In your opinion, what are the fairest rules to have a TRUE human versus machine match, regardless of which machine and which human? I would like to see the same standard rules be applied to ALL human versus machine matches. In this case, there can be no implication of unfairness.
– Should the opening and endgame tablebases be taken out from the computer?
– Should humans have the same opening and endgame tablebases?
– Should the time control be longer or shorter?
– Should the length of the match be longer or shorter?
– Should there be adjournments where humans can use machine to analyze?
What are your suggestion? 
Chess Daily News from Susan Polgar
There should not be special rules. The time has come to admit that computers are better than humans at chess. There is no reason to “dumb” the computer down to make it easier for a human player.
If there are special rules, then allow more time for people and, also, allow a team of GM’s to play a computer. Otherwise, the programs have us.
For a true match the machine would have its opening book and tablebases (any size tablebases that are available, 6 and even partial 7 man etc).
For the human, we all have access to the same openings via ECO indexes and so forth but wouldn’t be able to have these materials on hand, just memorized.
As for adjournments? Only if the GM has no access to phones, computers etc. As the computer in the match should also not be left on all night to analyze.
As this is impractical, the games should probably be SHORTER rather than longer games. This effectively handicaps the computer too. So would lopsided time controls like Kramnik having classical controls while Fritz had rapid controls etc.
All the same, I think it impossible for any human to equal the raw calculations that a computer can do. Give the human more time.
I think that if possible both players should play with “same” conditions:
No opening trees, no tablebases, no game database.
It’s a bit difficult, because players keep a lot of that info (info about openings, endgames, known games), but I think that if the computer programs is strong enough or well designed, should be capable of playing as strong as the player is.
So my selection is: Fritz without tablebases, game database or openening trees.
By the way, Susan, your sister Judith will be playing a blindfold match against Topalov in Bilbao con 7, 8 and 9th of december, in the famous Guggenheim Musseum. Surely I’ll be there one or two days.
Best wishes from the Basque Country.
We spent ages trying to make the machines strong. Now, Are we gonna try to make them weaker so we can please ourselves?
It isn’t human vs machine, but rather human vs human/machine. The contest cannot be fair, from the get-go. The computer has all the advantages of both man and machine.
I think an interesting contest would be Kramnik/Fritz vs. Kasparov/(place computer program here). Why not? Computer assisted competition might be the thing of the future. Think of it as a sport like cycling, stock car racing, etc…. the human has a machine as part of the contest.
I think the computer people shouldnt be allowed to change engines during a game or match. Thats like having more than one brain. Giving the human player time to analyze with computers defeats the purpose of computer matches.
If you wish to test ‘man’ against ‘machine’, there should be no constraints on either. VK-F10 fails to be that test, especially as F10 is likely not the best chess engine around by a distance.
If you wish to constrain the players in some way, that is a more focussed and specific test. I would constrain the human not to blunder horribly. When they do, we all have our time wasted. One could also halve or quarter the time available to the computer rather than penalise it in some more artificial way.
It takes skill to defend some 5-man endgames and it is a pity that VK is being relieved of that task. It takes even more skill to defend a 6-man or 7-man endgame. I think this rule was first proposed and accepted for a Kasparov-DJ match in New York.
I doubt that we shall see the irony of VK resigning six KNNKP endgames, all with 50-move draw-claims available.
you have to give kramnik the same resources that the computer has on it’s hard drive as it relates to openings and tablebases … in other words, things that don’t require calculating
No special rules… either beat the computer or not… but to change rules around is ridiculous… Kramnik is so horrible for chess… he does everything to gain all kinds of advantages, I personally can’t stand him.
This match is a sham…
This whole argument of opening and endgame tablebases is quite silly.
I’ve been looking at this thread and wondering why you would handicap the computer but allow it to play Kramnik.
To remove the endgame tablebases from the computer is unfair if Kramnik has EVER studied the endgame. How is that fair to Mr. Fritz?
You want the machine to purely calculate everything in every position?
The ‘remove this and that’ argument is just plain silly.
-vohaul
– Should the opening and endgame tablebases be taken out from the computer?
This isn’t really possible. The databases can be taken away, but the engine will still have special logic for these two situations. Much better would be to allow the human to use any databases they wanted to (but no computation).
– Should humans have the same opening and endgame tablebases?
Yes, as indicated above. These databases are really seperate from the game itself.
– Should the time control be longer or shorter?
Classical time controls would give the human his best chance. In many cases, if an engine is going to miss something because of a hole in its chess knowledge, it will miss it no matter how long it has to ponder. Giving the human a time advantage is also a very good option and at some time advantage there would be possibilities for the human, no matter how strong the engine was.
– Should the length of the match be longer or shorter?
Computers don’t care, so this should be determined by the preferances of the sponsor and the human player.
– Should there be adjournments where humans can use machine to analyze?
This is contradictory to the goal of the excersize.
Properly staged, I think these matches are actually very interesting, pitting chess knowledge against brute force speed. I think the matches that Kasparov played against engines could have been very valuable for the chess community and its a shame that he felt it necessary to accuse the IBM team of cheating. Up until that point, IBM got good publicity for the match and probably would have sponsored additional events.
Regards,
Alan
What question are these matches designed to answer? In Kasparov vs. Deep Blue, the question was, “Can a computer beat the best chess player in the world?” Now that we know the answer, the question has become, “How much must the computer be handicapped, for the human to still have a chance?”
The silliest of the suggestions is that the computer should not have an opening book. Every top-flight human player knows hundreds, or even thousands of openings by heart; so, why shouldn’t the computer?
There is perhaps a better argument for removing tablebases, since humans tend to play endgames with general knowledge, not by memorizing every possible position. So, if you remove tablebases, the games would test the general chess knowledge of the engines, rather than memorization. But many endgames happen in a time scramble, when the computer (which lacks the human’s gut instinct) would be at a disadvantage. Also, in recent years I don’t think programmers have focused on developing endgame knowledge, because they assume tablebases will be present. Remove them, and you deprive the computer of what it does best.
I tend to think that, in blitz or rapid time controls, computers would invariably prevail, due to the human’s increased tendency to blunder in shorter games, and computers’ tendency to punish those blunders instantly.
I think adjournments should be permitted only if the machine is also allowed to analyze the position all night, since clearly the human will be doing the same thing. Most modern programs are not designed for adjournments, but I think this could be programmed quite easily.
I have no opinion about the length of the match, except that I suspect a long match favors the computer, due to fatigue.
Historically the human was much better than the computer. so humans gave up all kinds of things to help out the computer so that the computer would look stronger.
Now these advantages have been institutioinalized into the computer even though the computer has gained tremendously in strength.
I do not see the contest being on equal footing if the computer has opening books and endgame table bases.
I never have used an opening book with my fritz program. I use it right from move 1. the computer is capable of making good choices right from move 1. that is the way to match them up equally. the object is to see if the computer can play the game. the GM has to figure out the best moves right from the beginning, so let the computer do the same.
When the endgame arrives, the human has to work out the position on the board. Let the computer do the same.
The computer program needs to play all its moves based on its programming and not on look up tables with the answers. If the computer program falls short in the opening or the endgame then it does not play as well as humans and the programmers need to work on that.
also I always felt the human should get a little more time. the compter does certain tasks almost instantly and the human takes a few seconds. those seconds need to be given back to the human.
If the computer can calculate so deep in say 3 minutes then the human needs whatever time it takes him to go so deep. maybe he needs 5 minutes or more. that is only fair that both sides be given the time necessary to calculate to a point where the player is comfortable making a move. right now the human is rushed and the computer has plenty of time.
with an opening book, the computer can play 30 moves in one second. a rediculous speed gaining unfair time on the human.
What we really want to test with a computer program is if the program can calculate correctly and make good decisions about its move choice. Right now it depends more on simply waiting for the human to make a mistake under time pressure and going in for the kill. not necessarily proving that the move choices of the computer were better than the human choices when not making a blunder.
One problem with the upcoming match is that although Mr. Kramnik is no doubt an exellent chess player, he isn’t the most interesting guy in the world and isn’t a good promoter. I think if Mr. Friedel had really thought this through, he would have had DF10 playing against the Polgar sisters playing as a team. With Susan’s promotional abilities, this match would draw ten times the interest of the planned event. The team approach would have the added benefit of reducing the chances of blunders. Of course this would require the team to actually be able to move the pieces around on an analysis board. This would add additional interest by showing the thought processes of the human players during the match!
Regards,
Alan
Alan,
I challenged Deep Blue when I was the reigning WC. I also challenged others. They all politely said no.
Best wishes,
Susan Polgar
http://www.PolgarChess.com
Perhaps future chess engines will incorporate accurately calibrated FIDE-scale rating controls, so a user might “dial in” a rating and expect the same challenge that a human of the same rating would offer. Players could then use their computers to accurately gauge and track their strength.
Also, the “Man-vs.-Machine” argument would be rendered moot, as any computer with a top-end rating significantly higher than the world champion would be obviously better.
1) Allowing the computer to consult an opening “book” (even if it is a virtual book in its “memory”) is against the rules of chess that do not allow outside help. If it is allowed for the computer, then the human should also be able to bring all his books and his personal notes with him to consult freely during play.
2) Same goes for the table bases.
3) The computer should also have to make all its own moves, physically, on the board, and punch the clock, just like a human!
Although the comp can be expected to beat Kramnik, there are still shreds of hope left for humankind. The imperfection shown in that Kramnik game last time when he resigned the drawn position, that the computer had badly miscalculated, has not to my knowledge been resolved. Even the strongest computers still over-emphasize material advantage in calculating some positions.
I also recall a game a couple of years ago in which Kasparov beat the machine with a slow Q side pawn strategy. The comp had no good plan. And a game about 4 years ago in which Van wely smashed the comp with a K side attack and the comp didn’t have a plan at all
People seem to be missing a very significant point about all of this… Human’s are better than computers at chess… Haven’t you noticed that it is Deep Fritz, ‘Deep’ implying that it uses a special machine. Open your engine and look how many positions it analyses per second and you will see that it is about 700 000. Now Hydra analyses 200 000 000. This roughly means that it is 200 times more powerful than your computer. If Kasparov played your engine be it Fritz or Rybka, he’d win…
I, also, am a bit unnerved by everyone’s absolute hatred of draws. If two players play equally well than there should be a draw. The ultimate level of performance by both players would result in a draw. I’m not encouraging stage draws in 15 moves though.
So complaining about draws that are certain to come is pointless. Also to this point, if you were playing against an opponent that was a superb tactian would play a highly tactical opening or game? Most people would answer with no.
Playes have to take advantage of theirs strengths and the weaknesses of their opponents.
Alan,
Why would IBM waste its time and money playing someone else after they beat Kasparov? Not to take away from the other players mentioned, but what challenge is left for them? Their machine would easily crush anyone several hundred points below Garry if it beat him so easily.
Chess matches between humans and computers will NEVER be fair. They can’t be and that isn’t the point. However, I believe, they can be interesting and instructive.
Remember, this is the 21st century but not science fiction. A computer (alone) cannot play chess against a human. With software (programmed by humans) and human operators they can. Yes, modern technology has advanced a tremendous amount allowing faster computation, more memory, etc. but the real advance has come in the ability of the programmers. A computer is a tool. In this case, a tool used by humans (programmers, operators, and users) to help play chess faster and hopefully better. Usually, a machine or a human with a tool will outperform a human alone. I don’t think we should be looking for the John Henry of chess.
Just like a swim meet between humans and dolphins, or a race between athletes in wheelchairs and runners (wheels are faster than feet), a chess game between a human and computer will never be fair. Let’s not worry about fair but rather how we can make it interesting, educational, or even entertaining.
-mm
Machine should definitely get to use its databases.
Whether the human gets to use databases, adjournment, etc. depends on whether you want OTB chess or a watered down version of correspondence chess. Both can be ok.
Human should be able to get a fairly large number of rest days on request, more than in a human-human match. Whether human chess skill is still ahead of machines is an exciting question. But we all realize that humans do not have machines’ endurance. A machine can play 24 hours a day for months at a time! Traditional matchess without much resting takes a toll on both players, so it’s even, but in a human-machine match, only one player is affected.
So there should be more rests.
I believe that the human should have the same opening book resource as the computer; the machine has millions of preprogrammed lines, so the human should not have to rely on memory – heck, give him a copy of ECO and the freedom to consult it!
I also believe that the human should be given the ability to move pieces on an analysis set – either a physical one, or an electronic display. The computer is not relying on it’s ‘mind’ to envision future positions with possible inaccuracy; when they say that it has analyzed each position 12 moves deep, they mean it – it’s program has tables and arrays that allow for exact analysis of each potential position. The human should have the same resource.
The contest should be about making choices; until the game of Chess is mathematically solved to the last move, one has to make choices – based on instinct or analysis techniques. I believe that THIS is where the real contest between human and computer lies, and I offerred the suggestions above in order to give the human a ‘fair’ playing ground.
Unlike some posters in the previous thread who think that Kramnik’s conditions are giving the human an unfair advantage, I believe that it is computers that are enjoying several ‘unfair’ advantages. The human must be provided the means to play with the same tools.
I would not handicap the computer; I would enable the human.
What we’re seeing here is the difference between “Man vs. Machine” and “Man vs. Program”.
The “machine” has a hard drive and can fill it with databases so that it doesn’t have to compute endgames or openings. Humans have been passed — but not necessarily because the machine is better at analyzing chess. Maybe it’s that the machine has so much memory.
The “program” is the logic whereby the computer determines what moves to make. This is by far the more interesting matchup now. Here, I agree with an earlier anonymous poster — if the program cannot compete without opening and ending databases, perhaps that is an AI challenge suggesting that the *program* has room for improvement.
Honestly, I cannot see how a computer can use precomputed data on millions of positions and claim that its programming wins. Maybe it’s the data that wins instead!
The fair rules are that both sides get databases, or both sides do not. Since no one wants to see grandmasters looking up positions while at the board, and no one wants to see the computer declare a draw in a 6 piece endgame, it’s obvious to me that no databases is the way to go.
With Databases ~ correspondance.
Without databases ~ over the board.
The use of endgame databases by programs in match play is ridiculous. Imagine Topalov showing up to the WC with reams of printouts about how to play perfect endgames in certain situations. This is what computer programs have access to, but humans aren’t allowed the same advantage.
Computers will ultimately destroy chess. Man will ultimately destroy chess. Xiangzi is our future.
Allowing the computer to consult an opening “book” (even if it is a virtual book in its “memory”) is against the rules of chess that do not allow outside help.
This comment, and a few others like it, show a fundamental misunderstanding. The computer’s opening book is simply the machine equivalent of what any grandmaster knows.
When Kramnik as Black sees 1.e4, he doesn’t calculate his options as if he’d never seen that move. He relies on an immense memorized database of openings. It is irrational to say that Kramnik is allowed to “memorize” openings (which clearly he does), but the machine cannot.
Obviously the machine’s opening book had to be created by a human, because without human intervention there wouldn’t be a computer in the first place.
But Kramnik’s “opening book” isn’t fully his own creation either. Any modern chess player is relying on centuries of theory developed by others.
The rule of “no outside help” applies only during the game. For instance, the programmers can’t change the opening book on-the-fly, and Kramnik can’t phone a friend for advice. But between games Kramnik certainly can, and does, seek outside assistance.
The only fair way: no special rules at all. If we start to restrict the computers, give special advantages to humans, the entire issue becomes an artificial and arbitrary manipulation.
Based on the same concept, I could play against Kramnik, he gets 5 minutes, I get 5 months, and quite possibly I may end up winning. But what would be the point?
One anon wrote:
“We spent ages trying to make the machines strong. Now, Are we gonna try to make them weaker so we can please ourselves?”
I agree, 100%.
Gabor
Let the human consult printed books; this may or may not consist of the printout of the machine’s table-bases (depending on what the human wants to consult). This will be most fair because, the computer is affectively doing the exact same thing while it is playing. The human consulting the book via a computer is incompatible with the spirit of a “human” vs “computer” concept.
Rgds M.
we write 2006 – computer technology and programming methods evolve two or more times faster, than any chess knowledge can evolve.
the sundown of human chess superiority on computers did not start yesterday – but with the upcoming of software like Fritz, Junior, Shredder or Tiger a few years ago.
the use of opening books, hashtables and tablebases by chess software (we do not talk about machines – hardware – what hardware can do to super grandmasters was already shown by IBM or by Donninger’s Hydra project)is, strictly according to FIDE rules – CHEATING (hashtables – remember: it is forbidden in OTB matches to WRITE your next move on the score sheet BEFORE you have done it!).
to be honest – i’m not really interested in the Kramnik vs. Deep Fritz!10 match – it is a nice advertising for chessbase products, it is a nice income for Kramnik – maybe a spectaculum – but nothing more. so, the playing conditions are even of a lower interest to me, because there won’t be any “fair conditions” in a match between a humans and a software (“machine”) – memory, i’ve read here; stamina, i’ve read here – nothing counts but tactics!
be sure – if you can calculate out your opponent simply by one move ahead – you will win the game with over 90 percent chance. horizon effects are not known to happen in computer chess only… and “understanding” of chess is much overestimated amongst class players – and master players and grandmasters, too, i suppose… (too much sovjet old school…)
so – end of my lecture – the match Kramnik – vs. Fritz10 will end in a 3:3 draw – Kramnik will take the money and chessbase earn the money back – and everything will be allright.
– some idiots like me will buy fritz10 and rybka2.2, will start a tournament to watch the investment working – inhale a daily dose of chess and for the rest, they stay waiting for the next weekend – btw i’ll be white next weekend and i intend to play 1.e4 … – to make chess what it is – A WONDERFUL FULLTIME JOB WITHOUT SAVINGS …, no, no, no – but the greatest game this world has ever seen…
^^
the use of opening books, hashtables and tablebases by chess software … is, strictly according to FIDE rules – CHEATING (hashtables – remember: it is forbidden in OTB matches to WRITE your next move on the score sheet BEFORE you have done it!).
