Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Knight Odds Match with FM
1 2 3 Previous Next  
Poll What will the match score be (remember draws are wins for Rybka)? (Closed)
Meyer by 4-0 8 10%
Meyer by 3-1 24 29%
Tie Match 2-2 28 33%
Rybka by 3-1 17 20%
Rybka by 4-0 7 8%
- - By lkaufman (*****) Date 2008-06-02 01:20
     On Sunday, June 8, we will hold our next official handicap match. For a change of pace, instead of giving a pawn to a Grandmaster, Rybka will attempt to give knight odds to a FIDE Master in a serious four game match. Rybka has done this before in blitz/quick chess, but it has never been attempted at a serious time limit. The time limit will be 30 minutes plus 30" per move increment, roughly equivalent to the 45'+10" limit in the pawn handicap matches with GMs Ehlvest and Dzindzichashvili. In view of the huge handicap and the fact that the human can almost always draw at will by repeating moves, Rybka gets draw odds in each game, meaning draws count as wins for Rybka. Rybka always takes White and removes the b1 knight, as was customary in knight odds play in the 19th century.
     The opponent is FM John Meyer, an active amateur player with a long distinguished record. He was U.S. Junior Champion back in the '60s and was U.S. Open Co-Champion in the early '80s. He was probably of IM strength around that time. Now at age 63 his play has declined somewhat, but he is still active enough for his current FIDE rating of 2284 to be a fairly accurate estimate of his current strength, especially in view of a current string of 9 consecutive victories in league play (which is not FIDE rated). John is playing this match to test the proposition that a chessmaster should be able to win consistently at knight odds against any opponent no matter how strong, with an hour (for 60 moves) on his clock. He will get a nominal prize of $100 for a match victory and $50 for a drawn match, but I know him well enough to say that he would do his best to win every game even with nothing on the line. John feels that he should win every game, but he is aware that other master level players have failed to win matches at much faster time limits and so will not underestimate the challenge.
     Rybka will be our latest version as of match time, running on my 3 GHz Octal, with a small knight odds opening book. Play will start at noon Eastern Daylight Time, with coverage expected to be on Chess Planet. Details of coverage and schedule to be posted later.
Parent - - By FWCC (***) [us] Date 2008-06-02 01:33
Very interesting Larry.It seems to me that knight odds even with the latest Rybka may be too much to give an opponent of master level(I have changed my view on this as at first I thought that it might be possible) but we will see.In light that Rybka's strength is increasing almost daily and that she just won Leiden this should be an interesting match.Taking nothing away from FM Meyer is it possible to get an even stronger player to take on this challenge(though FM level is of course adequate.)
Parent - - By lkaufman (*****) Date 2008-06-02 02:42
Well there's no point in looking for a stronger opponent if Meyer can win the knight odds match. If he fails then we might look for an IM. My son Ray would be one possibility, but there are other factors here than just chess strength. Since we can't justify a significant prize fund for a knight odds match, we need someone who will be strongly motivated despite the nominal prizes, and it's hard to find an IM with the right temperament and willingness to risk public embarrassment should he lose. I do have one in mind if Meyer loses the match, though I don't expect him to do so. Another point about Meyer is that he doesn't keep up with current opening theory, which tends to hurt his rating but would not impact a knight odds match obviously.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-06-02 04:15
I know that similar questions have been asked before, but you acquire new data and knowledge on this all the time: would you currently say that Anand, if given knight odds, should be able to beat a very strong and tricky player who is jockeying 32-piece tablebases?
Parent - - By lkaufman (*****) Date 2008-06-02 04:24
I would say yes, Vas would say no. We will both have a better idea after this match.
Parent - - By Vasik Rajlich (Silver) [hu] Date 2008-06-02 21:56
I don't remember what I said before. It seems that this is very unclear - it could be yes, it could be no.

