'Chess Style' and 'Quality Chess'

[Ed Seedhouse]

Re: 'Chess Style' and 'Quality Chess'

Apparently my version is a variation. As per Wikipedia, "The period is often quoted as 18 months because of Intel executive David House, who predicted that chip performance would double every 18 months (being a combination of the effect of more transistors and their being faster)"

You still have it confused. Moor predicted the the number of transistors on a chip would double roughly every 18 months. That is not the same as "performance" though it is related to it. Modern chips also run at far, far faster clock speeds, and this also contributes to their increased performance, not to mention the exact configuration of transistors on a given chip. Moor did not say anything about clock speed increases or RISC chips, among other things. His original prediction was only about the number of transistors that could be fitted to an integrated circuit chip.

[Garland Best]

Re: 'Chess Style' and 'Quality Chess'

Well if you want to nit-pick, he actually said the number of transistors would double every 2 years, not every 18 months. And yes, I got that it was the number of transistors that doubled with my last reply. The part I copied just didn't make that clear.

[Paul Bonham]

Re: 'Chess Style' and 'Quality Chess'

Well if you want to nit-pick, he actually said the number of transistors would double every 2 years, not every 18 months. And yes, I got that it was the number of transistors that doubled with my last reply. The part I copied just didn't make that clear.

Hey Garland, you could have nit-picked right back at Ed. He spelt it "Moor" TWICE (making it unlikely as a typo) when it is actually "Moore". :)

[Paul Bonham]

Re: 'Chess Style' and 'Quality Chess'

Paul, you completely missed...

Sorry that I misinterpreted your point regarding the "Chess for Tigers" book. I do question though whether someone could beat Houdini by playing for complications? It would seem not, and that only seems to reinforce the notion that human emotion, as Spock would have it, is a weakness in the realm of logic and reasoning. (Chesstalk postings seem to confirm Spock's view.)

I think we are getting closer to the heart of my question, and it's not important whether I like the answer(s) or not. I can acknowledge and accept that many people are almost by nature going to be subjective about it. After all, there are people who watch 'competitive' bowling and I can't find any objective reason why they would do that. They like to see bowling pins knocked flying through the air? Someone like Freud might say the pins represent the establishment or status quo, seemingly strong in formation, and the bowling ball is like an unanticipated disruptive force that shards it all to pieces. Some people have a psychological need to see that again and again. I wonder if more leftists watch bowling than conservatives?

So we know that computer engines on specialized hardware play the best chess known to man, and we know that computer engines have no emotion. The empirical evidence seems to suggest: play without emotion. Getting back to poker, this is a key tenet for many pros in that field. Suffering a bad beat, where you did everything right and still lost, is a challenge to handle emotionally that chess pros don't come up against. You can't do everything right in a game of chess and lose (although you might draw). People like to say "chess is life" but poker is much, much closer to the chaos of real life.

Incidentally, you can read about computer poker engines here: http://en.wikipedia.org/wiki/Computer_poker_players and the one you may be thinking about could be the University of Alberta project. But it will NEVER be the case that any computer bot will dominate human players in the same way that chess engines do. Poker is too volatile, and that was my point about table tennis also. You can't extrapolate that if the robot player gets faster and faster, it will ultimately win all matches and even points versus humans. The game has elements that can make an individual point (or rally) independent of speed and calculation alone. If you really disagree with that, put a time frame and some money on it. How about 10 years and $1,000? Then we just have to agree on how many points won by the robot without a loss of a single point to the human constitutes proof that the robot is unbeatable by the human. It's not enough that the robot occasionally beats the human. The robot must ALWAYS beat the human. And in fact, since every point in table tennis is a game in itself (no one ever 'sacs' a point in table tennis, except maybe out of fatigue), the robot must win EVERY point against the human. The occasional point won by the human means that even if the human is totally inferior to the robot in speed and calculation, there is still SOMETHING that can allow for the inferior player to win once in a while.

And this gets to my point that competitive chess is not a sport. If logic and calculation alone can win or draw every time against humans and their 'intuition' or their 'chess understanding', then chess as a competition is pure mathematics. The heuristics that an engine uses to evaluate quiescent positions don't involve questions like 'Houdini, how do you FEEL about this position?' There is only numeric calculation. A computer understands nothing more than that. And if such numerics can overcome all other influences WITHOUT FAIL, then you have no randomness or volatility left to speak of.

Every sport, to my mind, must have at least some element of randomness that influences the play. If you want to check out one example of how such randomness could become part of competitive chess, check out my variant called Expedition Chess:

http://www.chesstalk.info/forum/show...ght=Expedition

As we both understand, FIDE will never allow either computers or randomness in chess. You mentioned lack of sponsors, but even if sponsors were to show some interest, FIDE would be sure to stifle that interest because they want nothing to do with proving chess isn't a sport and is pure math. So they have to keep it a sport at least to the minds of some, and the element of randomness that is involved is... human emotion, or perhaps more accurately, the complete state on a given day of particular human beings. People being what they are, some will buy that as a sport.

Earlier in this thread, you said my questions about whether we should outlaw human drivers once self-driving cars are perfected was 'off topic'. Maybe now you can see it wasn't, assuming you have at least a partial inclination to philosophy. Will some people actually encourage human driving as a type of 'sport' in which errors and accidents and injuries and deaths occur, based on the faulty state of the humans involved? Will the law uphold that activity even knowing that the computer drivers, like the chess engines, are vastly superior (maybe even perfect) at preventing accidents? Since it already is accepted at NASCAR and Indy events that there will be accidents and drivers and spectators alike can be killed, I am sure the courts will decide it's worth a few lives to uphold driving as a sport. And thus humans will be allowed to drive and have accidents, all for the entertainment / amusement of some sporting audience.

Regarding Moore's Law, we must remember that chess engines have over the years been encapsulated into more and more chess-specialized hardware, culminating with Deep Blue. So it wasn't just the 18 months doubling of performance, it was a MUCH more rapid increase in performance. The same thing has been going on with computer graphics, where it's both software AND hardware contributing to the performance increases until we get movies like Avatar.

Your point about better and better evaluation functions actually fits in with competitive chess as pure mathematics, and thus deterministic and not a sport.

It's not that I don't want anyone watching chess because it's not a sport, or for any other reason at all. Watch chess if you want to watch it. What prompted this thread is that even the chess players and watchers (who are almost unanimously players) speak constantly about improving the quality of chess. They want the elite GMs to play more and more perfect. They cheer on Carlsen, or the 'next' Carlsen, because he's young and going to get better and they want to see what 'better' is. And as I wrote earlier, we have books, DVDs, opening databases, ending databases and tablebases, all devoted to the pursuit of perfection.

What I say to all that is "be careful what you wish for" because the computer engines show us what that would be like, and no one seems to like it for its own sake (i.e. the highest quality chess itself, regardless of who or what played it). That's why there's no book analysing the World Championship between Houdini and Stockfish.

Perhaps what we need in chess education is not lessons geared to playing better and better quality. Instead, lessons geared to finding the moves that will take the games into the deepest, darkest and wildest recesses of the chess universe imaginable. And instead of rewarding wins, we reward creativity. Something like gymnastics or diving or figure skating, where there's the whole concept of a 'degree of difficulty' and scores are awarded by judges.

Those critical of such a notion should recall that in Carlsen - Anand 2013, I believe it was Game 8 where everyone knew Anand playing Black was on the ropes and there was no doubt he had to 'come out swinging'. And instead, he played possibly the most safe, conservative game of the match, as if he was totally unable to even imagine such a thing as taking a chance. The commentators were dumbfounded. This is what it is today to be an elite GM? Totally unable to take any risk whatsoever?

And then we had Game 9 where Anand did actually succeed in thinking up ways to create complications... only to blunder like a child! Anand may have been a great tactician at one point, but years of elite chess have severely dulled his edge. A repeat of Carlsen - Anand this year is going to be a hard sell.

Now instead, imagine... playing a game of chess with 3 judges seated nearby looking on, making notes and scoring every one of your moves based on things like adventurism, creativity, complexity level of the resulting position. And if you made a great move on those factors and your opponent had a 1-move checkmate in reply, well you would still lose the game, but instead of losing it 1-0, you might lose it something like 10-9. Like losing a round of boxing because you were knocked down, but throughout the round you were the creative aggressor and only made that one mistake.

[Garland Best]

Re: 'Chess Style' and 'Quality Chess'

The standard advice given to play against computers was to play for positions was to avoid tactics, and instead go for positional lines where one had to understand the long term advantages. That strategy no longer works, because the evaluation functions in quiescence positions have improved so much over the years. Playing for complications worked against strong humans instead of computers, because computers do not make tactical errors, while humans will, and in increasing frequency in more complicated positions.

I concede that in poker/table tennis, there is an aspect of chance that means that once in a while the human will win a hand or a point. However, long term the computer/robot will win out, simply due to the law of averages. So for both events my condition would be winning every match, not winning every point. That I could see myself betting on.

In chess, the element of chance is the human. The decisions he makes at the board are influenced by so many non-chess related factors. It's free will, if you believe in it. Chess programs must have randomness put in as well, to choose between 2 moves with identical scores in the evaluation function. Otherwise every game between 2 programs running the same software would always make the same moves. In theory a human can still defeat a computer, simply by playing better moves than the computer (Such moves exist. Otherwise the programs would NEVER lose, assuming chess is not ultimately a theoretical win for white/black.). However the odds of this have become so small it is not worth trying. It is entirely conceivable that the same will happen in table tennis.

The fan favorites are the ones that show the greatest "style" as you define it, but unfortunately subjective scoring is not likely going to ever happen (Sexist aside of the day: ever notice that sports involving judging are the ones that are most popular with women?). The only exception is the brilliancy prize. Tournaments should offer more brilliancy prizes.

Bottom line, I think we can both agree for the average person, chess between two persons still meets the requirements of a sport (set of rules, winners/losers, element of chance (in this case the free will of the person)), and that for the majority of people it is the sporting aspect that will encourage them to follow events and purchase books on matches and tournaments. And that is why matches between computers would not sell.

[Ed Seedhouse]

Re: 'Chess Style' and 'Quality Chess'

So we know that computer engines on specialized hardware play the best chess known to man, and we know that computer engines have no emotion. The empirical evidence seems to suggest: play without emotion.

This seems like a non sequitur to me. Humans are not computers. Computers don't have emotions and play better chess, but it might just be that if we somehow added emotions to computers they would play even better. Since the experiment hasn't been done we can't rule out the possibility.

So on that basis alone your conclusion does not follow from it's premise.

Humans can't play "without emotions". We are animals and animals have emotions all the time. All normal people can do is pretend not to have emotions for a period of time.

We do know that people who are brain injured in such a way that they are no longer aware of their emotions simply cannot manage their lives and almost always end up in hospitals or jails. There's lots of clinical experience in that area. And we have lots of evidence that people who suppress their emotions for long periods tend to get seriously neurotic. And this weakens their chess skills, not improves them. I speak from personal experience.

Most of the grandmasters play with "intuition", which is just another word for "feelings". They calculate what they can, recognize what patterns they can, and when they come to the limit of that they play what they feel is best. They play for the position they like more. I have read several books that claim that grandmasters actually calculate less than weaker players on most moves, and there seems to be a good deal of research supporting that thesis.

Being humans, not machines, we have emotions and can't afford, in my opinion, to divert our attention into suppressing them at the chess board any more than in the rest of our lives.

[Paul Bonham]

Re: 'Chess Style' and 'Quality Chess'

This seems like a non sequitur to me. Humans are not computers. Computers don't have emotions and play better chess, but it might just be that if we somehow added emotions to computers they would play even better. Since the experiment hasn't been done we can't rule out the possibility.

So on that basis alone your conclusion does not follow from it's premise.

Humans can't play "without emotions". We are animals and animals have emotions all the time. All normal people can do is pretend not to have emotions for a period of time.

We do know that people who are brain injured in such a way that they are no longer aware of their emotions simply cannot manage their lives and almost always end up in hospitals or jails. There's lots of clinical experience in that area. And we have lots of evidence that people who suppress their emotions for long periods tend to get seriously neurotic. And this weakens their chess skills, not improves them. I speak from personal experience.

Most of the grandmasters play with "intuition", which is just another word for "feelings". They calculate what they can, recognize what patterns they can, and when they come to the limit of that they play what they feel is best. They play for the position they like more. I have read several books that claim that grandmasters actually calculate less than weaker players on most moves, and there seems to be a good deal of research supporting that thesis.

Being humans, not machines, we have emotions and can't afford, in my opinion, to divert our attention into suppressing them at the chess board any more than in the rest of our lives.

The language I used ("The empirical evidence seems to suggest...") showed that I was not drawing a firm, written-in-stone conclusion. So yes, your scenario of adding emotions to computers would be an interesting experiment. Since it's impossible, I suppose we'll never know.

I certainly understand what you say about emotions, that we are biologically [evolved or created, take your pick] to need them, but suppose we could reverse your experiment. Suppose we could (safely) remove all emotion from a human being, and suppose we did it on a chess GM. Would that GM play better or worse? We don't know.

And so it would seem that your own conclusion, which you do note is your opinion, is itself a non sequitur. We'll never know either side of this issue: would computers play better with emotions? / would humans play better without emotions?

On the human side of the issue, I'll again mention poker because poker players have a much bigger problem with emotion that chess players do. And every poker pro will tell you, strive to play without emotion. Do not let either the ups or the downs impact your emotions whatsoever. Do not feel great when you win, and do not feel terrible when you lose. Either reaction will have you subsequently playing sub-optimal poker.

Since this is so much a part of poker, perhaps a study should be done on professional poker players to get insight into whether or not suppressing emotional reactions, which definitely is better for poker play, is in the long run harmful to the poker player.

But we do have now 2 strong pieces of evidence that less emotional involvement in a game involving mental work produces better results. We have the poker pros telling us it is the case for them, and we have the computer chess engines showing us that their lack of emotion doesn't seem to affect results when playing emotional humans (although we can't say that 'all else is equal' in that case).

Einstein wrote hypotheses about spacetime from much less evidence than this, and some of his hypotheses may yet be proven to be wrong, but he went ahead and wrote them anyway. That's all I'm doing here on a different subject matter.

[Ed Seedhouse]

Re: 'Chess Style' and 'Quality Chess'

We'll never know either side of this issue: would computers play better with emotions? / would humans play better without emotions?

It is an empty question since to be human is to have emotions, with only rare and accidental exceptions. Such people are brain damaged and they can't function very well. Our emotions are involved in every thinking process and the idea of emotionless play is self contradictory. What people really want when playing is to be alert but serene. Serenity is an emotional state, it is not a state of freedom from emotion. Wanting to win is an emotion too. If we didn't want to win why would we even be playing?

[Paul Bonham]

Re: 'Chess Style' and 'Quality Chess'

The standard advice given to play against computers was to play for positions was to avoid tactics, and instead go for positional lines where one had to understand the long term advantages. That strategy no longer works, because the evaluation functions in quiescence positions have improved so much over the years. Playing for complications worked against strong humans instead of computers, because computers do not make tactical errors, while humans will, and in increasing frequency in more complicated positions.

I concede that in poker/table tennis, there is an aspect of chance that means that once in a while the human will win a hand or a point. However, long term the computer/robot will win out, simply due to the law of averages. So for both events my condition would be winning every match, not winning every point. That I could see myself betting on.

In chess, the element of chance is the human. The decisions he makes at the board are influenced by so many non-chess related factors. It's free will, if you believe in it. Chess programs must have randomness put in as well, to choose between 2 moves with identical scores in the evaluation function. Otherwise every game between 2 programs running the same software would always make the same moves. In theory a human can still defeat a computer, simply by playing better moves than the computer (Such moves exist. Otherwise the programs would NEVER lose, assuming chess is not ultimately a theoretical win for white/black.). However the odds of this have become so small it is not worth trying. It is entirely conceivable that the same will happen in table tennis.

The fan favorites are the ones that show the greatest "style" as you define it, but unfortunately subjective scoring is not likely going to ever happen (Sexist aside of the day: ever notice that sports involving judging are the ones that are most popular with women?). The only exception is the brilliancy prize. Tournaments should offer more brilliancy prizes.

Bottom line, I think we can both agree for the average person, chess between two persons still meets the requirements of a sport (set of rules, winners/losers, element of chance (in this case the free will of the person)), and that for the majority of people it is the sporting aspect that will encourage them to follow events and purchase books on matches and tournaments. And that is why matches between computers would not sell.

One thing I agree wholeheartedly with you about: Tournaments should most definitely award more brilliancy prizes. If we want to increase new membership AND retain existing membership in organized chess, reward creativity as much as you reward winning.

In an earlier post, you mentioned the Fischer - Spassky 1972 match and the effect it had on chess at the time. Imagine for a moment that it were not chess, but rather checkers that Fischer and Spassky were contesting for the world title. Imagine that chess did not exist at all, and so checkers held sway as the greatest mental game and so the Fischer - Spassky checkers title match received just as much attention as the real chess one did.

Now imagine that between 1972 and now, chess was invented, and checkers was not only solved, but was held in derision as a mere footstool to chess, if you will. In this scenario, the Fischer - Spassky checkers match would be laughed about as an anachronism, a sign of how stupid we all were back then.

Who is to say that chess isn't really our version of checkers from the scenario I just described? In other words, there is some other game out there waiting to be invented that would make chess look absolutely juvenile by comparison? And that all the problems elite chess has today -- excessive number of draws, humans vastly inferior to computers, no spectator appeal aside from the small niche group of chess players -- aren't in fact signs telling us that chess isn't 'it'? Chess will become solved (within 20 years if engine development continues at its former pace, that is my prediction, and since the search tree is so large, my definition of 'solved' is that no top computer engine will be able to defeat another top computer engine under normal tournament game conditions -- all such games will be draws. However, I tend to think engine development is about to die out because programmers are smart enough to realize that as an exercise in AI, chess has become a dead end.)

Chess in its essence isn't a sport, it is a puzzle... to be solved. Just as checkers was solved. People still play checkers for pleasure, and maybe somewhere in the world, there are still checkers tournaments. Thus it will be for chess. Anyone who wants to can insist chess is a sport, and can point to 'human free will' as the chance element that makes it a sport. But as large as that search tree is, it is FINITE. And so mathematically, chess is a deterministic game whose outcome with best play on every single ply must be one of only 3 possibilities and it must be eventually proven which of the 3 it is. What the 'human free will' is essentially doing is traversing all the different paths through the maze. We call it 'competition', but really it is 'exploration'. It is just us being stupid, as I showed in the checkers scenario.

[An aside: in the original Star Trek series, there was the famous 3D chess being played, with 3 levels of 8x8 boards and pieces could move from level to level. Perhaps that was the series creator trying to tell us even back then that 2D chess was essentially.... checkers.]

Whereas a game like hockey, a true sport, is absolutely non-deterministic. Good luck proving that there is a finite number of possible paths (not outcomes -- paths) that a hockey game can take from the opening faceoff.

Even poker is non-deterministic, if we take the typical 9-handed Texas Hold 'Em table as the starting point. Again, good luck if you try and prove mathematically that there is some finite number of paths that can be taken from such a starting point, even given some standard (constant) betting stucture and allowing that the levels change not based on time elapsed (as is actually the case), but on some number of hands completed.

Deterministic versus non-deterministic: to my mind, this is the line that disambiguates puzzle from sport. You certainly can disguise a puzzle as a sport, as is being done with chess. But eventually it dawns on everyone that... we're just solving a puzzle.

If you really want to wrap your head around this, set up a random chess position with maybe 24 total pieces on the board, 12 White and 12 Black. If it's not a legal chess position, make some adjustments one piece at a time until it is. Now, from that position, try and determine what legal chess moves that aren't absolutely ridiculuous could lead to that position.

I have tried this with many such random positions, generated in a provably random fashion. And the fact is, with most such positions, there is no conceivable and reasonable line of moves that could lead to that particular position. At some point, you reach a situation where only a series of hideous blunders and oversights could lead to the position, and often not even that -- i.e. it just seems that no legal set of moves could lead to the position.

What that indicates is that there are legal chess positions that can seemingly can NEVER be reached by any legal set of chess moves. For now, this is a hypothesis, not mathematically proven for any given random position P.

But all it takes to 'prove' it is to set up the standard opening chess position... then 'move' White's Queen to d3. Without moving any White pawns. Legal chess position, if you ignore any question as to how it was arrived at. But never reachable by any sequence of legal chess moves from the standard opening position.

I'll leave it for you to decipher in your own way what this tells YOU about chess. You could start by asking... "Standard opening position? Huh? Why do we need THAT???"


By the way... that video of the table tennis robot. I got to thinking about that. The robot was somehow anchored to the floor. I noticed that several of the highlights show the human player diving to one side to hit a return, only to be out of position to get to the next shot. It makes me wonder, what is the 'wingspan' of the robot? Is it comparable to the human's? I almost think it must be a lot larger than the human's. For it to be stationary and hit so many returns -- against the World #1 player mind you -- it must have a tremendous reach. Possibly giving it a huge unfair advantage. A proper robot would be mounted on legs and would have to move side to side just like the human.

And one last point: every table tennis point is a game unto itelf. In chess analogy, every shot is a move, every return of that shot is a counter move. Therefore you cannot bet on the 'long run' of a match. You must bet on every single point just as you would bet on an individual chess game between human and computer engine. If the computer can win every single such chess match, the robot must likewise win every single table tennis point.

However, table tennis has no analogy to draws. Thus we simply remove chess draws from the results as if they never happened. Only decisive games count. And when was the last time any human defeated Houdini under tournament conditions?