The Chessengine  RobboLito is the best Engine at the Swisstest see http://glareanverlag.wordpress.com/2008/08/08/64-schach-verfuehrungen/

The Swisstest is a set of special problems for Chessengines. Robbolito solves 55, rybka only 52.

RobboLito is a free UCI Chess Engine running only on one CPU with a strength higher then Rybka 3.0 Single CPU. The Program is free of charge and Open Source.

Download with sources here http://www.cyclonechess.com/robbolito.htm

This is good news: plenty of competition at the top.

Of course, special problems are not the best test of engines unless one can be absolutely sure that the developers have not seen the problems beforehand. Engines can be tuned to find problem answers by skillfully weighting internal criteria.

On the other hand, problem sets can be composed which specifically call out those trouble spots that engines sometimes seem to have. All in all, it's fun to have such tests, and the results are often intriguing, as they are here.

If an engine tops both the problem set and the competitive shoot outs, one can be sure that there's real chess playing power there.

Cool, very cool.

It just shows that RobboLito is not cloned from Rybka 3 like some said. But RobboLito is different to Rybka 3, higher depth, stronger at endgames, better in Swisstest ...

Just downloaded that, thanx for the link.

Works well in Scid too : )

Chronicle level game
1) 2 hours during the first 40 moves
2) 1 hour for the next 20
3) 1 hour for the end of the game

RobboLito 0.085g3 x64 - Terminato 0.033r2 x64 [B80]
(1), 29.12.2009
[0.10;0.00]

AMD Athlon(tm) 64 X2 Dual Core Processor 3800+W=24.9 ply; 1.020kN/s B=24.4 ply; 886kN/s 1.Ng1-f3 c7-c5 2.e2-e4 d7-d6 3.d2-d4 c5xd4 4.Nf3xd4 Ng8-f6 5.Nb1-c3 a7-a6 6.Bc1-e3 e7-e6 7.f2-f3 b7-b5 8.Qd1-d2 b5-b4 9.Nc3-a4 Nb8-d7 10.0-0-0 Qd8-a5 11.b2-b3 Bc8-b7 Both last book move 12.Kc1-b1 0.10/20 5:31 Nd7-c5 0.00/21 4:37 13.a2-a3 0.00/21 1:29 Nc5xa4 0.00/19 0 14.a3xb4 0.28/22 8:56 Qa5-c7 0.00/23 0 15.b3xa4 0.28/21 0 d6-d5 0.00/18 0 16.Be3-f4 0.28/21 0 e6-e5 0.00/15 0 17.Bf4-g3 0.28/19 0 d5xe4 0.00/19 0 18.Nd4-b5 (Qd2-g5) 0.28/19 0 18...a6xb5 0.49/20 5:44 19.Bf1xb5+ 0.16/20 5:28 Bb7-c6 (Ke8-e7) 0.44/21 23 20.Bg3xe5 0.51/23 15:50 Qc7-b7 (Qc7-c8) 0.40/23 3:03 21.Be5xf6 0.47/22 5:03 g7xf6 0.40/21 0 22.Rh1-e1 0.47/22 0 Bc6xb5 0.40/21 0 23.Re1xe4+ 0.39/23 3:44 Bf8-e7 0.35/24 0 24.a4xb5 0.39/23 0 0-0 0.35/21 3 25.Qd2-d7 0.39/22 0 Qb7-a7 0.35/21 0 26.Qd7xa7 0.39/21 0 Ra8xa7 0.35/20 0 27.Re4-g4+ 0.39/21 0 Kg8-h8 0.28/22 0 28.Rg4-c4 (Rd1-d5) 0.26/25 14:54 28...Ra7-b7 (Rf8-b8) 0.25/26 6:26 29.Rd1-d5 0.25/28 10:07 Rf8-g8 (Rf8-b8) 0.25/28 0 30.Rc4-g4 0.26/24 4:36 Rg8-c8 0.25/26 0 31.h2-h3 (g2-g3) 0.26/23 0 31...h7-h6 0.10/26 13:40 32.h3-h4 (c2-c4) -0.05/26 7:45 32...Rb7-c7 0.00/27 9:59 33.Rg4-e4 (b5-b6) 0.00/27 3:06 33...f6-f5 0.00/28 8:58 34.Rd5xf5 0.00/29 4:16 Rc7xc2 0.00/29 10:28 35.b5-b6 0.00/30 0 Rc8-c3 0.00/28 46 36.Re4xe7 (Re4-e1) 0.00/29 0 36...Rc2-c1+ 0.00/36 3:13 37.Kb1-b2 Draw accepted 0.00/36 0 1/2-1/2




Terminato 0.033r2 x64 - RobboLito 0.085g3 x64 [B80]
120'/40+60'/20+60' (1), 29.12.2009
[0.10;0.00]

AMD Athlon(tm) 64 X2 Dual Core Processor 3800+W=23.0 ply; 821kN/s B=24.0 ply; 1.014kN/s 1.Ng1-f3 c7-c5 2.e2-e4 d7-d6 3.d2-d4 c5xd4 4.Nf3xd4 Ng8-f6 5.Nb1-c3 a7-a6 6.Bc1-e3 e7-e6 7.f2-f3 b7-b5 8.Qd1-d2 b5-b4 9.Nc3-a4 Nb8-d7 10.0-0-0 Qd8-a5 11.b2-b3 Bc8-b7 Both last book move 12.Kc1-b1 0.10/20 5:15 Nd7-c5 0.00/21 4:54 13.a2-a3 0.11/21 14:29 Nc5xa4 0.19/23 9:47 14.a3xb4 0.11/20 0 Qa5-c7 0.19/23 0 15.b3xa4 0.11/21 0 d6-d5 (Bf8-e7) 0.19/22 0 16.Be3-f4 0.29/20 3:11 Bf8-d6 (Qc7-c8) 0.29/21 7:26 17.Bf1-b5+ 0.33/20 4:47 Ke8-f8 0.32/22 1:08 18.Bf4xd6+ 0.33/20 0 Qc7xd6 0.32/22 0 19.Bb5-e2 0.33/22 16:20 d5xe4 0.34/24 0 20.c2-c4 (Nd4-b3) 0.33/21 0 20...Ra8-b8 0.30/19 7:38 21.b4-b5 0.37/20 0 g7-g6 (a6xb5) 0.30/18 0 22.Nd4-b3 (Nd4-c2) 0.54/19 3:46 22...Qd6-c7 (Qd6-e7) 0.51/21 6:11 23.Kb1-b2 (Qd2-b4+) 0.78/20 6:24 23...a6xb5 0.70/21 10:24 24.a4xb5 0.78/18 0 h7-h5 0.69/22 6:24 25.Qd2-a5 (Qd2-d4) 0.83/23 13:21 25...Qc7xa5 (Qc7-e7) 0.86/21 5:28 26.Nb3xa5 1.02/23 3:25 Kf8-e7 0.95/23 1:31 27.Kb2-b3 0.94/23 1:37 h5-h4 0.95/23 0 28.Na5xb7 0.94/21 1 Rb8xb7 0.95/23 0 29.f3xe4 (Kb3-b4) 0.94/21 0 29...Nf6xe4 0.86/25 11:35 30.Be2-f3 (Rh1-e1) 1.18/27 3:25 30...Ne4-c5+ 0.84/23 5:10 31.Kb3-b4 1.18/27 0 Rb7-c7 0.84/22 0 32.Bf3-c6 (Rd1-d4) 1.18/14 0 32...Rh8-h5 1.26/23 9:27 33.Kb4-a5 1.26/25 20:39 Rc7-a7+ (Rc7-c8) 1.28/25 0 34.Ka5-b6 1.05/25 12:28 Ra7-a4 1.28/25 2:13 35.Rd1-d4 1.33/25 0 h4-h3 (Ra4-a3) 1.28/25 0 36.g2xh3 1.29/24 3:01 Ra4-a3 1.37/24 11:20 37.Rh1-f1 1.45/23 0 Ra3-a2 (Rh5-e5) 1.37/23 8:48 38.Rf1-d1 1.27/22 4:35 Rh5-f5 1.44/22 16 39.h3-h4 1.43/21 0 Nc5-a4+ 1.53/24 7:36 40.Kb6-b7 (Kb6-c7) 1.67/25 0 40...Ra2-b2 1.60/22 2:19 41.Rd1-a1 2.24/23 0 Rb2-b4 1.60/21 0 42.Ra1xa4 (Kb7-a6) 2.36/18 0 42...Rb4xa4 3.15/24 5:42 43.b5-b6 1.90/16 0 Ra4-b4 4.28/25 18:22 44.Rd4-d7+ 4.86/24 0 Ke7-f6 4.42/25 9:30 45.Bc6-b5 3.73/21 0 Rf5xb5 4.51/25 4:26 46.c4xb5 5.82/27 0 Rb4xh4 (Rb4xb5) 4.91/27 4:07 47.Kb7-c8 (Kb7-c7) 5.82/27 1:31 47...Rh4-h5 5.85/27 2:57 48.b6-b7 5.82/25 0 1-0

The Terminato got his shorts Robbolito ... 

last i heard, no proof has ever been shown it was clone just the maker of rybka and some fans claiming it is one.

after all when rybka first came out people also claimed it was a clone of the engine fruit.

robo could be a clone but maybe we should wait before being so harsh towards robo and at people who post about robo, sooner or later it will be proven or disproved anyways.

it doesn't sound hard to believe at all.

Rybka often makes shocking misevaluations about the endgame. I've seen it not realise that a lone king draws against a rook pawn and a bishop that doesn't control the queening square of the pawn. It was giving +5 or +6 to the side with the bishop and pawn.

It's not the only example either....

Till now i did see no hint why Robbolito should be a clone, just rumors and beleves.

Robbolito cant be a clone. Robbolitos depth is much higher then Rybkas, and you can read the sources to see its not fake.  Robbolito is better, so why should it be a clone?

Why are Rybkas producers not going to court with it? They can compare their code to to robbolito's? But they dont...

i checked that, you are right.

rybka wit 4 cpu several minutes:
4k3/8/8/8/8/8/7P/K6B w - - 0 1

Analysis by Rybka 3 32-bit-4:

1. +-  (5.52): 1.Bf3
2. +-  (5.27): 1.Kb1

(,  30.12.2009)

Robbolito 1 CPU vew seconds, depth already 2 time Rybkas depth:

4k3/8/8/8/8/8/7P/K6B w - - 0 1

Analysis by RobboLito 0.085g3 w32:

  =  (0.05)   Depth: 33/41   00:00:05  11880kN
1.Kb2 Kf7 2.Bd5+ Kg7 3.Kc3 Kh8 4.Be4 Kg7 5.Kd4 Kg8 6.Ke5 Kh8 7.h4 Kg8 8.Kd4 Kh8 9.Kd5 Kg8 10.Ke5 Kh8 11.Kd4 Kg7 12.Kc3 Kg8 13.Bd5+ Kg7 14.Kd4 Kh8 15.Be6 Kg7 16.Kc3 Kh8 17.h5 Kh7 18.Bf5+ Kh8


(,  30.12.2009)

It looks like Robbolitos Endgames are clear superior. The Position is a draw but Rybka cant see it.

But Robbolino is not allways thaat superior

A draw position:

8/K1k4n/P7/8/6Pp/8/4B2P/8 w - - 0 1

Analysis by Rybka 3 32-bit-4:

1. +-  (4.40): 7.Bf3 Ng5 8.Bh1 Ne6 9.Bb7 Ng5 10.Bg2 Ne6 11.Bd5 Nf4
2. +-  (4.40): 7.Bf1 Ng5 8.Bg2 Ne6 9.Bd5 Nf4 10.Bf3 Ne6 11.Bb7 Ng5 12.Bg2 Ne6 13.Bd5 Nf4 14.Bf3 Ne6 15.Bb7 Ng5 16.Bg2 Ne6 17.Bd5 Nf4 18.Bf3 Ne6 19.Bb7 Ng5 20.Bg2 Ne6 21.Bd5 Nf4 22.Bf3

(,  30.12.2009)

8/K1k4n/P7/8/6Pp/8/4B2P/8 w - - 0 1

Analysis by RobboLito 0.085g3 w32:

  +-  (4.11)   Depth: 26/47   00:00:15  24392kN
7.Bc4 Ng5 8.Bd5 Nh7 9.Ka8 Ng5 10.Bc4 Nf3 11.Be2 Ng5 12.Bf1 Nf3 13.Ka7 Ng5 14.Bg2 Nh7 15.Ka8 Ng5 16.Bd5 Kb6 17.Bb7 Kc7 18.Ka7 Nh7 19.Bh1 Ng5 20.Ka8 Nf7 21.Bb7 h3 22.Ka7 Ng5 23.Bd5 Nh7 24.Be4 Ng5 25.Bh1 Nf7 26.Bb7
  +-  (4.11)   Depth: 27/47   00:00:22  34991kN

(,  30.12.2009)

here Robbolito s only a few percent better, but still wrong

This Position is from Jussupow ( private conversation )

Everything is different with the Ippolit family and Rybka. The entrie internal structure is rewriteen. This has been mentioned on other places, but usually the Rybka team demands its deletion. If it were the same internally, the cloning proof would be obvoius. Plus they fixed the bugs, and improved it by 50 elo via a vast rewriting, starting from nothing but an decompiler dumping.

@ Kotomitsuki

I checked both positions with some of my engines: Fritz 8, Commet B50, and crafty 19.01 all evaluate the first position as a draw in less than a secound (Although, I must point out that I cannot remember whether I have endgame tablebases on my laptop.)

The secound position however, Fritz gets wrong: evaluating it as +5

You have to compile for the Mac, but it will work. Ian Osgood worked it on a Mac.

Can anyone send me the chess positions used in the Swisstest? I'm curious to know what they are.

Here are some links!

http://www.beepworld.de/members38/eigenmann/swisstest.htm

http://forum.computerschach.de/cgi-bin/mwf/topic_show.pl?tid=2027

http://go2.wordpress.com/?id=725X1342&site=glareanverlag.wordpress.com&url=http%3A%2F%2Frapidshare.com%2Ffiles%2F316097630%2FSwissTest4_Dez09.zip

http://glareanverlag.wordpress.com/2008/08/08/64-schach-verfuehrungen/

I ran Robbolito 0085g3 against Deep Rybka 3.0.  I tried Robbolito 9, but it crashed in the first game after the book opening.  Based on the fact that I also saw 85g3 eventually crash once, I will give version 9 another try later.

Here are the results so far for two different blitz time controls with a standard narrow book, Intel Quad Q6600 at 2.4 GHz:

G5 - Rybka won +31=66-19, winning ~27% of the games.

G10 - Rybka won +12=18-4, winning ~35% of the games.

Though I want to see more games played a the longer time controls, it seems that Deep Rybka is stronger with more time.  Additionally I observed this version of Robbolito was continually in time trouble.  It did lose one game on the sparse side of an opposite-color-bishop-and-rook-pawn ending, where with 0 seconds showing on the clock was feverishly moving its king back and forth, eventually lost on time after many moves of Rybka avoiding repetition of position.  Clear draw which Rybka never saw.  The draw was clear with king on h8 and pawn on g6, just after Rybka took a knight on move 86.  Robbolito eventually "saw" the draw on move 133, losing on time on move 160.  Thank goodness for the invention of tablebases.

So, I also agree with those discussing the weak endgames these things still play.  I am amazed to see their "brilliance" in middlbe game combinations that I do not see from playing over most GMs games, but at the same time to see the miscalculations/misevaluation of relatively simple endgames.

I do have to say that I sit in amazement over some of the middle game combos where the computer evaluates to a significant material advantage that does not come to fruition for ten moves.  The ones I find most impressive are the ones where it makes a small sacrifice to get the big prize, but is not a forced set of moves.  In one instance I saw the computer make the sacrifice, followed by a series of "intermezzo" moves and then pick up the advantage later.  Even if I can't create beauty over the board, I can usually follow the logic of any GM games I review, so I find it interesting to see what these programs come up with, but still do not catch the theme early in the unfolding of events.

Robbolito doesn't support multiple processors yet, so your result doesn't mean one engine is weaker than the other. If you want to test the relative strengths of the engines, you should match the single CPU verions or multiple versions of both.

If you want to test a multi processor engine of the same family against your Deep Rybka, try Firebird. You'll notice it's stronger than Rybka.

 I will definitely try firebird (and Iggorit v7).

BTW I both agree and disagree with your statement. 

Agree:  If by position evaluation/scoring given a fixed depth, then I whole-heartedly agree.  Then we are looking at the "chess knowledge" of the software.  Which I appreciate even more when the computer does not play a stupid endgame.

Disagree: Part of the beauty of the programming and maximizing the strength is the smart use of parallel processing.  If a programmer chooses not to use it, then why penalize the one who does.  I make that statement only for "standard" off the shelf configuration of current processors, though.  If Deep Rybka required daisy chained computers or some other expensive configurations (Deep Blue - hows that for dating myself), then I would agree.

Not that I can beat any of these unless my computer experienced a Single Event Upset in a key position, but it interesting to see the computers getting better to the point that no one on the planet can beat them.  It is a far cry from when I used to beat up on my Novag SuperConstellation as a college student.

Robbolito was programed for single processor systems. At a one processor Computer Robbolito is better than rybka. If you want to use a multiprocessorsystem use FireBird. If you want to compare  vehicle-engines you have to put them in the same vehicle. A engine for a smart should not be compared with a tank engine in a tank and vice versa.

No one on the planet can beat chess engine since jears. Might be interesting to figure out, when the last game happend, where a human won against a 1 class engine.