Talk:Hydra (chess)

From Wikipedia, the free encyclopedia

Contents 1 Non copyvio 2 Search techniques 3 Unfair 4 Top player = Hydra? 5 The handle Zor_Champ

[edit] Non copyvio

The following text was found common between this article and www.hydrachess.com.

"evaluates about 200,000,000 chess positions per second, roughly the same as the much older Deep Blue, but with several times more overall computing power. Whilst FPGAs generally have a lower performance level than ASIC chips, Moore's law allows modern-day FPGAs to run about as fast as the older ASICs used for Deep Blue. The engine is on average able to evaluate up to a depth of about 18 ply (9 moves by each player), deeper than Deep Blue, which only evaluated to about 12 ply on average. Hydra's search uses alpha-beta pruning as well as null-move heuristics [1]. The extra search depth over Deep Blue is due to its use of more modern type B forward pruning techniques that are slightly less perfect, but generally play better due to the greater search depth these techniques permit."

This seems to have been copied FROM the wikipedia; and this text dates back to about June 2005 and earlier in slightly different wording in the wikipedia, whereas the www.hydrachess.com text seems newer.

We're obviously doing good work or something :-)WolfKeeper 18:12, 21 August 2006 (UTC)

[edit] Search techniques

I think Deep Blue surely used alpha beta pruning and maybe null move huristic as well. The algorithms are very old. Also just 12 ply? dubious. Could anybody provide the official sources from IBM or the creaters of Deep Blue? IndiBoy 28 June 2005 23:16 (UTC)

Here's a good article from ICC with Feng-Hsiung Hsu: http://www.chessclub.com/resources/articles/interview_crazybird1.html Basically, Deep Blue would do a brute force search for 12 plies, and once that was over, it would start a selective search. Clearly, this wasn't smart at all, a shorter brute force search would have clearly been smarter.

As for Wolfkeeper, FPGA's are definitely not as fast as ASIC, FPGAs are cheaper and slower. The only reason Hydra can calculate chess positions as fast as Deep Blue is because of Hydra's superior processing power. Hydra can do 100 billion calculations per second, whilst Deep Blue could only do 11.38 billion calculations in 1997. Deep Blue was the 259th most powerful super computer in 1997: http://www.top500.org/list/1997/06/

Well today's FPGAs are pretty comparable to 10 year old ASICs, so in that limited sense you're wrong, and that was my point, although I admit I could have been clearer. Certainly the main processors have improved in speed much more than the chess 'coprocessors', but the bottleneck still seems to be the coprocessors, and they haven't sped up since Deep Blue (although they have been designed better, with more modern knowledge about search techniques).WolfKeeper

[edit] Unfair

As an aside, I think the human-computer matches are currently unfair; a strong gransmaster can look at the game history of another strong grandmaster to look for weaknesses; a system like Hydra usually only plays a couple dozen games--not nearly enough games to look for weaknesses in its playing style. Samboy 29 June 2005 06:37 (UTC)

Yeah, it's a new player. And?WolfKeeper

Omg, these two have nothing to do with chess and chess-programs. :))

[edit] Top player = Hydra?

My logic is: a) Hydra seems stronger than Deep Blue b) Deep Blue beat Kasparov c) Kasparov probably has got somewhat better since then, but still... d) It slaughtered Adams, which suggests a rating more than 200 ELO points higher than him (arguably 300), that gives a rating *well* over 2900. Kasparov is rated at 2812.

So I've put 'probably' in the article.

Still, as Hydra plays more the other players may start to find weaknesses in its play, but I expect it will adapt. WolfKeeper

I have followed these matches and I think Hydra is no more than Fritz running on a typical multiprocessor cluster.

It may well be roughly equivalent to that. Now you come to mention it, I think Brutus used Fritz software. Still, the FPGAs apparently give an order of magnitude speedup; so they probably are cheaper than just throwing raw processors at it. WolfKeeper

The FPGA enhancements did not seem to change the move choices since Fritz running on my laptop gave same move preferences 95% of the time. I think Fritz or even freewares like Crafty running on a bigger cluster / a supercomputer will beat this machine hands down. I believe MIT's Socrates like program running on a 1800 node(yes!) supercomputer will kill Hydra. Therefore the opening statement that it is the best chess computer is absolutely wrong. IndiBoy 29 June 2005 09:18 (UTC)

Of course Kasparov drew against X3D Fritz; and that was on a single CPU. So even if it's basically just Fritz running on 32 nodes, it would be likely to beat Kasparov anyway. But the FPGAs aren't there just to keep the processors warm; they're apparently about 10x faster than the main processor.WolfKeeper

Maybe it's the 5% of the time it does give different choices that win the match? Nil Einne 13:30, 14 October 2006 (UTC)

Well Hydra has beaten every engine it has faced.

This is completely false, it has lost several games and tournaments, all of which documented.

It completely trashed Shredder 8, the former world computer chess champion, in 2004 by the score of 5.5:2.5 http://www.chessbase.com/newsdetail.asp?newsid=1875 So it's fair to say that Hydra is probably the best over the board chess player at the moment.

In the latest freestyle tournament, it finished the final at 10th place, among ten finalists. Two unaided multiprocessor Rybkas, in contrast, were among three players that were tied for third place in the same tournament; one of these unaided Rybkas had been winner of the main tournament. All of this happened in spite of Hydra receiving human assistance; the latter fact can be verified quite easily by inspection of these games, because unlike a stand-alone engine, Hydra used a significant amount of thinking time right from move one in all of these games. With these facts in mind, I think it is not unreasonable to at least have some doubts about Hydra being stronger than Rybka running on an eight-way box.

It's obvious that many chess programs running on sufficient fast iron would very probably beat Hydra. However, even then, one match isn't statistically significant unless the win is very large. Also the freestyle nature of the tournament muddies the water, perhaps Hydra had human 'assistance' and they blew it, or Rybka did, and they totally didn't blow it. Unless we have a cite specifying exactly how ZOR_CHAMP and Rybka played, it doesn't really prove anything.WolfKeeper 09:07, 24 October 2006 (UTC)

In light of these considerations, I also do not understand why my additions to the tournaments section were reverted.

They just seem(ed) to be arbitrary OR, without any references, and appeared to be deliberately restating existing information so as to denigrate Hydra, but without adding any new information. The latest Freestyle tournament stuff is a new data point though, that probably needs to go into the article.WolfKeeper 09:07, 24 October 2006 (UTC)

No scientific basis at all for this statement. A score of 5.5 to 2.5 does not determine that Hydra is even just stronger than Shredder with any reasonable degree of significance. It would help if Hydra played matches against other computers, but Hydra does not even participate in the world championships.

Shredder didn't win a single game,

Irrelevant, the score was still only 5.5 to 2.5, which means that Shredder could still easily be hundreds of points stronger than Hydra. Just do the maths. You simply can't make such conclusions until you have more games, and Hydra refused a rematch (sounds familiar?).

That's theoretically possible. However, the likelyhood is that it's the other way around and that Hydra is much stronger than Shredder. You're also failing to take into account that Hydra has gained extra hardware recently; that means that its strength is likely to have increased rather significantly. And they've worked on the openings to very good effect. Shredder is apparently stronger at the endgame- assuming it ever reaches that point.WolfKeeper

That's more of an argument for Shredder. Software is easier and faster to improve,

Uh huh. If that's the case why isn't Shredder a lot stronger than it is already? The programmers just decided that 2800+ was good enough, and they couldn't be bothered? It's not that easy.WolfKeeper

and nothing prevents you from running Shredder on a bigger machine (quad dual core e.g.).

Prevents? No, clearly. The Shredder team haven't though. WolfKeeper

but sure, taken on its own it means little. In context of all the games it has played, and the comments of its owners (who claim a rating of over 3000, which is consistent with its current playing), I think we have a pretty fair idea of its strength.WolfKeeper

The claims of the owners cannot be considered fact.

I think they can, to a degree. If they're obviously out-to-lunch then everyone will laugh at them. Right now they seem to be reasonably consistent with it's over-the-board performance. My rough estimate is 2900-3000 equivalent rating.

Realistically there is almost nothing to go on, as any modern computer program is expected to crush a human.

It's not just that it beat Adams; it's the number of times it did it. And it's not that Adams played badly. Hydra seems to be over 200 points higher than Adams right now. That puts it at about 2900+; comfortably ahead of Shredder.

You can see very strong GM's getting crushed every day on the internet chess clubs. Again you are using an extremely small amount of games to make a totally wild speculation about rating and strength.

However, I do believe any of the top 500 supercomputers would destroy Hydra, even the 500th fastest supercomputer is several hundred times more powerful than Hydra's hardware. Don't be surprised if one of the top 500 supercomputer makers challenges Hydra later this year.

"Entity" is a vague term that could possibly include Advanced Chess teams. I think it's not hard to say that if you let Anand play a match agaist Hydra with his laptop, the score would be 5.5 to 0.5 -- in the human's favor. Besides, Hydra has not yet played a match against any of the strongest humans, so I think it's speculative to make grandiose statements of this kind. --Malathion 5 July 2005 09:02 (UTC)

Well, thrashing the 7th in the world, he is a top player. As is Shredder. And it annihilates Shredder at 2800! And the strength is very much thought to be much higher than Deep Blue, which took Kasparov. Hydra is 6 ply deeper searching- that's quite a lot. I think the rule of thumb is 30-50 ELO per ply. The big question is the positional strength. Anyway, we'll see, it should be interesting. And the flipping thing is still only playing with half its processors :-) WolfKeeper

Incidentally, when Garry lost to Deep Blue his rating was thusly (taken from the IBM/Deep Blue website): 1 Kasparov, Garry RUS 2795 11.00 2806.00

So he was just 2806. This implies that Deep Blue was roughly the same, probably slightly higher. That's only the same strength as Shredder 8. Either way, Hydra is a lot stronger than that; Kasparov is only slightly stronger.WolfKeeper

Who estimated the rating of Shredder 8? Probably its marketers. --Malathion 8 July 2005 16:55 (UTC)

Ratings are easily calculated given enough games; and the ratings can be connected to the human's ratings given some matches between them.

The Shredder rating is based of the SSDF rating list, which consists of computer-computer matches. It was seeded against human ratings, but this was a long time ago, and it's very doubtfull they are still accurate (but they are internally consistent).

From the games that have been played, they're not that far out.WolfKeeper

The played games don't allow any ratings estimate that doesn't at least have an error of several *hundreds* of ELO's. Again, _please_ do the maths. You are wrong and you will understand why.

I'm not only looking at the games. I know something about the way it works, and adding the extra hardware has a fairly predictable effect on the strength. I think that's why they're claiming 3000, they've done the maths. You're probably right that there are pretty large error bars on the rating right now; particularly if you include games from 2002. But I think only games from the last year or so count.WolfKeeper

Ratings are necessarily unambiguous- anyone who does the calculation gets the same answer. Ratings are *not* a matter of opinion. It doesn't matter if the marketers calculate them or not. Short of outright fraud. If you're claiming that; I'd like to see your working to prove it.WolfKeeper

I think all these grandiose statements about Hydra's strength are speculative. It is certainly exremely strong- but even in other chess articles on Wikipedia, where we have far more evidence (in the form of games against top level competition) -- for example, Garry Kasparov -- we have shied away from claiming that so-and-so is the "best player ever" or "strongest in the world". --Malathion 8 July 2005 16:55 (UTC)

It's impossible to prove 'best player ever'. Strongest human in the world is reasonably clear for Kasparov.WolfKeeper

The truth is that we know very little about Hydra's capabilities, so I think Wikipedia should be wary of rushing to judgment here. --Malathion 8 July 2005 16:55 (UTC)

I agree. I think 'probably the strongest in the world' is defensible; more than that isn't. It's theoretically possible it has a glass jaw against humans, but it doesn't look like it so far. And if it did, I expect they'd redesign the 'jaw'.WolfKeeper

Well Hydra crushed its opposition in this year's PAL/CSS Freestyle tournament without losing a single game. It is interesting to note that the second place and third place finishers used the strong newcomer program Rybka. I think it's safe to say Hydra is the strongest OTB chess player in the world today. Dionyseus 16:39, 13 April 2006 (UTC)

[edit] The handle Zor_Champ

Hydra has always used the handle Zor_Champ in the Playchess server, this has been known for years. When you say "team," it makes it appear as if they use a commercial program along with their Hydra engine to decide on what moves to play, which is untrue. Dionyseus 21:35, 27 April 2006 (UTC)

I didn't say team; their website says team, and I refered to where they said that; you didn't provide any evidence at all.

But what they mean by "team" is that they as a team created Hydra, in other words they want some credit too. Log into Playchess and ask them yourself, they regularly test their engine modifications in the Engine room. Their entire goal is to prove to the world that Hydra is the strongest chess entity, it would make no sense for them to use the aid of other engines, or human aid during games. Dionyseus 22:09, 27 April 2006 (UTC)

No, that's not the way it works; you have to prove your shit. You say by 'team' they don't mean what everyone else means by team. Prove it; and until you have proved stop violating NPOV.WolfKeeper 23:30, 27 April 2006 (UTC)

And even if what you say is true (and I've seen contrary claims elsewhere, and the team do have GMs on tap); that doesn't prove that Hydra has the highest Elo; or establish what it is, they haven't played enough games yet; it takes more than a couple of matches. Even if they beat the other centaurs and wotnot that could be luck. Hydra's probably running at 3000-3100 anyway, you wouldn't expect it to lose many against the top players.WolfKeeper 22:01, 27 April 2006 (UTC)

I'd like to know why you insist on putting in the article that centaurs regularly outperform Hydra. Where is your proof of this? The recent 2006 PAL/CSS Freestyle Tournament clearly shows otherwise. Dionyseus 22:09, 27 April 2006 (UTC)

I'd like to know what makes you think that a centaur that includes Hydra wouldn't outperform Hydra; since humans+machines > machines or humans is quite a normal finding. In fact that's the main point of the freestyle tournament, to work out one way or another. Seems to me you're just deleting stuff you don't agree with; and you're doing it with absolutely no references; in fact you actually expect me to provide references that your stuff is true!!! As in no; YOU log onto playchess and get a quote. If you can't be bothered to reference your own claims, then your claims don't deserve to be in the wikipedia; and you certainly shouldn't be removing other peoples POV. Do I need to sic the admins on you? They won't put up with your stuff anymore than I do.WolfKeeper 23:30, 27 April 2006 (UTC)

First of all Hydra cannot be duplicated so it cannot be used against itself, it is a machine not just a program. Second, who knows whether a human can actually provide helpful suggestions to Hydra, for all we know any advice from a human, even from Garry Kasparov, would be weaker than what the machine can produce. I think that we may indeed have to undergo mediation because it seems you are either refusing to listen to what I'm saying, or you are not understanding what I'm saying, and you insist on including your claim that centaurs regularly outperform Hydra when I have demonstrated that claim is clearly untrue, and you refuse to provide references. Dionyseus 23:37, 27 April 2006 (UTC)

Bold text