Vas
Parent - - By Kreuzfahrtschiff (***) [de] Date 2008-06-03 00:33
Anand would always win that, without a doubt
Parent - - By Roland Rösler (****) [de] Date 2008-06-03 02:19
Anand can´t even win a rapid chess tournament against Chucky :-).
Parent - By BB (****) [au] Date 2008-06-03 02:34

>Anand can´t even win a rapid chess tournament against Chucky :-).


Yeah, but that was a book loss :) : 1.d4 Nf6 2.c4 e6 3.Nc3 Bb4 4.e3 0-0 5.Bd3 d5 6.Nf3 c5 7.0-0 dxc4 8.Bxc4 Nbd7 9.Qe2 cxd4 10.exd4 b6 11.d5 Nc5 12.Rd1 Qe8 13.Nb5 exd5 14.Nc7 Qe4?? {Strangely, this move was played in Conquest,S (2490)-Zaichik,G (2510), Tbilisi 1988, where White played 15.Nxd5 Qxe2 16.Nxf6+ gxf6 17.Bxe2 Ba6 18.Bxa6 Nxa6 and drew after 42 moves. Ivanchuk had a refutation:} 15.Rd4 Qg6 16.Nh4 1-0.
Parent - By Roland Rösler (****) [de] Date 2008-06-03 02:34
Sorry, this wasn´t fair to Anand. :-(
Parent - - By elemeNt [de] Date 2008-06-07 16:47
Question: Whats your Elo or DWZ Rating ? Sometimes I think you be a GM with 1500 Elo?!
Parent - By Roland Rösler (****) [de] Date 2008-06-10 02:06
F: Spielst Du auch oder red´s Du nur?
A: Na, ich red nur drüber!

PS: Geklaut aus Vorspann zu "Waldis EM-Club". Trifft´s genau für mich :-)
Parent - By turbojuice1122 (Gold) [us] Date 2008-06-02 01:44
This should be interesting; I'm guessing that Rybka wins one game outright, draws one (thus being awarded a win), and loses the other two, thus giving a drawn match.
Parent - - By Permanent Brain (*****) Date 2008-06-02 01:53 Edited 2008-06-02 01:55
Rybka with 8 CPU cores against a 2284 seems brutal, even if he gets knight odds. If my calculation is (roughly) ok, then 8 cores are ~5 times as fast as one core, providing ~2(+) plies of depth more in the same time, for Rybka.

I understand that such matches should be run using decent hardware, but I think top hardware is only required for "big" matches against the 2700+ category... what about this idea:

Opp.Elo    # cores
------------------
<2300         1
2300...2499   2
2500...2699   4
2700+         8
Parent - - By lkaufman (*****) Date 2008-06-02 02:47
Well, you're rather missing the point from Meyer's perspective. He feels that knight odds is enough to win a match against even perfect play, he wants to play the strongest entity on earth. This is really the question we are testing here, although of course we can only prove him wrong in this match, since if he wins we can always claim that a stronger program/hardware might have won. If we were playing him at pawn and move odds like the last Dzindzi games then perhaps 1 core might make a reasonable match. But we are basically interested in showing what Rybka can do with the best hardware an ordinary person can buy, and an octal though expensive certainly qualifies.
Parent - - By Permanent Brain (*****) Date 2008-06-02 04:03 Edited 2008-06-02 04:11
The problem is, many people having single core, duals or quads will wonder how big the share of the (better) hardware was, for a particular performance. In other words, I guess that many Rybka fans want the software to do the slaughtering and the hardware to be the platform, only. With an 8 core computer in this match, I don't think it will make a noticeable difference if it's Rybka 2.99 or Crafty against this FM, but I'm of course not sure...

