Not logged inRybka Chess Community Forum
1 2 3 4 Previous Next  
Parent - By lkaufman (*****) Date 2007-04-12 21:36
     Since the GM will get only around a thousand dollars for a drawn match vs. six thousand for a match victory, he is very unlikely to set out with the goal of making six draws, especially since it is very unlikely that he would be successful. Of course, depending on who the GM is, he may not take excessive chances, so we may very well have several draws; that's just the way high level chess usually is played. Rybka will have some contempt factor to try to minimize this. Depending on the GM, I might expect four draws and two wins for Rybka. Rybka will presumably get a poor position from the opening every game, but hopefully not a lost one.
Parent - By theoak (**) [us] Date 2007-05-05 22:08
This is a highly acceptable playing style! Imagine a player like Petrosyan and Karpov only way stronger. What is wrong with that?
Parent - - By Berfomet (**) Date 2007-04-12 18:17
Rybka is a little bit drawish, what if the GMs aim for a draw and not for the win? What are your views about this?

And also I can draw rybka 2.3.1 (with a shallow opening book on a single 2.66) on a 4sec to "my choice of time (indefinite)" handicap match on a lucky day, and I am only around 1500-1700 rating.

One Example is this:
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=797
Parent - - By lkaufman (*****) Date 2007-04-12 21:42
    This shows the danger of playing book without having enough book. If you want to try an experiment, try playing ten games with Rybka, using the same time handicap (infinity vs. 4'), you take white, but every game have Rybka play a non-book opening move, like 1...h6 or even 1...f6, so she's on her own. Also turn off ponder (thinking on your time). Also set contempt for half a pawn. Let us know if you can score at all.
Parent - By Berfomet (**) Date 2007-04-13 19:19
Hi Larry,

I decided to play with rybka in a handicap, but I did not influence the rybka's moves but I play unorthodox.

Its here:
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=819

For now I only played 2 games more in the future.
Parent - By Uri Blass (*****) [il] Date 2007-04-12 21:47
This is only because you use the default setting so rybka does not know she is playing against weaker player.

Rybka does not play Qc2 with a positive contempt factor of 0.15 pawns.

Here is the analysis
Rybka 2.3.1 32-bit - Me
r2qk1nr/ppp2ppp/2n1b3/1Bb5/3pP3/5N2/PP3PPP/RNBQ1RK1 w kq - 0 1


Analysis by Rybka 2.3.1 32-bit :

8.Qd1-c2
  =  (-0.09)   Depth: 2   00:00:00
8.Qd1-c2
  =  (-0.16)   Depth: 3   00:00:00
8.b2-b4
  =  (0.00)   Depth: 3   00:00:00
8.b2-b4
  =  (-0.02)   Depth: 4   00:00:00
8.b2-b4 Bc5-b6
  =  (0.01)   Depth: 5   00:00:00
8.b2-b4 Bc5-b6 9.Bc1-b2
  =  (0.01)   Depth: 6   00:00:00  7kN
8.b2-b4 Bc5-b6 9.Bc1-b2 Qd8-d6
  =  (-0.18)   Depth: 7   00:00:00  14kN
8.Qd1-c2 Bc5-b6 9.Bb5xc6+ b7xc6 10.Qc2xc6+ Be6-d7 11.Qc6-c4
  =  (-0.11)   Depth: 7   00:00:00  21kN
8.Nb1-d2 Ng8-e7 9.Nd2-b3 Be6xb3
  =  (0.11)   Depth: 7   00:00:00  28kN
8.Nb1-d2 Ng8-e7 9.Nd2-b3 Be6xb3 10.Qd1xb3 0-0
  =  (-0.03)   Depth: 8   00:00:00  34kN
8.Nb1-d2 Ng8-e7 9.Nf3-g5 Bc5-b4 10.a2-a3 Bb4xd2
  =  (-0.03)   Depth: 9   00:00:01  84kN
8.Nb1-d2 Ng8-e7 9.Nf3-g5 Ne7-g6 10.Ng5xe6 f7xe6 11.Nd2-b3 Qd8-d6
  =  (-0.06)   Depth: 10   00:00:03  154kN
8.Nb1-d2 Ng8-e7 9.Nf3-g5 Bc5-b4 10.Ng5xe6 f7xe6 11.Qd1-g4 Qd8-d7 12.Qg4xg7 0-0-0
  =  (-0.06)   Depth: 11   00:00:05  252kN
8.Nb1-d2 Ng8-e7 9.Nf3-g5 Qd8-d7 10.Nd2-f3 a7-a6 11.Nf3-e5 Qd7-d6 12.Ne5-c4
  =  (-0.11)   Depth: 12   00:00:13  637kN
8.Nb1-d2 Ng8-e7 9.Nd2-b3 Be6xb3 10.Qd1xb3 0-0 11.e4-e5 Qd8-d7 12.Bc1-d2 a7-a5 13.a2-a3
  =  (-0.11)   Depth: 13   00:00:24  1248kN
8.Nb1-d2 Ng8-e7 9.Nd2-b3 Be6xb3 10.Qd1xb3 0-0 11.e4-e5 Qd8-d7 12.Bc1-d2 a7-a5 13.a2-a3 Ne7-f5
  =  (-0.03)   Depth: 14   00:00:39  1998kN
8.Nb1-d2 Ng8-e7 9.Nd2-b3 Be6xb3 10.Qd1xb3 0-0 11.e4-e5 Ne7-g6 12.Qb3-c4 Bc5-b6 13.Bb5xc6 b7xc6 14.Qc4xc6 Rf8-e8
  =  (-0.04)   Depth: 15   00:01:09  3621kN

(,  13.04.2007)

Uri
Parent - - By Uri Blass (*****) [il] Date 2007-04-12 21:53
I understood the offer was to any grandmaster who can win a six game match from Rybka.

Of course I understand that you prefer stronger GM's but based on my understanding the offer in the first post was for all GM's who are interested in it.

Uri
Parent - - By lkaufman (*****) Date 2007-04-12 23:08
My offer was for any GM, but some of the offers to add to it are only valid once, for the highest rated player to accept. If we play the match and win, we'll see if there is interest on either side for another match. If Rybka wins decisively against a strong GM, it may be clear to all that subsequent matches must be at pawn odds or the like.
Parent - - By plicocf (***) [br] Date 2007-04-17 04:40
Please IM LK, no pawn odds, let´s to play chess. On the subsequent matches, if Rybka
wins decisively, give more advantege to the GM. If Rybka lost decisively, less advantages.
Paulo Soares
Parent - By Uri Blass (*****) [il] Date 2007-04-17 07:13
The only more advantage that I can think about that can make the match interesting is a correspondence game when the human is allowed to make notes to analyze but not to use engines.

The main problem with a match like that is to be sure that the human does not cheat.

Senior Master Steve Ham lost 2.5:1.5 against computers in these conditions some years ago but it is possible that there are significantly better players than him.

I know that world champion GM Léotard, Christophe won 3.5:0.5

By searching with google I could find the following links for his games:

http://ajec-echecs.org/analyses/lh.php
http://ajec-echecs.org/analyses/lct.php
http://ajec-echecs.org/analyses/ctl.php
http://ajec-echecs.org/analyses/hl.php
Parent - - By lkaufman (*****) Date 2007-04-17 13:21
   If we're talking about live games (not correspondence), what other advantages are left to offer, besides material (or free moves, or bad openings, etc.)? We can of course make Rybka play worse by cutting her time even more, not allowing ponder, etc. but that defeats the point of these matches -- to show off what Rybka can do at her best. What would you suggest (in case of decisive Rybka victory)?
Parent - - By Uri Blass (*****) [il] Date 2007-04-17 15:09
I can think of the following possible advantages:

1)allow the human to use notes or to analyze the game in an additional board during the game.

2)allow the human to take back moves during the game when the time of the human is reduced for doing it and the machine earns the time that is reduced(if the human has an increasement of 1 minute and 2 minutes is reduced from the clock for every take back then it is impossible to take back moves forever).

Uri
Parent - - By Gaмßito (****) [cr] Date 2007-04-17 17:50
allow the human to take back moves during the game when...

Take back moves?

Please, please, stop to degrade the chess in that way. That is not true chess.
First : Odd pawns, then more available time to think, then take back moves (??).

That is totally ridiculous. A jeer to our game!

Regards,
Gambitto.

Parent - - By Uri Blass (*****) [il] Date 2007-04-17 18:44
What you say is totally ridicilous

I do not degrade chess.
The opposite
Chess when the human is allowed to take back move is chess of higher level.

Uri
Parent - - By Gaмßito (****) [cr] Date 2007-04-17 19:52 Edited 2007-04-17 20:03
Of course that the game will be better (will be less errors) for the reason of being able to return the movements.  But that is not real chess.

If I play with X program in a match and i can take back some moves, then have it for for sure that I will be able to draw easily or maybe to win. And that does not have any sense. That is 100 ridiculous%. And if a really strong player would play with those incredible advantages? Certainly the games, would be with less errors but also very boring, without any emotion. And undoubtedly a strong GM should be able to win a match against a very strong Computer. But, how sense has to win this way? Does some emotion exist in playing this way? There is not beauty in playing this way.

Take backs is the lowest thing that you can arrive. But of course, maybe in the future when we humans cannot play against the machines with any positive results, it will be the appropriate option. The chicken option. Excuse me but this is true.

Regards,
Gambitto.

 
Parent - - By Uri Blass (*****) [il] Date 2007-04-17 20:42 Edited 2007-04-17 20:44
I am not sure that you will be able to easily draw by taking back moves.
If the program is strong enough even taking back moves is not going to help you when you cannot take back moves forever.

The games are not going to be boring
It is the opposite(in case that rybka with 3 move opening book move wins the match convincingly and larry kaufman asked what to do after rybka wins the match convincingly)

If the GM gets no advantage the games are going to be boring without emotion because no GM is going to believe that he has even 1% chance to win the match or even to draw the match.

Uri
Parent - - By Gaмßito (****) [cr] Date 2007-04-18 01:45 Edited 2007-04-18 01:48
I am not sure that you will be able to easily draw by taking back moves.

Maybe not so easy for me, but for any other very strong player, it will be easy and it is a lot of advantage. I think that with only 3 take backs per game, any TOP GM, can win against Rybka or against any other program in a match under normal time controls. It would be very difficult that a TOP GM can lose under those conditions.

The games are not going to be boring

Yes, boring is not the appropriate word. I did not want to say that. With take backs there is not real emotion in the game.
The errors are part of the chess game. And without small errors there are not brilliant games.

If the GM gets no advantage the games are going to be boring without emotion because no GM is going to believe that he has even 1% chance to win the match or even to draw the match.

That is not true. All the GM´s of the planet, can think that some expert GM in anti-computer chess, maybe does not have any possibilities to win a single game in a match against Rybka under normal conditions, but this expert GM can play and bring us  some very interesting, good and pretty games. Don't you believe?

Regards,
Gambitto.

Parent - - By Uri Blass (*****) [il] Date 2007-04-18 06:02
This is only your opinion
My opinion is that it is not clear that top GM's are going to win a match against rybka even if 3 take back per game are allowed.

take back does not mean no errors because there are errors that you understand that they are errors too late.
The GM may understand that he did a losing mistake only 4 moves after the losing mistake and in that case 3 take back will not going to help
the GM.

Uri
Parent - - By Gaмßito (****) [cr] Date 2007-04-18 13:40
take back does not mean no errors because there are errors that you understand that they are errors too late.
The GM may understand that he did a losing mistake only 4 moves after the losing mistake and in that case 3 take back will not going to help the GM.


Yes, this is a good point. It would not be useful 3 or more take backs if the error was made several moves before. That
is very clear. But imagine, what it had happened if Kramnik has played with that rule in the previous match against Fritz?  He had probably been able to draw game 2. He had returned his ugly move Qe3?? and another story had happened. He had a better position in that game. In his  match in Bahrein, against the previous Fritz version, also with his move Qc4?? followed by Ne7+ 1-0. If he has been able to return this move, maybe he had been able to win the match in Bahrain against Fritz. We do not know.. 

However my opinion at the moment about take backs, is that is too much advantage for the human. And of course, for a very strong GM like Kramnik or many other Top players is too much advantage.

Reagards,
Gambitto.
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-04-19 11:45
I'd guess that three takebacks per game are worth around 50-75 Elo for a human.

The problem with this idea is that it's awkward for the audience.

Vas
Parent - - By Gaмßito (****) [cr] Date 2007-04-19 14:17
Hi, Vas

Yes, difficultly the idea will be received in a good way by the public.  I do not know. But personally, I don't like that.

I strongly believe that we still can offer much more fight under normal conditions with tournament control times. Of course, at the moment it is already quite very difficult, but our options are not still lost. There are very strong players with extraordinary capacities and they know enough how to play against the machines.

Regards,
Gambitto.
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-04-21 11:34
Gambitto,

humans and computers are good at different things. At many things, top players and even amateur players are better than Rybka. The problem is that, under tournament time control conditions, the things that Rybka is good at are just more important. She doesn't blunder, she doesn't make the sort of minor oversights that humans constantly make, she is objective and practical, she is consistent and sticks to her plan, she doesn't get tired, etc. At 40/120, this is enough to win.

Vas
Parent - By rivaldo (***) [de] Date 2007-05-06 01:40
I'd be happy if kramnik accepted the challenge :-)
I mean it would be cool, if he demonstrated, that he thinks he can win. he's probably not interested in 1500$ appearence fee.
if he took this challenge for relatively few money, everybody would cheer him.
Parent - By Legendary (***) Date 2007-05-09 23:09
The GM should easily win this match.
Parent - - By bluemax (**) [ie] Date 2007-07-06 13:35
The Kramnik Fritz10 match and offering big money just to play brings up a problem: to get a top GM you have to pay them to play in the first place. Will the fans and organisers get their moneys worth.?
I think the K-F10 match was the worst yet of all the computer-human matches, except for the great Fritz win (maybe Kramnik was tired after the WC).

Good and fair conditions are the best thing for the future of computer-human chess and Vas and co. are leading the field.
Parent - - By Gaмßito (****) [cr] Date 2007-07-06 22:21
I do not agree with you that K-F10 match was the worst match. Fritz did not show a clear superiority in this match.  
 
In game two, Kramnik had a better position before the blunder. And, in the last game Kramnik choose an opening that it was not appropriate to their style. Very strange that he tried to play a risky opening, perfect knowing that it is important to avoid the risks and the shape openings. For me, he threw the last game.

Regards,
Gambito.
Parent - By insipid (**) [us] Date 2007-07-06 22:32
Well he took the risk in the last game because he needed a win to even the match.  You're right though, Fritz didn't exactly show a clear superiority, however blundering a mate is a human thing and can't just be thrown out.  It is still true that Kramnik was unable to win a game by any means, and even if the match situation lead to his loss in the last game, it can hardly be said that he played better than the computer.
Parent - - By Legendary (***) Date 2007-04-19 12:59
Winning against Rybka isn't that hard to do. You just need to have anticipation and the ability to understand Rybka evaluation. IMO Rybka play's like that average 2000 player, due to the fact that she doesn't play a very Dynamic game. Rybka's game is very drawish, all you have to do to win is form what I call the "Bow and Arrow" pawn formation against her and she'll play some ridiculous moves that will ensure your win.
Parent - By Uri Blass (*****) [il] Date 2007-04-19 13:36
This is simply not correct.
Somebody who plays like average 2000 player has no chance to beat a GM with pawn handicap.

Uri
Parent - - By gala.martin (**) Date 2007-04-17 21:12
Are you kidding? Try this. Play rybka with its own book; you are allowed to take back 3 times, but your clock will keep running while you do that. Play 10 blitz games like this, and then tell me how easy it is got a draw/win. If you are not a computer chess expert, you will score below 1.0.
Parent - - By Gaмßito (****) [cr] Date 2007-04-17 21:37
What would it happen if the next GM would play with those rules in a match against Rybka?  Without any doubt, a GM will have a high percentage to draw or win the match. And I do not talk about Blitz games. In Blitz games, we human are very weak against strong computers, still with take backs.

But what sense exists, in winning to a strong program in this way, with 3, 4, 5 or more take baks? 

Regards,
Gambitto.
Parent - By Uri Blass (*****) [il] Date 2007-04-18 06:06 Edited 2007-04-18 06:10
I disagree that there is no doubt that the GM is going to win or draw the match.

I expect the machine to win the match in case that the machine is strong enough because even with take back humans do not play perfect chess.

Human may get draws with the white pieces (I am not sure if it is possible to beat top humans with black in every game even without handicap) but I expect that they will have problem with the black pieces.

Uri
Parent - - By Banned for Life (Gold) Date 2007-04-17 22:20
1) Allow the GM to access any database (openings, games, tablebases),
2) Allow the GM to work on an one or more analysis boards
3) (Impractical) Have a team of GMs
Parent - - By lkaufman (*****) Date 2007-04-18 02:05
     Allowing the GM access to opening databases or game databases is almost useless for him if Rybka has only a three move opening book. Rybka will be playing unusual openings or will leave book early anyway. Allowing the GM access to endgame tablebases while the computer has turned them off is just bizarre - that's making the human more like a computer than the computer! Allowing the GM analysis boards is reasonable, but in my opinion for most really strong GMs this would be a very small advantage, maybe 10-20 Elo. They are so good at blindfold play and analysis that this should not really help -- in fact I bet many top players would not even use analysis boards if allowed to do so.
     As for a team of GMs, they would have to be of nearly equal strength to be stronger than the strongest one by himself, and even then I'm not sure whether the advantages of less oversights would outweigh the time lost due to discussion/debate.
     Of course we can always give the GM more time, in fact the only limiting factor here is that if we are to play two rounds a day (as we must to keep costs reasonable), too long a time limit might actually hurt play by being too tiring. If a GM accepts my offer but asks for more time (keeping to the 2-1 ratio and the 2 round a day format), I won't object.
     The only idea that has been mentioned that really would make a noticeable difference (in real-time play) is take-backs. Rather than an arbitrary time penalty, you could just specify some number of takebacks as the handicap, and the number could be increased until it made for a fair match. However the idea doesn't really appeal to me; it's rather like computer playing computer -- the human is using the computer's play to modify his own moves, unless you specify that the takeback must be done before the computer moves, in which case it is a fairly small handicap, not likely to change the result of any but a close match.
Parent - - By Banned for Life (Gold) Date 2007-04-18 03:10
Your response seems reasonable with the exception of your assertion that you can negate all theory with a 3 move book. Either you are planning to escape theory entirely by playing very strange and most likely suboptimal 3-move openings, or you believe that Rybka can wade its way through theory unassisted (certainly not true). I guess going for strange suboptimal openings might be no more of a dissadvantage than pawn odds, but with the GM having white, your book would have to be very broad to ensure that the GM didn't throw you out of book first. If the GM can a priori figure out the game state at the end of the third move, Rybka is predictable enough to do some serious preparation (which of course would all be stored in the database :-) ).
Parent - - By lkaufman (*****) Date 2007-04-18 04:28
     I don't say we'll negate all theory with a 3 move book, we'll just avoid main lines where lack of book knowledge might be fatal. For example we might choose an offbeat defense to the Spanish, like 3...Qf6 or 3..Bb4. Natural moves suffice to give White an edge, probably between 1/3 and 1/2 a pawn. There is no shortage of defenses to 1e4 and 1d4 which although inferior are not totally awful. The GM won't benefit from a database; there will be some games by players weaker than he most likely, and probably Rybka will soon deviate anyway. If the GM wants to throw us out of book first by rare moves, I think that would be foolish strategy, because he cannot prepare exactly what will happen since we will not use a commercial version of Rybka. We just have to accept that we will always be somewhat worse after three moves, I just believe that we will not stand as badly in the opening as we did in at least the six non-edge pawn handicap games. Both the handicap in that match and the one proposed for this one are in reality somewhere in the ballpark of a half pawn disadvantage - in the case of the pawn odds match Rybka got White and two of the eight pawns were low-value edge pawns; in the present match the opening is not quite as bad as this, I think, but the time situation is much worse, so it's a matter of opinion as to which terms are more favorable for Rybka. I personally feel that my risk of having to pay up is less under these terms than under the Ehlvest match terms, but I could be wrong.
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-04-19 03:28
I'm in agreement with Alan, Larry.  He and I are opening book specialists, of a sort.  An opening book of the sort you imagine might well be a handicap, just as you say.  But an opening book of the sort Alan has (and he fatuously claims I have) could potentially do you harm.  Mind you I am not saying this would happen every time, or even the majority of the time, but often enough to make a material difference.  The good news for you: I doubt any potential GM opponent will come armed with Alan's book.  (You had better hope not!) 
Parent - - By Vasik Rajlich (Silver) [hu] Date 2007-04-19 11:49
Ok, just for fun, let's try Larry's two examples:

I) 1. e4 e5 2. Nf3 Nc6 3. Bb5 Qf6
II) 1. e4 e5 2. Nf3 Nc6 3. Bb5 Bb4

Can you check your book and tell me if your analysis of these two lines has significant practical value?

Vas
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-04-19 12:31
Heh, well...that would entirely depend on your replies to subsequent moves, wouldn't it?

For Bb4 my book has a 41% result for black.  For Qf6 I have a 33% result.  Both lines produce well below-average draw percentages.  Both moves occur in roughly 1-in-1000 games.  (Pardon me for not revealing my Ns.)

I don't see anything necessarily fatal in these lines; there are countless others you could take with a similar profile.  If Larry's approach is to have the book take similar obscure lines I think he's strategically on the right track against a GM aided by a strong book--neutralizing theory to the greatest degree possible and forcing early calculation.

But Alan's approach would force you to do this in a particular line that he has analyzed so broadly and deeply that you'd be back on your heels from the git-go.  He would be your worst case scenario: somebody who doesn't play 1.e4 at all but rather repeatedly plays some other move or line that doesn't transpose particularly well and branches into a different theoretical universe where your three moves would not be adequate.
Parent - - By lkaufman (*****) Date 2007-04-19 14:36
     If you play unusual lines yourself as White in the first three moves, you are apt to take Rybka out of book early anyway, in which case the three move book limit will not prove to be much of a disadvantage, right? I think that your strategy is more appropriate against a fully booked Rybka.
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-04-19 15:57
What is an unusual line to you might be a deeply-analyzed line to someone else.  That is what Alan was alluding to--a prepared book that forces you into deeply-analyzed lines.  This isn't a question of having a database of historical games; it is all about having pre-analyzed positions in very deep permutations down a particular path in a non-commercial book.  This is Alan's forte and certainly the same technique could be used in variations other than his.  Anybody who can stay in book 10, 12, 15 moves longer than you, in well-analyzed positions, can potentially convert this into significant advantage on the board and the clock.
Parent - - By Uri Blass (*****) [il] Date 2007-04-19 16:32
there is memory limitation and the GM need to prepare books for hundreds of unusual lines in order to have a good practical chance
to use this strategy.

If the human starts with 1.h4 then he will not going to know if the computer has 1.h4 a6 or 1.h4 a5 or 1.h4 b6 or .... in the book.
If the human start by common line like 1.e4 then he is not going to know if the computer starts with 1.e5 or 1...c6 or 1...c5 or ...
Even if the human guess correctly that the computer has 1.e4 e5 in the book and decide to prepare for 1.e4 e5 2.a3 then he still will have problem to know what the computer has in the book after 1.e4 e5 2.a3

If you build a deterministic book for one game then the book needs to have reply for every possible move 1 of white(20 moves) reply for every possible non losing move of white in move 2(near 400 moves) and reply for part of the possible moves of white in move 3(if you prepare something like 1.h4 a6 2.a4 c5 there is no need to prepare third move because the probability of the human to guess it and prepare is very small).

I think that a book of 2000 moves may be enough.

Humans can prepare against 1.e4 e5 2.Nf3 Nc6 2.Bb5 Qf6 if they know that the first moves of this line are going to be played but when they do not know both the first move and the second move the probability that they will prepare to some rare line is very small because we do not talk about match when humans can bring big theory books with them.

Uri
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-04-19 17:18
If the GM can't use a book then this whole discussion is beside the point; I was talking strictly about somebody who could use an opening book during the match, which is what started this conversation I think.  If we're talking an unassisted 2600 GM then I think we know the outcome of this match already regardless of Rybka's book: imagine a bug hitting your windshield at 150 kph.

I don't disagree with your analysis, Uri.  Indeed there are seemingly infinite permutations and nobody can analyze them all beyond a certain point.  But it is quite possible to narrow the field of likely responses in potential out-of-book positions if you know the engine you are facing, so the exponentiality is not quite as extreme as you think.  This is how people on Playchess play 30, 35 (and sometimes more!) moves out of book in some often-played lines.  In an unusual line you could manage 10-15 book moves I think, and that offers a pretty fair advantage against a 3-move book.  There's no doubt of this statistically if you compare relative moves in book over a lot of games.
Parent - - By lkaufman (*****) Date 2007-04-20 02:26
     I see your point; if a GM spent countless hours preparing a special book on a slightly offbeat but not completely harmless opening, and had the right to use that book during a match, he would indeed get very good positions from this book against 3 move book Rybka. But it is not at all clear to me that he would get better positions than he would get simply by playing his normal mainline openings and letting Rybka play whatever inferior defenses she selects. Either way, the GM should get advantages in the 1/3 - 1/2 pawn range, I think. So what incentive would there be for the GM to put in all that extra work to prepare this special book under these circumstances?
Parent - - By Nelson Hernandez (Gold) [us] Date 2007-04-20 03:24
There's the rub.  You are willing to concede up to a half-pawn to the GM because you are confident you can win it back and then some.  The question is whether the "special book" could force you to yield even more than a half-pawn and thereby put Rybka's prospects in peril.  Answer: yes, it could from time to time.  But statistically the question is what the median result would be, and under that precise definition I seriously doubt even a half-pawn advantage could be attained over a span of games.  Not against Rybka powered by a quad.  This, I think, is where Alan and I part ways; he is much more optimistic that his book could bring you to grief.  (I'd be content if my book could just keep me out of trouble consistently.)
Parent - By Vasik Rajlich (Silver) [hu] Date 2007-04-21 11:43
Yes, exactly.

We all agree that that:

a) Under Larry's match conditions, Rybka should avoid theoretical lines.
b) The white player can play a normal theoretical variation, taking advantage of the fact that Rybka can't afford to get into a theoretical debate, and get some sort of nice advantage.

If the match is scheduled, I am sure that Larry will try to quantify these advantages, for example by running the Randomizer on the position after 1. e4 e5 2. Nf3 Nc6 3. Bb5 Qf6, etc. For now, 1/2-pawn seems like a pretty good estimate.

The question is if the white player could do even better than this by himself steering the game into unusual territory in the first three moves. Personally, I highly doubt it. Rybka is not that predictable on a 4-way machine. In the freestyle tournaments, we've played many games against unassisted Rybkas and you just don't get that much mileage out of knowing your opponent, even if it's a 1 CPU Rybka and you have the exact version yourself.

It's harder to quantify this second effect, but it should anyway also work in normal games.

Vas
Parent - - By Gaмßito (****) [cr] Date 2007-04-20 21:28
Larry,

I believe that a three move book is not enough to play against a very strong and prepared GM. I think that Rybka needs a bigger book. In games with odd pawns, are very important to play actively from the first movements. If after 3 good opening book moves Rybka plays passively, the GM can try to play very solid, exchange some pieces, and get a slightly better endgame position with a pawn more.

A very good prepared GM could attempt something like that in a match. Do not you believe?

Regards,
Gambitto.
Parent - - By lkaufman (*****) Date 2007-04-20 22:03
You seem to be confusing the three move book (and White) handicap with the pawn handicap, which are two separate topics. There was no particular limit on the book in the pawn odds games, although I think only one or two of the eight pawn odds games with Ehlvest stayed in book past move three anyway. In normal chess, Rybka will get inferior games as Black with a three move book, no doubt, but I expect her to draw most of them and to win some.
Parent - By Gaмßito (****) [cr] Date 2007-04-21 13:14
Of course, I thought that the 3 book moves, was related with the handicaped match (pawn odds games). 
 
Thanks for your explanation Larry.

Regards,
Gambitto.
Parent - By Uri Blass (*****) [il] Date 2007-04-18 06:43
I think that the right to allow the GM more time should be defined well.
There should be some maximal time control when you want to play 2 games per day.

When there is a game with increasement it is not clear

90+30 mean possible (1.5+n/120)*3/2 hours per game when n is the number of moves.

In theory in case that n is too big (  for example n=800) it is impossible to play 2 games per day but I know practically of no game that takes so many moves.

practically I think that games rarely take more than 200 moves so it is logical to have some limit that prevent 200 moves of both sides in more than 9 hours(it is possible to decide that games with 200 moves or more than it will be delayed or adjudicated) so you can decide that you allow every time control that force the human to play 200 moves in 6 hours when there is going to be a difference of 9 hours or less than it between games in the same day.

Uri
1 2 3 4 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill