.a5 d3 12.Kc3 Kc6 13.Kxd3 Kb5
+/= (0.45) Depth: 14 00:00:02 275kN
3...Bb5 4.Bxb5 axb5 5.Kb4 d4 6.b3 Kd6 7.a4 bxa4 8.bxa4 Kd5 9.f5 h5 10.h4 f6 11.a5 d3 12.Kc3 Kc6 13.Kxd3 Kb5
+/= (0.45) Depth: 15 00:00:03 367kN
3...Bb5 4.Bxb5 axb5 5.Kb4 d4 6.b3 Kd6 7.a4 bxa4 8.bxa4 Kd5 9.f5 h5 10.h4 f6 11.a5 d3 12.Kc3 Kc6 13.Kxd3 Kb5
+/= (0.45) Depth: 16 00:00:04 512kN
3...Bb5 4.Bxb5 axb5 5.Kb4 d4 6.b3 Kd6 7.a4 bxa4 8.bxa4 Kd5 9.f5 h5 10.h4 f6 11.a5 d3 12.Kc3 Kc6 13.Kxd3 Kb5
+/= (0.45) Depth: 17 00:00:06 721kN
3...Bb5 4.Bxb5 axb5 5.Kb4 d4 6.b3 Kd6 7.a4 bxa4 8.bxa4 Kd5 9.f5 h5 10.h4 f6 11.a5 d3 12.Kc3 Kc6 13.Kxd3 Kb5
+/= (0.45) Depth: 18 00:00:11 1134kN

18 plies and it is suggesting a truly dreadful move, though it apparently sees this a few seconds later. That said, its bastard brother, RobberLito, is worse:

Analysis by RobboLito 0.085e4 x64:

3...Bc8 4.Kd4 Kd6 5.b4 g6 6.b5 axb5 7.Bxb5 Bb7 8.f5 g5 9.a4 Bc6 10.Bxc6 Kxc6 11.f6 h5 12.a5 Kb5 13.Kxd5 Kxa5
+/= (0.29) Depth: 15/34 00:00:00 898kN
3...Bb5
= (0.21) Depth: 16/40 00:00:01 3093kN
3...Bb5 4.Bxb5 axb5 5.Kd4 Kd6 6.f5 h6 7.h4 Kc6 8.a3 Kd6 9.b3 Kc6 10.f4 Kd6 11.h5 Kc6
+/= (0.28) Depth: 16/40 00:00:02 3599kN
3...Bb5 4.Bxb5 axb5 5.Kd4 Kd6 6.f5 Kc6 7.a3 Kd6 8.b3 Kc6 9.a4 bxa4 10.bxa4 h6 11.h3 Kb6 12.Kxd5 Ka5 13.Kd6 Kxa4 14.Ke7 f6
+/= (0.33) Depth: 17/40 00:00:02 4069kN
3...Bb5 4.Bxb5 axb5 5.Kd4 Kd6 6.f5 Kc6 7.a3 Kd6 8.b3 h5 9.h3 Kc6 10.h4 Kd6 11.a4 bxa4 12.bxa4 Kc6 13.f4 f6 14.a5 Kb5 15.Kxd5 Kxa5 16.Kc5 Ka4
+/= (0.29) Depth: 18/40 00:00:03 5246kN
3...Bb5 4.Bxb5 axb5 5.Kd4 Kd6 6.f5 Kc6 7.a3 Kd6 8.b3 h5 9.h3 Kc6 10.h4 Kd6 11.a4 bxa4 12.bxa4 Kc6 13.f4 f6 14.a5 Kb5 15.Kxd5 Kxa5 16.Kc5 Ka4
+/= (0.29) Depth: 19/40 00:00:04 7121kN
3...Bb5 4.Bxh7 g6 5.f5 g5 6.f6+ Kxf6 7.Kd4 Ke6 8.b4 Kd6 9.Bc2 Be8 10.Bb3 Bc6 11.Bd1 Bd7 12.f4 gxf4 13.h4 Ke6 14.Bg4+ f5 15.Bf3 Kf6 16.Bxd5 Bc8 17.Bf3 Be6 18.a3
+/= (0.34) Depth: 20/48 00:00:09 15790kN
3...Bb5 4.Bxh7 g6 5.f5 g5 6.f6+ Kxf6 7.Kd4 Ke6 8.b4 Kd6 9.Bc2 Be8 10.Bb3 Bc6 11.Bd1 Bd7 12.f4 gxf4 13.h4 Ke6 14.Bg4+ f5 15.Bf3 Kf6 16.Bxd5 Bc8 17.Bf3 Be6 18.a3
+/= (0.34) Depth: 21/48 00:00:16 27908kN
3...Bb5 4.Bxh7 g6 5.f5 g5 6.f6+ Kxf6 7.Kd4 Ke6 8.b4 Kd6 9.Bc2 f6 10.Bd1 Bc4 11.a4 Kd7 12.Kc5 Kc7 13.Bc2 Kd7 14.Bf5+ Kc7 15.Bg4 Bb3 16.b5 axb5 17.axb5
+/= (0.41) Depth: 22/48 00:00:33 57450kN
3...a5 4.b4 axb4+ 5.Kxb4 Kd6 6.a4 h5 7.a5 h4 8.f5 Bc6 9.f4 f6 10.Bc2 Bd7 11.a6 Bc6 12.Bb1 d4 13.Bd3 Kc7 14.Kc5 Ba8 15.Bb5 Kb8 16.Kxd4 Ka7 17.Bf1 Bc6 18.Bc4 h3 19.Kc5
+/= (0.70) Depth: 23/59 00:05:04 527mN

Edit: Well, I'm amazed. I fed it to Fritz 12 out of curiosity, and after a minute, it seemed as if it could not err, and then.... lightning struck.

Analysis by Fritz 12:

3...Bc8 4.Kd4 Ke6 5.b4 g6 6.a4 Kd6 7.b5 axb5 8.Bxb5 Bb7 9.a5 Bc6 10.Bxc6 Kxc6 11.a6 Kb6 12.Kxd5 Kxa6
+/= (0.26) Depth: 16/24 00:00:00 1399kN
3...Bc8 4.Kd4 Kd6 5.b4 g6 6.a4 f6 7.b5 axb5 8.Bxb5 Be6 9.a5 Kc7 10.h4 Kb7 11.a6+ Kb6 12.Bd3 Ka7
+/= (0.36) Depth: 17/26 00:00:00 2203kN
3...Bc8 4.Kd4 Kd6 5.b4 g6 6.a4 f6 7.b5 axb5 8.Bxb5 Be6 9.a5 Kc7 10.h4 Kb7 11.Bd3 Ka7 12.f5 gxf5
+/= (0.36) Depth: 18/29 00:00:01 3087kN
3...Bc8 4.Kd4 Kd6 5.b4 g6 6.a4 f6 7.b5 axb5 8.Bxb5 Be6 9.a5 Kc7 10.Bd3 Kb7 11.a6+ Kb6 12.h4 Ka7 13.Kc5 h6
+/= (0.34) Depth: 19/27 00:00:02 5329kN
3...Bc8 4.Kd4 Kd6 5.b4 h6 6.b5 axb5 7.Bxb5 Bb7 8.a4 Bc6 9.Bxc6 Kxc6 10.f5 Kd6 11.a5 Kc6 12.a6 f6 13.a7 Kb7 14.Kxd5 Kxa7
+/= (0.36) Depth: 20/29 00:00:06 12304kN
3...Bc8 4.Kd4 Kd6 5.b4 h6 6.b5 axb5 7.Bxb5 Bb7 8.a4 Bc6 9.f5 Bb7 10.a5 Bc6 11.Bxc6 Kxc6 12.f6 gxf6 13.a6 Kb6 14.Kxd5 Kxa6
+/= (0.39) Depth: 21/36 00:00:12 23648kN
3...Bc8 4.Kd4 Kd6 5.b4 h6 6.b5 axb5 7.Bxb5 Bb7 8.a4 Bc6 9.f5 f6 10.h3 Kc7 11.Bxc6 Kxc6 12.Kd3 Kb6 13.Kc3 Ka5 14.Kb3 Kb6
+/= (0.37) Depth: 22/33 00:00:28 51704kN
3...Bc8 4.Kd4 Kd6 5.b4 h6 6.b5 axb5 7.Bxb5 Bh3 8.Be8 Bg2 9.Bxf7 Bxf3 10.a4 Kc6 11.h4 Be4 12.a5 Kb5 13.Bxd5 Bxd5 14.Kxd5 Kxa5 15.f5 Kb4 16.f6
+/= (0.40) Depth: 23/33 00:00:36 68526kN
3...Bb5 4.Bxb5 axb5 5.Kd4 Kd6 6.f5 f6 7.a3 Kc6 8.b3 Kd6 9.Kd3 Kc5 10.h3 Kb6 11.Kd4 Kc6 12.Kc3 Kd6 13.Kd3 Ke5 14.a4 bxa4 15.bxa4 Kd6 16.Kd4 Kc6 17.f4 h5 18.h4
+/= (0.39) Depth: 23/41 00:01:35 173mN
3...Bb5 4.Bxb5 axb5 5.f5 f6 6.Kd4 Kd6 7.a3 Kc6 8.b3 Kd6 9.h4 Kc6 10.f4 Kd6 11.Kd3 Kc7 12.Kc2 Kd6
+/= (0.39) Depth: 24/43 00:03:28 381mN

I would never have thought this position would cause top engines so much trouble.

[Rolf]

Here is an example:

[D]8/3bkppp/p7/3p4/5P2/2KB1P2/PP5P/8 b - - 0 1

I would never have thought this position would cause top engines so much trouble.

What is the problem with the position? It's won but the engine output shows only a very low advantage for White. Walter, ascomputerchess enthusiast even thought the position could be drawish. In truth it isnt.

I had a little experiment. I took Rybka 3 after, now from the heart, Bb5 xb5 ab5 and now Kd4 which is then already a winner and Rybka sees it. I didnt check the other engines. Rybka then wins after Kd6 f5 h6 h3 with afteranother a3 b3 then a4 bxa4 bxa4 now with a score already in the >3.--

It's sort of zugzwang what Rybka realises because after sometimes g6 fxg6 fxg6 it sees that the Black King must hold the a Pawn and loses time while White King captures the Black Pawns . The score then goes up to >30.--!

I think the output of Rybka in the beginning is not bad. A plus score isnt meaning a draw. But the zugzwang stuff is really a case for clusters and NOT what you and I had at home - BTW for me a little proof for the sense of online hardware boomers. Oerhaps it is also a consequence out of hash problems.

But now the side effects of my quick research!

What is with Robo* and Fritz 12???

Dont they all show the same scores? Are they not hereby proven as clones of the stolen Rybka whatever version? The same Crafty 23? They all show that + .23 stuff after Bb5.

Somewhere I read ok, to be a clone or not depends on the output in the game or position. Well this endgame reveils that they all copied from Rybka or public domain.

[Damir]

Would you stop with this NONSENCE Rolf ? You are out of your mind!