By that, I do not mean to underestimate Meyer, who has a very respectable Elo, but maybe I overestimate 8 CPU-Rybka? We will see. I voted 3-1 for Rybka :-)
Parent - By lkaufman (*****) Date 2008-06-02 04:15
Well, we know roughly that 8 cores runs about 5 times as fast as 1, which is a much smaller rating difference than Rybka 3 vs. Crafty. Also depth of search should not be as important as eval and avoidance of simplification when behind, so I think that the number of cores is of less significance in this match than in an even match with Anand let's say. When you are lost, searching deeper doesn't really help much, there's nothing good to find.
Parent - - By Gaмßito (****) [cr] Date 2008-06-02 04:05
I think also that a Knight plus is too much advantage, but not as much to really say that it is possible to win every game against the strongest entity on earth. At least not for a 2200-2300 player.
 
I think that the choose of the openings are quite important at Knight odds games. I do not like safe openings for Rybka at Knight odds, and I think that sharpest openings offer much better possibilities. 
 
If the FM plays quite safe and it is enough prudent in his play, that it is the correct way to play - trading pieces and avoiding all kind of risks- the match can be very hard for Rybka. We will see if Rybka is even capable of get a good result in this difficult challenge.

Thanks for this upcoming challenge Larry. It will be a very interesting match, indeed.

Regards,
Gaмßito.
Parent - By lkaufman (*****) Date 2008-06-02 04:17
Of course sharp openings would be nice, but there are no sharp openings if Black does not cooperate.
Parent - - By Sesse (****) [us] Date 2008-06-02 02:24
Knight odds is pretty wild. I'd guess for one blunder giving Rybka a win, and the other three a win for the FM.

/* Steinar */
Parent - By lkaufman (*****) Date 2008-06-02 02:51
I won't argue with your prediction except to say that in my experience blunders at knight odds usually only cost the win, not the full point. Even if you miss some tactic that wins a piece by a fork or trap, usually you can get a pawn for it and remain a pawn ahead which should suffice to draw if we're approaching the endgame. If Rybka wins a game I think it will be because John avoids any drawing lines even after an error, due to the match terms.
Parent - By Werewolf (*****) [gb] Date 2008-06-02 06:39
I am organising a very similar match, rememeber. Just waiting for the product!

My testing so far seems to suggest that getting a draw is pretty easy, beating the machine is hard.
Parent - By Graham Banks (****) [nz] Date 2008-06-02 09:02
I'll go with 3 wins to the FM plus a draw.   :-)
Parent - - By Vasik Rajlich (Silver) [hu] Date 2008-06-02 22:00
Just for my curiosity - how much Elo is a knight worth in engine vs engine play?

Meyer is spotting Rybka around 800 Elo, and draw odds is another 200 Elo.

Anyway, it seems to me that a human will react differently enough to these conditions that this type of Elo math may have limited value.

Vas
Parent - - By lkaufman (*****) Date 2008-06-02 23:00
My tests showed that Rybka e13 (roughly halfway between 2.3.2a and now) on my quad lost by 136 Elo giving knight odds in forty quick games (10'+5" and 4'+2") to Crafty 19.0 and Fritz 5.32, which would average around 2500 on CEGT rapid list (Crafty 19.0 isn't on there, I'm averaging nearby versions). Rybka 2.32a on Quad is around 3100 on their list, so this gives a knight odds value of about 600 + 136 + 35 (guess as to improvement from 2.3.2a to e13) for a total of 771. But the Meyer match will be at a much slower pace, so on an engine vs. engine basis I would imagine 900 Elo is a good guess at that time limit (my tests do confirm that even for engines, slower time limits make it harder to give handicaps). But at least one of those old engines, perhaps both, had no (or little) concept of trading off when ahead, and of course no engine has the concept of avoiding unclear variations when facing a much stronger opponent or when well ahead. Also a randomizer test of Rybka vs. Rybka at knight odds at only 8 ply gave 759 Elo, which would surely be much more at a serious time limit.  So even with a thousand points to overcome (including the draw odds) I would rate him the favorite.
Parent - - By turbojuice1122 (Gold) [us] Date 2008-06-03 01:56
So if a knight is 1000 elo, then presumably two knights would be another 1000 elo?  But wouldn't a 1000-level player not be able to even come close to beating Rybka at queen odds?
Parent - - By Sesse (****) [us] Date 2008-06-03 02:42
I'm about 1000. I've played Rybka at queen odds a few times, without much of a chance.

/* Steinar */
Parent - - By InspectorGadget (****) [za] Date 2008-06-03 07:30
"I'm about 1000. I've played Rybka at queen odds a few times, without much of a chance." :)

Sorry about that, why don't you trade down all the way?
Parent - - By lkaufman (*****) Date 2008-06-03 14:36
I'll just note that Rybka knows to try very hard to avoid even trades when she is far behind in material, so it's not quite as easy as it sounds. It comes down to whether he can trade pieces faster than he loses them to various tactical tricks.
Parent - By Felix Kling (Gold) [de] Date 2008-06-03 16:29
Btw., I guess even I would have reasonable chances to win against someone at around 1000 Elo who is a queen up. they blunder their queens relatively often ...  and don't forget that they overlook stalemate tricks and so on.
Parent - - By lkaufman (*****) Date 2008-06-03 03:19
You are right. The problem is that the Elo equivalent of a given handicap degrades as you go down the scale. A knight seems to be worth around a thousand points when the "weak" player is around IM level, but it drops as you go down. For example, I'm about 2400 and I've played tons of knight odds games with students, and I would put the break-even point (for untimed but reasonably quick games) with me at around 1800, so maybe a 600 value at this level. An 1800 can probably give knight odds to a 1400, a 1400 to an 1100, an 1100 to a 900, etc. This is pretty obviously the way it must work, because the weaker the players are, the more likely the weaker one is to blunder a piece or more. When you get down to the level of the average 8 year old player, knight odds is just a slight edge, maybe 50 points or so.
Parent - By turbojuice1122 (Gold) [us] Date 2008-06-03 03:30
Yes, you're right--this seems easy to see now that you state it that way; I knew that there was a breakdown of some sort, though I was thinking in the wrong direction--this makes perfect sense.
Parent - - By Vasik Rajlich (Silver) [hu] Date 2008-06-04 12:11

> My tests showed that Rybka e13 (roughly halfway between 2.3.2a and now) on my quad lost by 136 Elo giving knight odds in forty quick games (10'+5" and 4'+2") to Crafty 19.0 and Fritz 5.32, which would average around 2500 on CEGT rapid list (Crafty 19.0 isn't on there, I'm averaging nearby versions). Rybka 2.32a on Quad is around 3100 on their list, so this gives a knight odds value of about 600 + 136 + 35 (guess as to improvement from 2.3.2a to e13) for a total of 771. But the Meyer match will be at a much slower pace, so on an engine vs. engine basis I would imagine 900 Elo is a good guess at that time limit (my tests do confirm that even for engines, slower time limits make it harder to give handicaps).


Ok, so we could easily find an engine which would clearly lose to Rybka in this match at this time control and which would easily beat Meyer in normal chess at the same time control. In other words, we are truly measuring the human's ability to adapt to the situation.

Vas
Parent - - By sarciness (***) [gb] Date 2008-06-04 17:16

> Ok, so we could easily find an engine which would clearly lose to Rybka in this match at this time control and which would easily beat Meyer in normal chess at the same time control.


Really? I'd be surprised. What makes you say so?
Parent - By Vasik Rajlich (Silver) [hu] Date 2008-06-05 08:46
From Larry's comments. It sounds like Crafty with a 4:1 or so artificial slowdown would fit the bill.

Vas
Parent - By Kreuzfahrtschiff (***) [de] Date 2008-06-03 00:37
a knight is 1000 elo difference then (with good players)

with weak players like 1200 or so they need at least a rook for 1000 diff
Parent - By sarciness (***) [gb] Date 2008-06-02 23:31
I'm going out on a limb and saying 3/5-0.5, so 3-1 at draw odds. We'll see :)
Parent - - By gala.martin (**) Date 2008-06-05 21:58
I voted 3-1 for meyer, but the result is quite unpredictable. for knight odds, there is basically no opening theory. therefore preparation pays a lot in this match. If Meyer prepares some opening lines, rybka will have no chances. Conversely, if Meyer is completely unprepared while rybka has some tricky opening book, things can be quite different. I guess that the result of most of the games can be told at move 15 or so.
Parent - - By lkaufman (*****) Date 2008-06-06 00:00
I don't believe a "tricky book" is possible at knight odds. No matter what first move White plays, Black has multiple options to play safely. If Rybka is to score it will be by resisting simplification, slightly improving her position by her vastly better search, and posing problems for Black. A typical successful game for Rybka might see Black shed a pawn to trade queens, shed another one to trade some more pieces, and then only draw the endgame. It's hard to force trades without making concessions against Rybka, the question is how big these concessions will be.
Parent - - By irvstein1 (***) [us] Date 2008-06-06 01:50
i will be really really amazed if rybka is able to win at knight odds . i dont know this player but for me knight odds would be a slap in the face to lose that to anybody or anything (computer).my money is on the human in this case . im seeing what you are saying about how you must make a concessions but remember rybka must make some concessions to because he is behind already . i think if you are serious about making odds games human vs computer then draws should not count then you will have true fighting chess and find out who is really stronger . i myself have draw many game vs rybka but beating rybka straight up has been a serious serious hard challenge .if all i had to do was draw rybka at knight odds i would have a 50/50 chance on your machine i can beat it on mine ,,,, i know mine is not fast only dual core 2.5 ghz im only running 32 bit mp version ... but another observation i have made and you need to considder . while your machine is much faster than mine and im sure you will get a deeper search ,,, how many ply will you actually get and what is the % of time that rybka will chose a diff move with maybe 2 or 3 ply deeper and how much better move will it be ???? unless rybka 3 is some demon in a box and has lightning shooting out of it ,,,, im going with the human if he dont get nervous and make a silly blunder .
Parent - - By irvstein1 (***) [us] Date 2008-06-06 02:02
i went up and read where you are going to give some one that kind of time and knight odds ,,,, unbelieveable ,,,,i really like rybka and think its a great program but i think you are going too far . even up match great opn book good size hash tables , egtb ok rybka is a hard hard match and deserving of a 3000 ++ rating but - the good open bk and - the knight  you are putting rybka in a hole it can not crawl out of . what do you think annand would do to rybka under these conditions ?????????? dont ever try this vs a strong player and a good anti-computer player , ill be rooting for rybka but im sure rybka will lose .
Parent - - By lkaufman (*****) Date 2008-06-06 03:50
I'm not insane, we're not offering knight odds to Anand! John Meyer is certainly not a "good anti-computer" player, just a good normal chess player with vast experience and a good general understanding of chess. "Strong player" has no meaning; I'm sure that most chessplayers would consider an FM like John Meyer to be a "strong player", but to Anand he is a patzer. Reminds me of the time I talked to Hikaru Nakamura about our matches against GMs Ehlvest and Benjamin, and he asked me "But have you played Rybka against a strong player?". "Strong" is only meaningful relative to some standard.
Parent - By Roland Rösler (****) [de] Date 2008-06-06 04:48
Reminds me of the time I talked to Hikaru Nakamura about our matches against GMs Ehlvest and Benjamin, and he asked me "But have you played Rybka against a strong player?".

That´s best! Here you can see, that our childs are grown. No clue about history. What shall´s!
Parent - By BB (****) [au] Date 2008-06-06 05:03

>and he asked me "But have you played Rybka against a strong player?".


The right repartee [though maybe not, with the odds] would probably have been: "Other than self-play games?" :-P
Parent - By lkaufman (*****) Date 2008-06-06 03:44
Regarding "draws don't count", I guess you mean like the first Karpov-Kasparov match, where 6 wins were needed. After 48 games the match was called off when no one had the 6 wins. This didn't exactly lead to "fighting chess"! As for your results at knight odds (and your statement about getting many draws without handicap), this doesn't tell us anything without an approximate rating (FIDE, National, ICC, Playchess, whatever you have). Maybe you are 1500, maybe you are 2500; without this information your results have no meaning except to you. But if you have obtained many draws against Rybka without handicap (other than by tricks like total blockade or book draws) you are surely of master strength and should indeed have excellent chances to win a match like this one. As for extra ply etc. on my hardware, all that really matters is that each doubling of speed is worth about 75 Elo in self play, 60 against other engines, maybe 40 against humans. So it's worth at least a hundred Elo in this match to have my octal 64 bit instead of your machine, and at least 75 for the latest program, and perhaps another 50 for the book. As for the human in question, I don't expect "silly blunders" but I do expect lots of small mistakes that add up. Still, Rybka is the underdog in this match, you are right about that.
Parent - - By irvstein1 (***) [us] Date 2008-06-06 03:45
you know i had to check my memory banks hmmm 1985 maybe im mistaken but i remember you larry kaufman it was in chess life and it was a endorsement for mephisto amsterdam and i think i remember you was saying something like it was the first computer that you could not give knight odds to .....and still beat it ??? im sure you will remember better than me .and think about it ,,,, im sure your rating then was very high i think it was over 2500 + and if you will check the current rating of mephisto amsterdam you will see it was no where near your strength ,,,, but a knight is a lot and i imagine spotting mephisto amsterdam for me would be a loss of a match ,,, i would not be able to say i could win .any way  im glad you have so much believe in rybka 3 ,, im sure ill buy it as i have done with all the rybka versions . i guess what i would love to see is rybka without hadicap but on a much weaker machine like a dual core 2.5 ghz good open book 1 gb hash tb 5 man vs a high rated gm say at least 2700 or so  some one like kaidanov ect because there are alot of people who love computer chess like me but dont own a demon of a computer 8 core ,,, and want to know if what we are useing can beat a gm ????? not a gm thats long past his prime ,, but a solid strong gm .
Parent - - By lkaufman (*****) Date 2008-06-06 03:58
Yes, I did say that in 1986 I believe. Mephisto Amsterdam was around 2150 USCF level then (never mind what today's lists show, the ratings have been lowered by hundreds of points over two decades to make the ratings of the latest programs reasonable), and so of course at 2500 I could not give it knight odds; a knight is far more than 350 Elo in value. As for the match you request, Joel Benjamin is rated only a little below Kaidanov and is not much older, and he scored only 2 draws in eight games against Rybka all with the White pieces for Joel. True, Rybka was on a quad (I didn't have the octal then), but Rybka 3 on a duo should be close in strength to that version on a quad. So basically the match you request has already been played, except Rybka gave draw and White odds!
Parent - - By sharpnova (***) [us] Date 2008-06-06 07:37
Have you tried any games with Rybka giving you knight odds? If so, how did you fair?
Parent - - By lkaufman (*****) Date 2008-06-06 14:15
Yes, I reported on this about three months ago. I played several games (six I think) at a timit limit around 20' (I forget exactly) and came out one game ahead. But this was when I only had a quad, and now Rybka is stronger too. 
Parent - - By NATIONAL12 (Gold) [gb] Date 2008-06-06 20:50
just out of interest what is your fritz benchmark,and average KNs with 2.3.2a,from opening position.i am getting 639KNs at D20,i dont want to bugger around with it during match with Hiarcs as i dont know what i am doing.
regards Paul
Up Topic Rybka Support & Discussion / Rybka Discussion / Knight Odds Match with FM
1 2 3 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill