Test Rybka 3 game

[M ANSARI]

This is one of the games I chose played in a tourney yesterday at home that impressed me. I chose it since it shows some of the aspects of R3 playing style ... very different than R2.3.2a. Please note all comments are mine and were done with quick analysis from an average strength chess player like most of you here.


(101) Rybka 3 - Zappa Mexico II [A00]
Rybka 3 tourney–1 Home (34.2), 24.07.2008
[0.63;0.48]
Intel(R) Core(TM)2 Extreme CPU Q6850 @ 3.00GHz 4095 MHzW=13.3 ply; 298kN/s; 4 TBAs; Perfect 13x.ctgB=14.3 ply; 3,004kN/s; 3,750 TBAs; Perfect.ctg


1.h3 B/0 0 g5 B/0 0 2.d4 B/0 0 h6 B/0 0 3.e4 B/0 0 d5 B/0 0 4.exd5 0.63/14 7 Qxd5 0.48/15 25 5.Nc3 0.74/14 7 Qd6 0.58/14 11 6.Nf3 0.65/14 5 a6 (Nf6) 0.54/14 14 7.Ne4 (Be3) 0.92/15 6 7...Qb6 0.58/14 10 8.Bd3 0.81/14 6 Nf6 (Bf5) 0.58/15 19 9.0–0 (c4) 0.85/13 2 9...g4 (Bf5) 0.48/15 12 10.hxg4 0.84/14 6 Bxg4 0.48/15 13 11.a4 0.72/14 8 Nbd7 (Rg8) 0.47/14 5 12.a5 0.63/13 5 Qc6 0.45/14 5 13.Re1 (Qe2) 0.44/15 23 13...0–0–0 (e6) 0.26/13 8 14.b4 (Bf4) 0.26/14 27 14...Rg8 (e6) 0.18/13 4 15.Rb1 0.30/12 5 Nxe4 (Qa4) 0.02/14 6 16.Bxe4 0.54/13 3 Qb5 (Qd6) 0.18/14 6

[d]2kr1br1/1ppnpp2/p6p/Pq6/1P1PB1b1/5N2/2P2PP1/1RBQR1K1 w - - 0 17

OK, let us spend a few seconds looking at this position. White King looks a little uncomfortable with black rook on g file with no pawn in front ... but black is still slightly under developed. As a human you immediately look at c7 and think that it is a target ... so c4 looks really good ... but in computer play such speculative play without concrete calculation of a variation that wins material can backfire, especially with the defensive resources of engines.

17.c4!! (Bh7)

[d]2kr1br1/1ppnpp2/p6p/Pq6/1P1PB1b1/5N2/2P2PP1/1RBQR1K1 w - - 0 17

Of all the Rybka's ... Rybka Winfinder is the only one that likes this pawn sac. R3 sees that c7 is weak and whips up an attack. Other engines might play this but I could only find Hiarcs that likes this. These type of King attacks are what used to seperate strong GM human play from computer play ... as Kasparov would say "play the move and work out the variations later"...

0.37/13 9 17...Qxc4 0.07/13 8 18.Bf4 1.33/13 19 e5 0.07/14 6 19.Rc1 (Bxe5) 1.33/12 0 19...Qb5 0.98/13 9 20.Qc2!

[d]2kr1br1/1ppn1p2/p6p/Pq2p3/1P1PBBb1/5N2/2Q2PP1/2R1R1K1 b - - 0 20

It is obvious that R3 had seen this strong attack all along since it is playing instantly ... the attack on c7 is relentless

1.33/12 0 20...c6 0.98/13 0 21.d5!

[d]2kr1br1/1p1n1p2/p1p4p/Pq1Pp3/1P2BBb1/5N2/2Q2PP1/2R1R1K1 b - - 0 21

A nice follow up on the attack R3 sacs a bishop for a full bore attack on the black King ... again this shows completely different play than 2.3.2a

1.33/12 0 21...Bxf3 1.06/13 3 22.Bxf3 1.33/11 0 Rg6 (exf4) 1.34/13 7 23.Bd2 (Be3) 1.75/12 3 23...Kc7 1.26/13 2 24.Qa2 (Red1) 1.80/13 1 24...Kb8 (Rd6) 1.36/13 2 25.dxc6 1.80/11 0 bxc6 1.36/12 0 26.Qxf7 1.80/11 0 Rf6 1.55/13 2 27.Qh5 2.10/13 2 Be7 (Rd6) 1.55/12 2 28.Bc3 (Red1) 2.24/12 2 28...Ka8 (Rdf8) 1.78/12 4 29.Bxe5 (Rcd1)

[d]k2r4/3nb3/p1p2r1p/Pq2p2Q/1P6/2B2B2/5PP1/2R1R1K1 w - - 0 29

Now follows a brutal combination that ends it all

29.Bxe5 (Rcd1)4.20/13 20 29...Rxf3 4.98/12 3 30.Qxf3 4.20/11 0 Nxe5 5.47/13 4 31.Rxe5 4.20/11 0 Qxe5 5.47/13 0 32.Qxc6+ 4.20/11 0 Kb8 5.52/14 5 33.Qb6+ 4.20/11 0 Ka8 5.52/14 0 34.Qxa6+ 4.44/13 2 Kb8 5.52/14 0 35.Qb6+ 4.44/12 0 Ka8 5.52/14 0 36.Qc6+ 4.44/12 0 Ka7 5.52/14 0 37.Qc7+ 4.44/12 0 Qxc7 5.52/14 0 38.Rxc7+ 4.44/12 0 Kb8 5.99/15 6 39.Rxe7 4.74/16 1 Rd1+ 7.11/16 7 40.Kh2 4.74/14 0 Rd2 (Rf1) 7.11/16 4 41.f4 (Kg3) 5.51/12 9 41...Rd4 (Rd5) 7.46/13 3 42.f5 5.78/10 1 1–0

Scary stuff !!

[gladius]

Interesting game. What was the time control?

[d]

Fritz 10 finds Bxe5 here immediately with a game ending advantage. Rybka chooses the safer line here, and still converts nicely so just a question of finding the tactics quickly. I wonder how long it takes Rybka to come up with Bxe5?

23.Bxe5 Nxe5 24.Rxe5 Rf6 25.Rf5 Rxf5 26.Qxf5+ Kb8 27.Qxf7 Qxb4 28.dxc6 bxc6 29.Bxc6 Bd6 30.Qf5 Rf8 31.Qd7 Be5
+- (1.82) Depth: 9/27 00:00:00 412kN
23.Bxe5
+- (2.10) Depth: 10/33 00:00:00 772kN
23.Bxe5
+- (2.38) Depth: 11/32 00:00:00 1280kN
23.Bxe5
+- (2.66) Depth: 11/33 00:00:00 1605kN
23.Bxe5
+- (3.22) Depth: 11/36 00:00:01 2144kN
23.Bxe5 Nxe5 24.Rxe5 Rf6 25.Rf5 Rxf5 26.Qxf5+ Kb8 27.Qxf7 Rc8 28.dxc6 bxc6 29.Rxc6 Rxc6 30.Bxc6 Qxc6 31.Qxf8+ Kb7 32.Qe7+ Ka8 33.g4 Qc1+ 34.Kg2 Qc6+ 35.f3
+- (3.37) Depth: 12/38 00:00:02 4138kN

[M ANSARI]

This was a 5 0 game.

[Dr.Wael Deeb]

Thanks Majd for the game and the comments,it's realy a scary stuff
An outstanding positional understanding....I guess the supporters of the human strategy superiority over the machines will remain without comments reading this post....

[Marc MP]

Thanks Majd for the game and the comments,it's realy a scary stuff
An outstanding positional understanding....I guess the supporters of the human strategy superiority over the machines will remain without comments reading this post....

I identify myself as a "supporters of the human strategy superiority over the machines", for the time being (who knows maybe in the future that will change?).

In my opinion, the best humans are still far superior in strategic planning in closed positions. The game presented here (at least from what I see from the diagram - I didn't go through carefully) is semi-open or clearly opened. In that type of games I've no problems admitting that engines are superior (in general, not in every position) to humans. But I still firmly think that humans are stronger for planning and playing closed positions. Of course this is an opinion, not backed with data, but I'm sticking to it.

For the anecdote, I was analysing a game, closed position - King's Indian, with friends of mine that are good club players (say 2200-2300 elo) last year. We were analysing on one of the guy laptop equiped with the latest Rybka then. What happened? They quickly wanted switch off the engine because they felt Rybka was suggesting "stupid" or "nonsense" moves. I recall, this was a closed (but not blocked position). I could see Rybka's eval going down 10 moves later, when the pawn breakthrough came in.

With today's top of the line hardware and Rybka 3, I think those effect will be lessened ... to some extend only. To avoid the 50 moves rule, you must search 100 plies and deeper. From what I read yesterday on the rybka forum, Rybka 3 will only get you to 27-28 plies on a Skulltrail (some kind of powerful 8 cpu thing I believe) after 5 days in average. We are far from 100 and over.

Go ask the best player in your local club about engine usefulness in a closed position for strategic planning. I'm almost sure he will tell you it is worth close to nothing...

[Dr.Wael Deeb]

Thanks Majd for the game and the comments,it's realy a scary stuff
An outstanding positional understanding....I guess the supporters of the human strategy superiority over the machines will remain without comments reading this post....

I identify myself as a "supporters of the human strategy superiority over the machines", for the time being (who knows maybe in the future that will change?).

In my opinion, the best humans are still far superior in strategic planning in closed positions. The game presented here (at least from what I see from the diagram - I didn't go through carefully) is semi-open or clearly opened. In that type of games I've no problems admitting that engines are superior (in general, not in every position) to humans. But I still firmly think that humans are stronger for planning and playing closed positions. Of course this is an opinion, not backed with data, but I'm sticking to it.

For the anecdote, I was analysing a game, closed position - King's Indian, with friends of mine that are good club players (say 2200-2300 elo) last year. We were analysing on one of the guy laptop equiped with the latest Rybka then. What happened? They quickly wanted switch off the engine because they felt Rybka was suggesting "stupid" or "nonsense" moves. I recall, this was a closed (but not blocked position). I could see Rybka's eval going down 10 moves later, when the pawn breakthrough came in.

With today's top of the line hardware and Rybka 3, I think those effect will be lessened ... to some extend only. To avoid the 50 moves rule, you must search 100 plies and deeper. From what I read yesterday on the rybka forum, Rybka 3 will only get you to 27-28 plies on a Skulltrail (some kind of powerful 8 cpu thing I believe) after 5 days in average. We are far from 100 and over.

Go ask the best player in your local club about engine usefulness in a closed position for strategic planning. I'm almost sure he will tell you it is worth close to nothing...

Thanks for the input
While I do agree with you that in some closed positions the engines don't have a clue what to do,these positions are not much and how many times are you going to reach such positions in a 100 games match for example

[George Tsavdaris]

Thanks Majd for the game and the comments,it's realy a scary stuff
An outstanding positional understanding....I guess the supporters of the human strategy superiority over the machines will remain without comments reading this post....

I identify myself as a "supporters of the human strategy superiority over the machines", for the time being (who knows maybe in the future that will change?).

In my opinion, the best humans are still far superior in strategic planning in closed positions.

Actually i think you overestimate machines in the above.
In my opinion GMs and IMs and FMs are still far far far superior in strategic planning in closed positions.

And actually in all kind of positions, open, semi-open, closed FMs, IMs and GMs are far far superior at long term planning. Computers on the other hand are far far far far superior in short term planning which is tactics.

But Chess is such a game where in most cases these short strategic planning=tactics, are enough for taking the advantage and keeping it until the end of the game, while long strategic planning is far far less important.
If you can create an amazing plan that will bring you a good passer 20 moves ahead(40 ply), but miss a tactic in the midway, it's pointless.
The computer may not create any long term plan at all, but these short plans are enough.
That's how is the situation today with humans-computers.....

With today's top of the line hardware and Rybka 3, I think those effect will be lessened ... to some extend only. To avoid the 50 moves rule, you must search 100 plies and deeper. From what I read yesterday on the rybka forum, Rybka 3 will only get you to 27-28 plies on a Skulltrail (some kind of powerful 8 cpu thing I believe) after 5 days in average. We are far from 100 and over.

The situation is ever worse, since these 27-28 plies are pruned plies and not full searched plies, so the engine may lost many good moves with the pruning that took place.

[Marc MP]

Thanks for the input

Your are very welcomed Dr. Deeb! Even if I sound like cynic about computer play, believe me that I respect very much their play in most positions! Hopefully I made that clear in my last statement. I'm very happy to see you keep your mind open about this, that tells, in my mind, a lot about you now and your possibilities about improving in all domain, not only chess (beware of my kind words - sometimes they are cursed! ). Engines are getting stronger every day, I've no problem admitting that, they play bettey than Kasparov in a lot of positions, but not in all of them.

While I do agree with you that in some closed positions the engines don't have a clue what to do,these positions are not much and how many times are you going to reach such positions in a 100 games match for example

True, but we always thinks as humans playing "anti-computer" stuff. But I never seen a comp-booker saying "my openings are anti-humans", i.e. open positions only.

At the point we are at, opening prep is, in my opinion(!), the key for success or failure when comparing humans to computers for a game from start to finish. For a game from a random position, it will depend on the closedness of the position....

Pro-humans and pro-engines need to echange more.. it will benefit them both in the long run....

[Marc MP]

Thanks Majd for the game and the comments,it's realy a scary stuff
An outstanding positional understanding....I guess the supporters of the human strategy superiority over the machines will remain without comments reading this post....

I identify myself as a "supporters of the human strategy superiority over the machines", for the time being (who knows maybe in the future that will change?).

In my opinion, the best humans are still far superior in strategic planning in closed positions.

Actually i think you overestimate machines in the above.
In my opinion GMs and IMs and FMs are still far far far superior in strategic planning in closed positions.

And actually in all kind of positions, open, semi-open, closed FMs, IMs and GMs are far far superior at long term planning. Computers on the other hand are far far far far superior in short term planning which is tactics.

But Chess is such a game where in most cases these short strategic planning=tactics, are enough for taking the advantage and keeping it until the end of the game, while long strategic planning is far far less important.
If you can create an amazing plan that will bring you a good passer 20 moves ahead(40 ply), but miss a tactic in the midway, it's pointless.
The computer may not create any long term plan at all, but these short plans are enough.
That's how is the situation today with humans-computers.....

With today's top of the line hardware and Rybka 3, I think those effect will be lessened ... to some extend only. To avoid the 50 moves rule, you must search 100 plies and deeper. From what I read yesterday on the rybka forum, Rybka 3 will only get you to 27-28 plies on a Skulltrail (some kind of powerful 8 cpu thing I believe) after 5 days in average. We are far from 100 and over.

The situation is ever worse, since these 27-28 plies are pruned plies and not full searched plies, so the engine may lost many good moves with the pruning that took place.

Thank you for your thought George,

Believe me I learned from this. I (previously) wrongly associated you with hard-core pro-engines (At least for the ' hard-core pro-engines' that exist in my mind!) . You just proved me wrong. Thank you, that just tells about why the CCC forunm is there: exchange of ideas. Sometimes you teach someone, some other times that person is teaching you. That is how Robert J. Fischer viewed chess from what I recall.

Have a good day,

[M ANSARI]

Yes ofcourse ... there are many positions where humans are far superior to engines. Humans can quickly identify a situation where a blocked position is possible or a fortress is possible with no further possibility of progress even though there is a material deficit. This usually happens after the human has survived the opening and middle game stage ... and only in certain conditions. These situations are getting exceedingly rare ... and engines are starting to handle even those positions better. I would guess this human superiority over engines in these rare situations will also disappear as progress in hardware and software improves. Already engines are being programmed to avoid such situations.