The computer is not writing a move before it has been played. It is remembering past analysis. Kramnik also remembers past analysis, and there is no rule against that.
A hash table is simply an electronic memory of positions already analyzed. In a game against humans, Kramnik doesn’t “forget” everything he’s already analyzed or learned about the position. He retains the memory of what he previously calculated.
In other words, opening books and hash tables are just computer implementations of what a human player would do. Unlike the human, the computer’s memory is of course perfect. But to ban the computer from having a memory is not a fair rule, unless human memory is also forbidden (which, of course, would be absurd).
@marc – a hashtable is – in human terms – nothing more and nothing less than a NOTEBOOK …
sincerly
@marc – a hashtable is – in human terms – nothing more and nothing less than a NOTEBOOK …
Incorrect. A hashtable is nothing more than one of the many data structures that computer programs use remember things.
At some point the analogies break down, because computers are made of silicon, and brains are made of cells. We know precisely how computer memory works, because humans invented it; the precise mechanism of our own memory has not yet been discovered.
But clearly human chess players do memorize positions they have previously thought about, and “write” them to memory cells for retrieval later on.
Susan Asked:
Should the opening and endgame tablebases be taken out from the computer?
Yes.
Should humans have the same opening and endgame tablebases?
No. there should be NO endgame tablebase OR opening book for the computer.
Should the time control be longer or shorter?
Time controls should be FIDE STANDARD Match length, whatever that is at the time.
Should the length of the match be longer or shorter?
FIDE Length of WCC, whatever that is at the moment.
Should there be adjournments where humans can use machine to analyze?
NO Adjournaments. Period.
Chess Training Blog
Should the opening and endgame tablebases be taken out from the computer?
Yes.
Are you going to somehow force humans to forget their memorized openings?
Should the time control be longer or shorter?
Time controls should be FIDE STANDARD Match length, whatever that is at the time.
There are multiple standards.
Should the length of the match be longer or shorter?
FIDE Length of WCC, whatever that is at the moment.
Who would pay for that?
Should there be adjournments where humans can use machine to analyze?
NO Adjournaments.
Here I agree.
@marc – your exertion on “silicon minds” reminds me a little of the famous fight between man vs. windmill in the novel by Miguel de Cervantes… 🙂
interstingly i find you on the side of the windmills – but the windmills lost the fight against the last knight of Spain – and maybe the World – by demolition – an exercise man is really good in.
hashtables ARE notebooks … and memory is an inexpensive thing for silicon minds (the windmill), but not for biological neuronal networks (the knight) … to force a computer’s memory to act like a brain (human or fish, all the same), would assume a NON partitioned – or better – OVERLAPPING memory use … clearly not given nowadays – it is not only the processor to make a computer go round – and there is no processor in a biolagical brain – the brain itself is processor and memory in one piece.
so – the novel’s first part was published in 1605, the second part 1615 – and as i’m used to cite:
“What has been will be again, what has been done will be done again; there is nothing new under the sun.”
Ecclesiastes 1:8-10
Sorry, the Don Quixote analogy is lost on me. Hash tables are just a way of structuring data inside a machine, one of many that has been devised. If a computer is not permitted to store data, then you don’t have a useful machine any more. Fundamentally, I just have to assume that Vohaul thinks the hash tables are doing something more grandiose than they really do. They are just a way of organizing memory.
@marc – hashtables do indeed a lot more, than simply structurizing databases or memory – especially in chess software.
you can perform a little test of this “more” with your own chess program at home.
chose a little “brain teaser” – e.g. the mate in five position here in this blog. measure the time, your chess software needs to solve the problem with given hashtable size.
reduce the hashtable size about one third – perform the same task
reduce the hashtable size by one more third – perform the same task again
reduce the hashtable size to zero and perform the same task
do the same experiment with increasing hashtable sizes (up to the maximum your hardware can deal with)
as a result you will clearly see the “more” hashtables can add to chess software performance
remember – chess is a game with time limits – for both, the human and the software side…
for a human the use of a pencil and a piece of paper during chess position calculation or the possibility to move pieces around on a chessboard, would exactly do the same job, as hashtables do for chess software, it structurizes AND economically saves MEMORY PROCESSES to avoid uneconomic use of limited time
moreover computer storage techniques and increasing hardware memory made it possible, to STORE hashtables for later use … the nalimov is not a very bad example for this …
PS: in a romantic flash i believed Kramnik to be a modern Don Quixote, the last knight from ancient times to stand up against the proceedings of HUMAN technology. however, Vlad does not fight windmills by belief, but by persuation of his own sponsors instead, what makes indeed a big difference in motivation… 🙂