FREE - In Google Play
FREE - in Win Phone Store
I was curious if anyone had any good insight into how improving engines have maybe caused there to be changes to opening book theory.
It would stand to reason that as the ability of engines to see more deeply into move combinations, that either different openings would have been solved or perhaps a variation once thought to be not as strong as another, may have been proved to actually be better. I welcome any comments or opinions on this subject, as I know next to nothing about it.
So engines are interesting in relation to the opening. In fact, engines are useful to watch out for traps and tactics but are not considered up to snuff compared to a veteran GM. In engine vs. engine tournaments, their opening books are chosen in collaboration with Grandmasters who work a lot on opening theory because engines can't "see" well into strategic ideas until the position is quite established.
If you picked out a random Benoni line with an even-ish evaluation by GMs, you'd often find most engines will evaluate it as strongly in White's favor for as many as the first dozen moves, sometimes more. Engines often also like doing things like trading off the Dragon Bishop in the Sicilian Dragon for the sake of a pawn, though this idea typically is terrible strategically.
They had a tremendous impact on professional chess : most novelties these days are computer-generated. All of them are computer-checked.
The reason I brought this up is because, I have wondered if maybe there are a few openings that have moves that were maybe on the bubble , as far as being better than others.
I thought maybe the best GM's and thus programmers, chose particular moves over others because, it was easier to exploit certains lines over others, and with the advent of better programs such as houdini, certain lines that were typically pruned before were perhaps clarified.
I realize that this,in and of itself, most likely wouldn't change the opening anywhere near as much as the middle game, but considering my head starts to spin when trying to imagine, have programmers crunched every variation, for every opening? My first guess is yes, but I can see where they have probably spent far more time on the main stays of chess, with lessor machines and programs.
One reason I inquire about this, other than wanting to know if computers have reproved GM's prior theories is, I used to have a Kasparov program for a cell phone. If I tried any king or queen pawn opening on the hardest level, I almost always lost, but if I played certain other openings, I could actually win or in one case with a variation of the English, I could wipe the floor with it. I realize the English has been well explored by computers, but I don't know where to find the latest theory news or ideas for more obscure openings.
I noticed no one has seemed to delve into this idea too much lately. I had an idea I thought would stir things up. I have a Chessmaster program I have been using for some experiments.I realize it isn't the cutting edge in engines, but I found you can still do some interesting experiments that, probably are still rather accurate in overall assessment. I was curious as to what it would assess as the best opening for white to play, giving it the best chance to win versus all of the other possible openings. This idea is probably far less calculable without engines.
So I set up the program to play itself in the infinite mode. After a few hours no first move. So I click the force option. Once the first move was selected, it quickly realed off a succession of moves. That tells me it has the ability to weigh its chances with each opening. It just takes forever to calculate the possibilities.
I used the take back move feature to clear all of the moves, to set this game scenario up again. I was curious as to whether it would quickly put the same moves back it had previously selected. I was curious if it was still able to calculate the line from the point it had left off at, or if it needed to start over. I noticed an interesting phenomenon, it did start the calculation process over. It selected the King's Indian after the initial force, but after the take back, it chose 1. e4.
I let each of these play until mate. White won soundly with the King's Indian, but lost a close King's Pawn Game. I have wondered if Houdini or Rybka gets similar results. If this is accurate, we could use this to make a list of openings in which white can't be beaten if the strongest moves are played. I realize it would be a daunting task to check all of the possible variations to make sure this theory holds true, due to the propensity of engines seemingly miscalculate.
I am curious as to anyone else's ideas or experiments as it pertains to this ideology. For example has anyone set up their newer engine to play itself without a timer to see what opening it chooses and how quickly?
The reason is, I wonder if programmers just put in a program randomizer(using a finite set of predetermined relevant moves), for the computer to pick opening moves and defenses. I suspect this is also possible because of Chessmaster's multiple reactions with black, to the same opening move on multiple occasions.
Please, I implore you, let us reason together. Johnny 5 needs input...
Some openings that were hardly played years ago are popular or at least not uncommon now. The Scandinavian was regarded as unsound until Anand unleashed it against Kasparov in 2000..guess where he got his ideas from? The Rubinstein/Burn variation of the French Defence was hardly ever played at the top level until recently, purely because of new ideas found by computer analysis.
Tactical minefields like the Botvinnik and Moscow variations in the Semi-Slav are perfect for computer analysis so thats why the Slav is now much more common than the Orthodox defence to the Queens Gambit. The Najdorf and Dragon Sicilians also lend themselves to a brute force calculating approach. What is interesting is how someone like Morozevich has resurrected the Budapest Gambit, The Albin Counter Gambit and the Chigorin Defence in the Queens Gambit based entirely on computer aided analysis, that is the good side of using a chess engine to find novelties.
nameno1had, I see you've put a lot of effort into your post, but you should know that your suggestion that because chessmaster beat itself using the king's indian and drew with the king's pawn opening must mean that the king's indian is superior is absololute nonsense.
"we should try to make a list of openings where white can't be beaten when the strongest moves are made"....
I don't necessarily have a problem with Chessmaster's behavior or my assessment of it being inaccurate but, why do you say that? I have no problem with people make claims, even outlandish ones but, atleast give some evidence to back it up...
My example for how one opening could be counted as superior another is simple. If an engine plays(against itself) all of the best moves of an opening line based on strength(I realize that this is perhaps the controversial, due to the way the programs choose moves) and there is a clear material/positional difference, in terms of, things like fewer moves needed to win, less vulnerability, more options to win, etc, can make one opening seem superior to another.
I realize the human element changes this for most of us, but GM's play more like a computer (75-80% of their moves), so therefore, the opening itself could be considered better if it gives any noteable advantage.
It didn't draw itself with the King's pawn game, it lost with white. Are you suggesting it perhaps by random proposition( a programming feature) picked some lines in which white so happened to lose for my viewing pleasure, instead actually trying to do what I asked it to ....win?
nameno1had, where did you get the idea that a chess engine could be able to find the best move in any position?
Well, generally speaking, I never seem to be able to find better moves than it suggests. A few times I have wanted to play a different move than it suggests is best, but I think in those instances the result was going to be relatively the same.
I believe what I do based on the criteria with which it is told to judge what the best moves are. The engines will generally out perform any human in my experience. It does this by calculating (reasonable speaking) all of the possible moves and chooses the line that gives it the best material/positional/time advantage.
If you question the abilities of the best engines, is this because you or someone you know regularly out perform them? If so, please feel free to share...
No one regularly outperforms engines, that isn't the point. You don't have to be better than them overall to be able to see their weaknesses. It really means nothing that an engine beat itself once in an opening. You can't draw conclusions based on one game and you since engines consistently misevaluate positions in some openings like the King's Indian you couldn't draw conclusions from even 1,000,000 games.
Here's a very simple argument that engines don't play best chess.
Engines beat eachother, therefore engines don't play best chess.
It's not perfect but you will hopefully get what I mean.
nameof1had, I do not doubt the ability of chess engines to find excellent moves in most positions. They just do not find the best moves everytime, obviously (your argument seems to be based on that assumption).
Btw: are you aware of the fact that chess programs usually use opening libraries, because their opening play is known to be bad?
They beat each other, but if we can't beat them, doesn't that give what I am saying some relevance? If that doesn't, consider the following idea to better clarify what I was intending for my words to mean.
I wasn't assessing the engines performance for the sake of do they necessarily make the best moves, in terms of, whether the ones they are currently making, are the best possible that could ever be made, but instead was considering them the best possible we know of. They might not be the best possible, but still the best currently.
Its like comparing two athletes of slightly different ability, from different eras and calling each each one the best respectively, though there is a clear difference in their ability/performance. They each in their own way could be the best.
My post # 18 will clarify what I was meaning.
I will agree, when it comes to strategy, engines probably will only reflect the best ideas their programmers where able to try to instill in them. The trouble is, the rules for formulating strategy aren't like a set of absolute rules, that are without exception.
I see. But nevertheless, the best human chess players are still able to beat engines in correspondence chess, due to their superior strategical insight. In opening play, strategical insight is crucial - that's why chess engines use opening books created by humans.
So I'm afraid that your approach won't create advancements in opening theory - not with the engines available now or in the near future.
I guess will have to agree to disagree on this one. The only proof I have to offer is the last time the best player in the world played a computer, though he was able to win one game out of 6, he was thoroughly humbled. Need I add that both Rybka and Houdini are vastly superior to Deep Blue? Have you noticed that no GM has recently came forward to challenge Houdini to show their flaws? I watched a GM thoroughly demostrate, in a video on youtube,the best ways to try to beat engines. He said 10 years ago you had a chance, but the chance is literally all but gone with the latest advancements. If you wont take my word for it, maybe you will trust a GM's opinion...
1. I was talking correspondence chess.
2. If you are right, why do engines use opening books?
3. If you are right, why do Centaurs beat engines?
I really don't think the time mode for standard chess that is used would matter,humans won't out think a computer. Should it be held against the program as a flaw, because opening book theory was developed before the programs? In other words, do you really think that the programs calculating ability wouldn't allow to figure out what would be the opening book equalivalent?
As far as the Centaurs go, you are really only demonstrating the actions that would be used to make the latest cutting edge program be the best or to discover the best possible lines known. I don't recall saying that the programs weren't flawed. Therefore, it stands to reason that there could be glitches in the programs, that until worked out, otherwise would allow an opportunity for a human to beat one.
Lets put this in perspective though, If I took the best program available and I'll give you your pick, Anand/Carlsen, who ever you want, would you bet all that you have on one game between the human or the engine? I'll take the engine. I would be willing to bet that the engine will win 95% of the time or more against the best GM's. If indeed that is the percentage of the time the best GM's are able to win, wouldn't it be a sound system to use the lines and moves that Houdini suggests as the best? If you won't that tells me that this has something more to do with human pride more than your desire to find the best way to figure out what the strongest openings,moves,etc are...
Food for thought....
If I had to bet on the outcome of a game Kramnik vs. engine with a time limit of 14 days per move, I'd choose Kramnik without hesitation.
If you want to play the opening lines your engine suggests, I wish you good luck.
Edit: I don't care about human pride. I am quite sure that in a hundred years, chess engines will be vastly superior to anything humans could come up with. But not yet.