Rybka vs top human match

[Terry McCracken]

Let's make one thing clear. There are engines and then there is Rybka playing on a minimum of Quad core and 64 bit OS. This Rybka is a formidable opponent that cannot be compared to a single CPU Shredder 9 or CM series ... the difference is tremendous. I know that for a fact because I was able to get a win against almost every engine up till Shredder 10. Rybka is totally different ... it plays moves that I cannot comprehend that seem wrong ... but then are proven right. I gave up trying to beat it a long time ago. Ofcourse you can find little chinks in its armour ... but really it a formidable opponent against any Super GM and I doubt that any human can even come close to challenging Rybka on Quad core 3.2Ghz running on 64bit OS with 6 men TB's and a good book. Thus you HAVE to handicap Rybka to even the playing field. Vas estimates that removing Rybka's book will cost it maybe 50 to 80 ELO points. I have a feeling that might be true on general terms ... but I think a well prepared GM might take advantage of that and make the difference more like 200 or 300 ELO's. I would think that a 2650+ GM would have a 50% chance to win a match with such conditions as mentioned in the proposed match.

I think there are a few left ANSARI, but I'll concede not very many, and they'll need a good deal of computer chess play experience. I'd never compare such a beast with S9 single or smp or CM10. Rybka smp 64bit is simply in a league of its own.

Hydra would lose, and we know Hydra is very strong, but Shredder 8 three years ago playing the first ver. of Hydra gave it a very hard time.

Rybka would have cut off every head! So I imagine Rybka 64 on a dual quad, would do the same to the newer faster Hydra, that won't come out and play!

Shredder 8 didn't give the first version of Hydra a hard time at all, Shredder only managed 5 draws. You can read about it here:
Hydra beats Shredder by the score 5.5-2.5

I know all about it. Shredder was the only software program at that time that could do that well....drawing such a beast isn't easy.

The score isn't that bad for Shredder, imo.

[Vempele]

Shredder 8 didn't give the first version of Hydra a hard time at all, Shredder only managed 5 draws. You can read about it here:
Hydra beats Shredder by the score 5.5-2.5

I know all about it. Shredder was the only software program at that time that could do that well....drawing such a beast isn't easy.

The score isn't that bad for Shredder, imo.

Especially considering the version used in the match was buggy and about 150 points weaker than Shredder 9 according to Sandro Necchi.

That would have put Hydra at about S9's level but with huge error margins, of course. I guess Shredder just got extremely lucky.

[Leto]

Let's make one thing clear. There are engines and then there is Rybka playing on a minimum of Quad core and 64 bit OS. This Rybka is a formidable opponent that cannot be compared to a single CPU Shredder 9 or CM series ... the difference is tremendous. I know that for a fact because I was able to get a win against almost every engine up till Shredder 10. Rybka is totally different ... it plays moves that I cannot comprehend that seem wrong ... but then are proven right. I gave up trying to beat it a long time ago. Ofcourse you can find little chinks in its armour ... but really it a formidable opponent against any Super GM and I doubt that any human can even come close to challenging Rybka on Quad core 3.2Ghz running on 64bit OS with 6 men TB's and a good book. Thus you HAVE to handicap Rybka to even the playing field. Vas estimates that removing Rybka's book will cost it maybe 50 to 80 ELO points. I have a feeling that might be true on general terms ... but I think a well prepared GM might take advantage of that and make the difference more like 200 or 300 ELO's. I would think that a 2650+ GM would have a 50% chance to win a match with such conditions as mentioned in the proposed match.

I think there are a few left ANSARI, but I'll concede not very many, and they'll need a good deal of computer chess play experience. I'd never compare such a beast with S9 single or smp or CM10. Rybka smp 64bit is simply in a league of its own.

Hydra would lose, and we know Hydra is very strong, but Shredder 8 three years ago playing the first ver. of Hydra gave it a very hard time.

Rybka would have cut off every head! So I imagine Rybka 64 on a dual quad, would do the same to the newer faster Hydra, that won't come out and play!

Shredder 8 didn't give the first version of Hydra a hard time at all, Shredder only managed 5 draws. You can read about it here:
Hydra beats Shredder by the score 5.5-2.5

I know all about it. Shredder was the only software program at that time that could do that well....drawing such a beast isn't easy.

The score isn't that bad for Shredder, imo.

Shredder was the most popular and strongest engine of that time and had won many tournaments and championships, Hydra on the other hand was barely known. I don't even think Hydra had won any tournaments before its match with Shredder.

We know Hydra played in April 2004 in ICT4, it finished 1.5 points behind the winner Shredder 8.

Even up to a couple weeks before the match with Shredder, the Hydra team was having trouble improving Hydra's stability, as can be seen here:
http://www.hydrachess.com/main.cfm?midd ... cfm&id=101
http://www.hydrachess.com/main.cfm?midd ... cfm&id=102

In the match in Agust 2004, many described it as a shock after Hydra won the first two games: http://www.hydrachess.com/main.cfm?midd ... cfm&id=138

I remember that in those days, just before the match began, the playchess engine room was always filled with Shredder 8 users, I remember that I was sure Shredder 8 would win the match. So I too was shocked at the beating Shredder 8 took.

[Thomas Mayer]

Hi Mike,

[...]
Anyway, whatever the outcome may be, I doubt that odds matches should be considered as something "normal" from now on if GMs fail. I think odds matches are interesting and useful as experiments, but will remain something unusual. I still doubt that a "super" GM 2700+ would want to receive odds. It's like branding him to be weaker than the comp. Even if he most probably is weaker, it will be preferable for him to lose normal games in normal matches, not involving odds.

also I am quite unsure if odds are really odds... In my opinion a human might have even more problems in unusual situations then a computer which still computes all moves with his full power. I don't know which 3-moves openings Larry will chose, but I believe that it will be some uncommon not so deeply studied but still equal sidevariants. Of course Rybka might not yet be as strong to find moves that hundreds of years of chess studies have found, but at least it will find moves that are far from being easy to refutate. I hope that this match will happen, but I am wondering whether Rybka will open then with e4, d4 or answer on e4 with c5 or e5... If it does - well, then the Rybka Team consist of real Bravehearts...

Greets, Thomas

[ArmyBridge]

Hi
the topic is very interesting, personally I think that in some games, or in some positions engines perfomence is fantastic and it simulates "creativity" just check the style of Deep Junior,by the way the Mikhail Tal of compu-chess world, I played trough many incredible games full of spectacular sacrifices if someone hide the name of Junior from these games and ask you that guess who is the author of these games maybe u would affirm that the author of the game is the great Misha, check this that wrote maxchgr time ago "I really wanted to show you guys this. This is two games between Deep Junior 10 and Rybka 2.2, playing a position with reverse colors in the second game.

The position is from a Kramnik - Kasparov game that I found, where Kasparov sacrifices his queen in the opening for two minor pieces, and the resulting position is highly interesting.

Junior 10 beat Rybka 2.2 with both white and black. If you look at the games, Junior 10 handles the position much better for both sides, and its play is especially beautiful with Kasparov's side. As you'd see in Game 2, where it uses Kaspy's side, it rather keeps its dark square bishop than win the exchange, and I think this is the main reason that Junior understand the position better. I really recommend you to play over the games, at least game 2. "


Deep Junior 10 - Rybka 2.2 mp 32-bit [E92]
DJunior vs. Rybka2.2 Blitz:20'+1" (1), 18.11.2006

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 0-0 6.Be2 e5 7.d5 a5 8.Bg5 h6 9.Bh4 Na6 10.0-0 Bd7 11.Nd2 Nc5 12.b3 Nfxe4 13.Bxd8 Nxc3 14.Qe1 Rfxd8 15.Nb1 Nxe2+ 16.Qxe2 e4 17.Na3 Bxa1 18.Rxa1 Re8 19.Nc2 Kg7 20.Qe3 Re5 21.h3 h5 22.Rf1 Rae8 23.Qc3 Ra8 24.Nd4 f6 25.a3 Ree8 26.b4 Na4 27.Qb3 axb4 28.axb4 Kf7 29.Rc1 Nb6 30.c5 Ba4 31.Qb1 Nxd5 32.cxd6 cxd6 33.Qa2 Bc6 34.Qb3 Re5 35.b5 Bd7 36.Rc7 Ke8 37.Rxb7 Ne7 38.b6 e3 39.fxe3 Rea5 40.Ne6 Bc6 41.Ng7+ Kf8 42.Qe6 Ra1+ 43.Kh2 R1a2 44.Qxe7+ 1-0

Rybka 2.2 mp 32-bit - Deep Junior 10 [E92]
DJunior vs. Rybka2.2 Blitz:20'+1" (2), 18.11.2006

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Nf3 0-0 6.Be2 e5 7.d5 a5 8.Bg5 h6 9.Bh4 Na6 10.0-0 Bd7 11.Nd2 Nc5 12.b3 Nfxe4 13.Bxd8 Nxc3 14.Qe1 Raxd8 15.Nb1 Nxe2+ 16.Qxe2 e4 17.Nd2 Rde8 18.Rad1 f5 19.Nb1 f4 20.Qd2 a4 21.bxa4 f3 22.g3 e3 23.fxe3 Ne4 24.Qc2 f2+ 25.Kg2 Bg4 26.Nd2 Nxd2 27.Rxd2 Bf3+ 28.Kh3 g5 29.Qg6 Be4 30.Qh5 g4+ 31.Kh4 Bf5 32.Qxe8 Rxe8 33.Rdxf2 Bd3 34.Rc1 Rxe3 35.Kxg4 Bd4 36.Rd2 Bc5 37.h4 Be2+ 38.Kh3 h5 39.Kg2 Bb4 40.Rb2 Bf3+ 41.Kf1 Ba3 42.Rcb1 Bxb2 43.Rxb2 b6 44.a5 bxa5 45.Rb7 Re7 46.Kf2 Bg4 47.a4 Kf7 48.Ra7 Kf6 49.Rxa5 Re2+ 50.Kg1 Rc2 51.Rb5 Rxc4 52.a5 Ke5 53.Kf2 Rc2+ 54.Ke1 Ra2 55.Rb7 Rxa5 56.Rb1 Kxd5 57.Rb2 c5 58.Kf2 c4 59.Ke3 Ra3+ 60.Kf4 c3 61.Rf2 Kd4 0-1

So the humans have talent and creativity but they comment tactic mistakes that machines not comment. therefore I believe that the situation this made level. the humans have better estrategical understanding but the machine are better in tactis and is hard very hard beat them (in spite of its weaknesses of estrategica understanding) just ask Kramink overplayed Fritz , but....he Lose the match it is why at the present time the machines are better than Human top GM
Hasta la vista baby

[FICGS]

As a correspondence chess player, I always find that chess engines (even Rybka) are very far from real GM level. It's all a question of time and nerves. I expect Rybka to have serious opponents in the next months or years, probably Fritz has the best basis for this. To implement chess knowledge will become a question of time. We'll see the elo 3200 mark crushed, 3300, maybe more... Does it mean something ? Is Rybka really stronger than Fritz/Shredder against a GM ? I'm still not convinced...

In my opinion, Hydra could beat Rybka but as it seems to be his opposite, the result could be really aleatory in a 20 games match, anyway the games could be nice to see. Rybka tries to play like a GM, Hydra is a strong computer.. Hard to say

[M ANSARI]

I think that by far ... Rybka is the most difficult engine for a human to handle. Junior is a little too aggressive and plays some attacks that turn out to be unsound. With enough time, a super GM will have no problem refuting an unsound line. Rybka is a little more conservative, but it has an evaluation that will go into unbalanced positions that are very difficult for a human to evaluate ... until it is too late. I think Kamsky (who has played Rybka a lot) has confirmed this several times.

Having said that, if the GM manages to go for a line that say turns out to be some of the critical sacrificial lines in the Sicilian ... Rybka will easily get crushed, because it would be trying to refute theory that has been analyzed by thousands of GM's over many years, and most of the compounded analysis would have been done with the help of computers. The thing is ... will a GM be able to steer Rybka into such positions.