Author: Robert Hyatt
Date: 21:10:14 07/01/05
Go up one level in this thread
On July 01, 2005 at 22:35:21, Robin Smith wrote: >On July 01, 2005 at 21:38:28, Robert Hyatt wrote: > >>On June 30, 2005 at 21:47:25, Robin Smith wrote: >> >>>On June 30, 2005 at 20:29:42, Robert Hyatt wrote: >>> >>>>On June 30, 2005 at 16:08:32, Andreas Guettinger wrote: >>>> >>>>>On June 30, 2005 at 11:34:18, Evgeny Shu wrote: >>>>> >>>>>>http://chessbase.com/newsdetail.asp?newsid=2485 >>>>> >>>>> >>>>>Now this surprises me a bit: >>>>> >>>>>"I wasn’t really concerned about that possibility. In any case it would be >>>>>impossible for me to tell, because Hydra plays a very different game to any >>>>>other computer that I ever saw. Even in these six games it actually played >>>>>differently to anything I saw in its own previous games, so it’s not easy to >>>>>judge. But no, I don’t have any suspicions about human intervention. That’s not >>>>>something that concerned me." >>>>> >>>>>A replayed the matches live on Hiarcs 9.6 and Fruit 2.1 on my 2 computers, and I >>>>>would say above 95% of Hydras moves were suggested by at least one of them. >>>>>Especially Fruit did very well in predicting Hydras moves. >>>>>Therefore the sentence "Hydra palys a very different game to any other computer >>>>>that I ever saw" leaves me a bit out in the cold. >>>>> >>>>>regrads >>>>>Andy >>>> >>>> >>>>It's a little hyperbole and a lot of exaggeration. :) I had crafty analyzing >>>>most of the games live on ICC and it as well as most other programs predicted >>>>Hydras moves _very_ accurately... >>> >>>Please define "_very_ accurately". 100% of the time? :-) Or are you running >>>Crafty (or pehaps a stable of engines) and noticing that the engine(s), at some >>>depth or another, show the same move as Hydra most of the time. How often did >>>Crafty come up with the same move as Hydra when given the exact same amount of >>>thinking time? I haven't tested this, but I'll bet it is less than 95%. >> >>In one game where I kept the log, Crafty got 36 of 37 moves right (it keeps up >>with this in the log file). Crafty was searching as moves were relayed on ICC, >>so I have no idea how the moves were relayed there with respect to real-time at >>the game site. >> >> >>> >>>And even if Crafty did predict Adams' moves (once out of book) perhaps as much >>>as 95% of the time, even that does not mean that Hydra didn't put much more >>>pressure on Adams than Crafty or other PC engines would have. At the highest >>>levels of chess it only takes a move or two per game to make a big difference. >>>One slip by the computer and the presure is off. More presure->"very different >>>game" (at least from a subjective human perspective such as Adams') even if all >>>the other moves would have been the same. >>> >> >> >>My point was that Hydra is most _certainly_ not some new level of computer chess >>as stated by Adams. I wouldn't argue against it being the best computer chess >>entity at the moment. But it is absolutely _not_ head and shoulders above >>others. The advantage I have is that I have a lot of experience with parallel >>and distributed search, and know the losses that a distributed search entails >>compared to a pure SMP approach. And even if they are currently reaching 200M >>nodes per second, which I somehow doubt given the FPGA numbers they have >>published in the past, that is not _that_ much faster than other readily >>available hardware. I've seen numbers well beyond 20M for Crafty on a quad >>dual-core opteron, for example. I've seen numbers more than double that on >>other machines I can't really mention at the moment. So they are not _that_ far >>beyond today's programs. Clearly Adam's comments are based on some other >>reality or understanding that is not based on factual analysis. >> >> >> >> >> >> >>>Then there is also the issue of opening books. Hydra leaves book faster than >>>most top programs, because the Hydra team believes Hydra handles being out of >>>book and finding good TN's better than other programs. Leaving book earlier is >>>already, all by itself, a radically different game, in spite of how many of >>>Hydra's subsequent moves the PC's might find. >> >>Many programs have done this. All the way back to the 1970's. It is not a new >>idea at all. Many use very selective books for such matches also. And it has >>its dangers if the human chooses to attack such a book. Too many lines that >>appear to win a pawn to a 15 ply search, but 25 ply searches would show that the >>pawn was poisoned... That is a _huge_ risk for those willing to take it on by >>playing to a book weakness... >> >> >> >> >>> >>>I am certain Adams has played many games against PC engines. I am certain Hydra >>>seemed subjectively, to Adams, as stronger and harder to handle than these PC >>>programs. This means that Adams statement "Hydra plays a very different game" >>>would, from Adams' perspective, be completely true; even though PC's can predict >>>most of Hydra's moves. >> >> >>I simply believe it is hyperbole. I know too many GM players that both watched >>the games, and have played thousands of games against computers, and they simply >>said "it played pretty good chess, very good tactically, less good >>strategically, and the opponent simply played the wrong style of chess to beat >>the machine." I had too much exposure to deep blue, and saw the exact same >>thing back in the deep thought and deep blue days. OK strategic chess, >>excellent tactical chess, no giant breakthrough at all... Just a big >>computational edge. In 1996 DB's edge was 100X faster than the micros of the >>day. Today Hydra's edge is not even 4x. 4x is significant, but not >>unbeatable... >> >> >> >> >> >> >>> >>>I think your claim above regarding Adams could be a little hyperbole and a lot >>>of exaggeration. :) >>> >>>-Robin >> >> >>Think what you want. If Hydra blows through the WCCC undefeated, your point >>might have a chance. I doubt it will... >Hi Bob, > >Hooray! A reply that I (mostly) am in agreement with. For example I agree that >Hydra is not some whole new level of computer chess. I do believe it is the >strongest chess entity out there, but not by leaps and bounds, and a good >anti-computer player would likely take it down. > >One point of disagreement however, I did _not_ say that Adam's statement is >factually correct; I said that from Adams' _subjective viewpoint_ there are >plenty of reasons _he_ could believe it is correct. Accusing him of "a little >hyperbole and a lot of exaggeration", an expression generally reserved for >people who are knowingly overstating their case for purposes of >self-aggrandizement, could easily not apply here. > OK on that point. Personally, I am a bit suspicious of the remark myself. Because there are simply way too many "experts" that know his statement is highly exaggerated. It is almost one of those "OK, I got myself killed here, so if I can convince everyone this thing is stronger than any other machine around, I will not look nearly as bad. And since not many will actually have the opportunity to play it, it is unlikely that most will realize that it is not as strong as it appeared in this match..." >For example, you point out that many programs going back to the 70's have left >book early. Yes, but has Adams played these programs? Did the programs Adams has >played leave book early against _him_? If not, then from Adams' perspective it >will seem to _him_ like a "very different game". > >-Robin I obviously can't and won't try to speak from the perspective of Adam's experience since I don't know what he has done "in the dark". But GM players play my program all the time, and sometimes they find it playing very shallow book lines due to my testing things, or because book learning has cut off the key variation of a common opening leaving it with no viable book move pretty early, etc. And none take this is a "remarkable new development."
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.