Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / POLL: What kind of human-computer matches do you prefer?
Poll What kind of human-computer matches do you prefer to see? (Closed)
1)Human Vs computer at normal Chess. 63 42%
2)Human Vs computer at draw odds games. 12 8%
3)Centaur Vs computer at normal Chess. 10 7%
4)Human Vs computer with Pawn handicap. 5 3%
5)Human Vs computer with Knight for Rook handicap. 6 4%
6)Human Vs computer with f-Pawn and 3 move odds. 2 1%
7)Human-Comp, normal Chess,no tbs and opening book 11 7%
8)Human-comp, normal Chess, 6 ply book for comp. 5 3%
9)Human-comp at normal Chess with time handicap. 23 15%
10)Human-comp, normal Chess, human with op. book. 12 8%
- - By George Tsavdaris (****) Date 2008-08-08 22:56
Which kind of matches between computers(Rybka for example) versus human do you prefer?

•1)Human Vs computer at normal Chess without any handicap.
•2)Human Vs computer at draw odds games(if humans draws it counts as a win).
•3)Centaur(human+computer) Vs computer at normal Chess.  (The computer assistant of human should be less stronger than that that he will play with)
•4)Human Vs computer with Pawn handicap(for example an 8 games match were computer will not have each time a-Pawn, then b-Pawn,...,etc.).
•5)Human Vs computer with Knight for Rook handicap(computer will not have a Rook while human will not have a Knight).
•6)Human Vs computer with f-Pawn and 3 move odds(computer will not have the f-Pawn and human as white always, will play 3 starting moves).
•7)Human Vs computer at normal Chess but computer not to have opening book and tablebases.
•8)Human Vs computer at normal Chess but computer to have only a ~6-ply book.
•9)Human Vs computer at normal Chess with time handicap(for example 1 min/move for human versus 20 seconds/move for computer, or 60+20" for human versus 10+2" for computer, etc)
•10)Human Vs computer at normal Chess with human been able to see and take advice from an opening book for his starting moves and computer not allowed to use tablebases.

Suffice to say that i explicitly prefer the 1) option which is normal Chess. :-) OK 2) and 9) are good also.....
Parent - - By Martin Thoresen (***) Date 2008-08-08 22:57
Where is the poll? :)
Parent - - By George Tsavdaris (****) Date 2008-08-08 23:03
The damn Rybkaforum.net poll system is crap. :-P
You have to first write the question and press post.
And THEN you can create the questions of the poll. That was the reason of delay....
Parent - By Vempele (Silver) Date 2008-08-09 08:19

> The damn Rybkaforum.net poll system is crap. :-p


As are the attachment and ropic tag systems, for the same reason.
Parent - - By dareapa (**) Date 2008-08-08 22:58
1
Parent - By likesforests (**) Date 2008-08-08 23:02 Edited 2008-08-08 23:06
We all know by now that at tournament time controls engines usually win. It would be interesting to see what happens at correspondence time controls, or with a small team of humans vs a computer. I guess my option's not up there.  :)

I don't like handicaps as much because humans can't win. If they score a point, "it's only because of the handicap".
Parent - - By Banned for Life (Gold) Date 2008-08-08 23:02
I would prefer to see even-up matches between Rybka 3+ on the fastest available hardware and 3200 Elo humans. :-D

Regards,
Alan
Parent - By dragon49 (****) Date 2008-08-09 01:12
I would like to see a match with a long series of games ( 12 or 24) between a GM and Rybka at normal chess with the following conditions.

1 - Rybka can play on a fast octal, but can not use opening books or tablebases.
2 - SLOW time controls - something like 40/12 20/6 SD/3 with 3 minute increments.
Parent - - By BB (****) Date 2008-08-09 03:52

>I would prefer to see even-up matches between Rybka 3+ on the fastest available hardware and 3200 Elo humans. :-D


But at giveaway chess. :-P
Parent - By wem511 (**) Date 2008-08-09 13:00
lol +1
Parent - By InspectorGadget (*****) Date 2008-08-11 11:35
I would prefer to see even-up matches between Rybka 3+ on the fastest available hardware and 3200 Elo humans.

Hahahaha!! That's not fair Alan.

I voted 7) and 9), I these 2 are closer to even-up matches. No opening book and no tablebases so that we can see how much can the computer go without Jeroen Noomen and Eugene Nalimov's assistance :)
Parent - By Roland Rösler (****) Date 2008-08-09 04:18
I would like to see matches between Rybka vs human (IM and GM) with no odds! So you have to do something it´s thrilling. My proposal: Take away the queen and two minor pieces for both sides. And then let us see ...
Parent - - By Arkansaw (***) Date 2008-08-09 05:19
I think 7 will possibly be the most balanced without resorting to material odds. The challenge for the human will then be as follows:

1. try to survive the opening
2. try to survive the middlegame
3. try to survive the endgame

PS: the point is that the chances for 1 to 3 are not so negligible now
Parent - - By lkaufman (*****) Date 2008-08-09 05:46
With no book at all we might just get repeat games. That's why we specified three move book for Ehlvest match. Also we don't want the human to be able to prepare the game in advance by seeing what moves Rybka plays on his own computer. Of course we can avoid this by making a very different special version of Rybka, but that has its own drawbacks. Regarding no tablebases, it's no big deal, but Rybka has no knowledge of many simple endings because tablebases are assumed, at least for 3-4 men.
Parent - - By Vinvin (***) Date 2008-08-09 10:48 Edited 2008-08-09 10:51

> With no book at all we might just get repeat games.


Simply use a randomizer (0.10 to 0.20 pawns) for 5 first moves.
May be you can make some statistic against top5 engines (with and without book) to know how this would weaken Rybka 3.
Parent - By lkaufman (*****) Date 2008-08-09 13:14
This is one solution. It's not too different from using a very short opening book (as we did vs. Ehlvest) that is also designed to produce variety, except that the short book also tries to avoid mainline theory even at the cost of playing somewhat inferior moves. It would be quite a waste of time and money to arrange a match with a top player only to see Rybka fall into some heavily analyzed line and lose on move 15 while the opponent is just repeating standard opening theory.
Parent - By InspectorGadget (*****) Date 2008-08-11 11:38
With no book at all we might just get repeat games

That's fine, so that human can do what suits him most :) :D
Parent - - By Axel Caro (**) Date 2008-08-09 07:28
Good poll (and overdue...)!

>Suffice to say that i explicitly prefer the 1) option which is normal Chess. :-) OK 2) and 9) are good also.....


That's exactly my choice: on all others alternatives it's true, what "likesforests" said:

> I don't like handicaps as much because humans can't win. If they score a point, "it's only because of the handicap".


The only handicap I would accept, is time handicap. PocketFritz 3 won a tournament in italy - on a standard PocketPC which should be 100-200 times slower than a quad or octal. Nonetheless he succeeded...

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=87848;hl=Pocket%20Fritz :

>*Pocket Fritz 3 (hiarcs 12 engine) win the Mercosur cup 2008 with 8/10
>*Performance of 2700 elo whit GM norm
>* No loss


Reducing speed/time reduces the overall strength without affecting the basic nature of chess. I you are a box champion in light weight you will not fight against an opponent in super heavy weight...
Parent - - By Vinvin (***) Date 2008-08-10 09:48

> PocketFritz 3 won a tournament in italy


It's in Argentina ;)
Parent - By Vempele (Silver) Date 2008-08-10 09:54
Didn't you hear? They've renamed some countries - Hiarcs is now German instead of British, for example.
Parent - By Axel Caro (**) Date 2008-08-10 10:14
Things change... Men change.. Country change...  But your are definitely right: it's Argentina (and still is...) ;-)

http://www.diariodemocracia.com/diario/articulo.php?idNoticia=10818
Parent - - By BB (****) Date 2008-08-11 03:45

> > PocketFritz 3 won a tournament in italy
>It's in Argentina ;-)


I wonder how many Forum readers are aware of the large Italian-descended population in Argentina... (about half of the population, I think).
Parent - By Uly (Gold) Date 2008-08-11 06:38
Yes, even the accent is very similar :)
Parent - - By oudheusa (*****) Date 2008-08-09 08:51
Alas my preferred option is not there; normal chess but human gets to use  opening book and light engine to support analysis and prevent miscalculations because of fatigue etc. Also if engine uses tablebase human should be able to use as well; ie all hard coded knowledge (book and tb) to be equally available to both parties.

In this way we will get a match in which we can establish if human strategic insight is up to the strongest engine. In other cases, even the best GMs suffer from loss of concentration/fatigue in any match and this should not be a factor.
Parent - - By George Tsavdaris (****) Date 2008-08-09 09:11

>Also if engine uses tablebase human should be able to use as well; ie all hard coded knowledge (book and tb) to be equally available to both >parties.


In that case you should also forbid the engine to use its evaluation code, since all are hard coded knowledge.
You can't be selective to what you will forbid.

That's how computer works!
People have their memory for remembering the openings.
Computers have their own memory. This is opening book for the openings and tablebases for the endgame.
If humans can't accept that computer's memory is much better, then humans should not play these kind of matches if this leads to asking unreasonable things to be forbidden for their opponents.

With the same logic, computers should also forbid humans to use their pattern recognition since computers don't have that kind of ability.

I didn't see these kind of irrational requests in 90's for example where humans could still beat computers.
Now they are incapable of doing this, they try to degrade computers strengths with the most lame excuses.....
Parent - - By oudheusa (*****) Date 2008-08-09 10:23
To me the question is not at what piece or time handicap the strongest human player can still beat or draw the computer, but if the strongest GMs are still superior in understanding the game, ie in strategy. My proposed setup for a match serves exactly this purpose.
Parent - By Kappatoo (*****) Date 2008-08-09 10:30
A big part of 'understanding the game' is due to accurate calculation. The question ist thus: Which engine help to avoid blunders would you consider appropriate? It would have to be an engine which is clearly weaker than the human, I suppose. And than I am not sure how much it would help. Most of the time, strong grandmasters lose due to very subtle 'blunders' - they are just being outplayed by the engine.
Parent - - By George Tsavdaris (****) Date 2008-08-09 10:39

>To me the question is not at what piece or time handicap the strongest human player can still beat or draw the computer, but if the strongest GMs >are still superior in understanding the game, ie in strategy.


If this is the question that you are interested to, the answer is easy: Humans.
Top humans at Chess have far superior strategy than computers.

Bad thing for humans is that long term strategy doesn't mean too much in Chess in non-correspondence time controls.
Short tactics are far far more important.

While i don't really understand what do you mean by "understanding the game." What is the definition of it?
Parent - - By Uri Blass (*****) Date 2008-08-09 10:58
It is not so easy because it is not clear how do you define superior strategy.

Uri
Parent - By George Tsavdaris (****) Date 2008-08-09 11:53
Well i define strategy as the long term strategy and by the last one i mean planning about moves that their purpose will become obvious and have a consequence after many many moves(20-40) in the board.
For example an example of strategy is "i play 22.f4 to close the Kingside so i will try to have play in the Queenside where i'm a bit better and after some non-concrete yet variations it's obvious i will be able to create a Passed Pawn.".
Computers can't do that(since they have to calculate the in-between variations so they are too slow for that). Humans can very efficiently(obviously not efficiently enough to succeed most times).

The problem for humans is that the in-between moves are more critical than the plan's efficiency.
And the short-strategy, which is tactics is far more important.
So even if you have a very good plan that computer have no idea if you make small(or bigger) mistakes(tactical mistakes) in the execution of the plan, then the plan is of no value.

Of course as you say it's not obvious why by saying strategy we should mean long term strategy and not short strategy(=tactics) but that's the common use of the word strategy i think.
Parent - By wem511 (**) Date 2008-08-09 13:03
+1
Parent - - By tralala (*) Date 2008-08-09 15:10
It's there it is option 3 (which I prefer as well). The specifics which engine on which hardware the human may use has to be worked out (and experimented with), but it's clear it should be only a very light support. Let's say at least 400 Elopoints weaker than the opponent engine.
Regarding the use of an opening book for the human, I don't think that is such a big help to them, I think they have memorized their repertoire. And for TBs the arrangements are negligible.
Parent - By oudheusa (*****) Date 2008-08-09 16:06
Fully agree; a tactical engine like Fritz 6 should serve the purpose well.
Parent - By lkaufman (*****) Date 2008-08-11 04:10
One problem here is that 400 Elo points worse than Rybka 3 is still equal to or better than Kasparov/Kramnik/Anand/Carlsen. The supporting engine must be some significant number of elo points worse than the HUMAN player, not the opposing engine, or else it is just an engine vs. engine game with human centaur assistance for the weaker engine.
Parent - By exigentsky (***) Date 2008-08-09 10:52 Edited 2008-08-09 11:00
I'd like to see Rybka play against a GM at long time controls but with the GM having access to an analysis board next to his playing board. He can move the pieces however he likes and we get to watch what he's moving. This would help off-set Rybka's calculation ability and put the game more into the realm of chess understanding. Draw odds would still be necessary since I doubt the analysis board helps THAT much. Moreover, I think it's only fair that the GM gets access to the opening book and is allowed to use the perfect play of tablebases when he reaches such a state of reduced material. Why? These are not part of the engine, but they are stored information simply accessed during the game. It would be like me reading an opening book or looking at my notes while playing OTB. It's true that GMs have memorized some openings etc. but it is not the same as having the results from millions of games carefully selected. I think this is an interesting idea which hasn't been done yet. Moreover, it's still normal chess (or maybe closer to correspondence chess)... for the most part.
Parent - - By tralala (*) Date 2008-08-09 12:18
Strange that so many people still prefer option 1. Spectating a wipe-out gives me no satisfaction although I guess for some people it might feel good to have bought an engine for about 60 Euros which defeats the human Champion 5-1 in a 6 game match.

I like the centaur option, other handicap options are also interesting but I find it most annoying when the human plays strategically very well and keeps the balance or even obtains a slightly favorable position just to see him make a couple of innaccurate moves and the game is over. This would be reduced if the human could use a slow notebook with an older engine, just for blunder check (or if he missed something in a calculated line).
Parent - - By Mark (****) Date 2008-08-09 14:47
I like option 1 because there are still quite a few people (primarily on the other forums) that think that humans are as good as computers.
Parent - - By tralala (*) Date 2008-08-09 15:16
But I find it offending to arrange more slaughter matches (like the Adams-Hydra-Match) to prove the point. Those who doubt that the comps are stronger will always find excuses why in this particular match the human lost.

Btw. I am one of those who think that strictly playing for a draw with good knowledge how engines play and how to keep the game calm will allow the human to achieve quite a number of draws (say four out of five) however I don't think it will be valid for Rybka 3.
Parent - By Mark (****) Date 2008-08-09 16:30
Well, 4 draws and a loss out of 5 games would be 40% which is a rating difference of only 72, which would put the computer programs at under 2900.  I'd like to see the GMs try for this with Rybka 3!
Parent - By bluemax (**) Date 2008-08-09 17:52
Yes, until theres a general agreement that computers are better. Offering handicaps only complicates the debate, both sides (both pro computer and pro human), blaming the irregular conditions for the results.
Parent - - By Schiffermueller (*) Date 2008-08-10 22:29
I like all handicaps where the engine can show all its abilities. So time handicap is not interesting. Centaur option I find also interesting but only under following conditions:
- The human should have proven centaur abilities.
- The same engine is used by the human but with lower speed.

Centaur - human matches I find interesting to answer the question how we can estimate the ELO of a centaur from the ELO of the human and the ELO of the used engine.
Parent - By Schiffermueller (*) Date 2008-08-10 22:32
I mean centaur - computer matches.
Parent - By dareapa (**) Date 2008-08-11 06:50
How about a no handicap match with R1.2f sp, ...:)
Up Topic Rybka Support & Discussion / Rybka Discussion / POLL: What kind of human-computer matches do you prefer?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill