does a/o here know exactly how stockfish, AZ, leela, etc. comes to choose its move ? and also any comments on how NN'ing/AI & adv memory & stronger processors will change things.
also do ppl feel that a unbeatable engine solves chess ?...or does it need to win every game ? iows plz define 'solved'. and is this def agreed upon. thx...i hope.
btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd.
does a/o here know exactly how stockfish, AZ, leela, etc. comes to choose its move ? and also any comments on how NN'ing/AI & adv memory & stronger processors will change things.
also do ppl feel that a unbeatable engine solves chess ?...or does it need to win every game ? iows plz define 'solved'. and is this def agreed upon. thx...i hope.
btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd.
Im going to respond to this first because there are multiple people here who will spread misinformation. (optimissed, tygxc)
does a/o here know exactly how stockfish, AZ, leela, etc. comes to choose its move ? and also any comments on how NN'ing/AI & adv memory & stronger processors will change things.
also do ppl feel that a unbeatable engine solves chess ?...or does it need to win every game ? iows plz define 'solved'. and is this def agreed upon. thx...i hope.
btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd.
Im going to respond to this first because there are multiple people here who will spread misinformation. (optimissed, tygxc)
That means giving opinions you disagree with because you're uninformed, doesn't it?
"Elroch is trolling with his suggestion that the first moves of all 108 games are wrong and that 1 a4 wins for white. Demanding game trees for 1 a4 or 1 e4 e5 2 Ba6? is not rigor, it is stupid"
not trolling at all, you just have no idea what demanding mathematical rigor is.
"it is stupid" isnt an argument, sorry. It is rigor literally by definition. if you had taken a mathematical proof class you would have known this.
why do you continue to ignore the many objective lies of yours that I point out?
@tygxc needs to tell the authors of Stockfish to stop it from wasting time analysing variations after 2. Ba6. At present it shows that it is a weaker player than him by keeping analysing deeper and never reaching an evaluation that indicates certainty about the result.
My conclusion about unassisted @tygxc's chess strength can be verified by a 149 game match against Stockfish where I predict a victory by 1 point due to his superior ability to fail to analyse variations.
If the engine is set for a strong solution after 1. e4 e5 2. Ba6, it's never going to finish the analysis in a million years, which is rather convenient! 
Because, of course, you know I can out-argue you easily and that I also understand more about the subject than you. Otherwise, you'd treat me as you do tygxc. He understands more than you also. Clearly the monkey understands nothing but sometimes pretends he does and rambles on infinitely, displaying his existential angst for all to applaud. The hamster understands more but he hasn't a logical mind and doesn't do logical arguments. Never seen it, anyway. His strategy is to stop before he's made his killing point, because he underestimates our ability to understand an argument from zero valid premises. Or maybe overestimates our ability to join him in constructing one. Could be either or both.
I suppose you must be the mouse, considering your name here.
LMFAO LOL.
I had suspected that you were pissy about me ignoring you, but this confirms it. I'm going to go back to ignoring you with impunity now .
I actually read many of your posts, but theres literally zero substance for me to consider them worth responding. This one I chose to, not because of any substance, but because i found it so funny how easily you revealed the cards in your hand.
You've taken hold of quite the wrong end of the stick, mousey. I know you're going to ignore me because some of my posts are very much on topic and you don't have the ability to answer them well. All I've been doing is finding that out. I like to know whom I'm talking with and what motivates them, since there are lots of people who pretend and very few if any who genuinely attempt to communicate honestly.
You're all mouth and no content.
You've taken hold of quite the wrong end of the stick, mousey. I know you're going to ignore me because some of my posts are very much on topic and you don't have the ability to answer them well. All I've been doing is finding that out. I like to know whom I'm talking with and what motivates them, since there are lots of people who pretend and very few if any who genuinely attempt to communicate honestly.
You're all mouth and no content.
...he said, posting no content.
Maybe you could edify us all about how opening theory and endgame tablebases are "crystalized structures"...you know, try to turn that into some kind of real content instead of vague made-up fluff you tossed out when challenged about your false claims.
There is actually no such thing as winning an argument. It's not a game like chess, there is no independent objective arbiter.
What @Optimissed means when he praises his own arguing skills (I noticed he did not qualify this with any reference to being right) is that he can in all circumstances convince himself of his superiority over everyone else in every way that matters to him. I am not sure what he thinks about convincing other people, but he often makes bold proclamations about what other people think that are a very long way from reality.
Some people you can never convince in an argument. They are utterly invulnerable. Eg flat earthers. But essentially anyone whose ideas are not subject to any valid testing.
Others, you can never even get to understand what the argument is about.
Fact: Stockfish has never won the World Computer Chess Championship.
Komodo has won it most recent years, but in last year's competition Stoofvlees beat it.
A bit deceptive, since TCEC is generally regarded as the "real" WCCC, and Stockfish has done well in it. The last four finals have seen it against LCZero. It only won by 52:48 in seasons 24 and 25, but managed 57:43 in season 26. Leela picked up 17 wins though.
Dumme luete.
This is like someone posting "your stupid"...if you are trying to call someone/something dumb, but you misspell your own insult, well...
Atleast i have a life!
Yes everyone here lives
We all know you have a life thank you for your helpful input lol
Atleast i have a life!
1. You misspelled 'At least'.
2. I think that everyone reading these forums has a life.
3. Get better at chess before calling someone stupid, and I actually agree with Elroch, and if you do not, that does not mean that you can call him stupid.
4. Bye.
You've taken hold of quite the wrong end of the stick, mousey. I know you're going to ignore me because some of my posts are very much on topic and you don't have the ability to answer them well. All I've been doing is finding that out. I like to know whom I'm talking with and what motivates them, since there are lots of people who pretend and very few if any who genuinely attempt to communicate honestly.
You're all mouth and no content.
...he said, posting no content.
Maybe you could edify us all about how opening theory and endgame tablebases are "crystalized structures"...you know, try to turn that into some kind of real content instead of vague made-up fluff you tossed out when challenged about your false claims.
hi hamster. Your cheeks are a bit puffed out. You're eating the content.
Since Mega is now identified as a troll, that's relevant content. Means he's in the same boat as you are. This is getting like Noah's Ark.
First, nobody here knows EXACTLY how the top chess engines are coded. if we did, then we would be out there coding the engines instead of wasting time on this forum. but I can probably give you a general idea. My strength is in mathematical analysis, so I might be a little off/vague in the description of the chess engines.
in general, a chess engine works by:
1. Choosing which paths down the game tree to look, and how deep to look.
2. Doing a "small" calculation with each individual position looked at (these are often formally referred to as "nodes")
3. Compiling each of the "small" calculation values into "large" calculations
4. Choosing the best move based on the results of the "large" calculations.
Engines will often do these steps multiple times, in a variety of orders. What differentiates different engines is: the order and computing power that is put into each step, the type of "evaluation"/"calculation" used, and how the 'calculator' was trained/created.
"and also any comments on how NN'ing/AI & adv memory & stronger processors will change things." AI and NN'ing allows for greater/more effective processing on the individual node calculations and a better evaluation of those values. Engines like Leela use less nodes/look at less positions than a general engine, but they have a much larger and complex system of evaluation. This evaluation system has been trained on neural networks extensively. Stronger processors allow us to host larger evaluation matrices, in addition to allowing us to calculate over a much larger pool of nodes.
"also do ppl feel that a unbeatable engine solves chess ?...or does it need to win every game ? iows plz define 'solved'. and is this def agreed upon. thx...i hope."
By game theory, there are multiple ways to "solve" a game like chess. Wikipedia has great base explanations, so I'll borrow those, and add notes on the current state of this 'discussion' and chess with regards to those definitions. There are three ways to consider a game "solved" in game theory.
Strong Solution: "Provide an algorithm that can produce perfect play for both players from any position, even if imperfect play has already occurred on one or both sides." it is universally agreed upon (even by the misinformation spreaders) that humanity is nowhere close to strongly solving chess with current technology and chess knowledge. An 'algorithm' could either be the game tree in its entirety, a shortcut strategy with an accompanying mathematical proof that the strategy leads to perfect play, or a combination of the two.
Weak Solution: "Provide an algorithm that secures a win for one player, or a draw for either, against any possible play by the opponent, from the beginning of the game." This is the main type to be considered in a conversation of "solving" chess. The current consensus is that humanity will likely achieve a weak solution some time in the future, but that it is beyond current technology. tygxc claims that humanity is very close to a weak solution, however his claims rely on faulty calculations that have been shown to be off by a factor of over 1 million.
Ultra-Weak Solution: "Prove whether the first player will win, lose or draw from the initial position, given perfect play on both side" - Note that we do not have to know what perfect play looks like to Ultra-Weakly solve a game. A great example of this is Chomp. Tygxc has claimed that chess is ultra-weakly solved (he will sometimes add "for all practical purposes" to cover his tracks, but the phrase "for all practical purposes" is an oxymoron in mathematical analysis and game theory, a solution is either rigorously proved or it isn't). There have been many flaws pointed out in tygxc's 'proofs' of these claims by many people.
On the "Unbeatable engine" - Something important to consider is that an algorithm which guarantees white at least a draw against black no matter what black plays isn't a Weak solution. Such an algorithm does not prove that white cannot win from the starting position with better play.
However, an algorithm that forces at least a draw for one side would still be a massive accomplishment, and could be referred to (casually) as a "solve" of chess.
"btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd."
Simple. The 7-table is calculated off of the entire game tree of each position in the table. it is a "Strong Solution" of that position, acquired from many years of work. Current chess engines are not able to calculate every single possible position at that depth, so they can occasionally miss sequences of moves that could change the game.
hi hamster. Your cheeks are a bit puffed out. You're eating the content.
Since Mega is now identified as a troll, that's relevant content. Means he's in the same boat as you are. This is getting like Noah's Ark.
Who you decide is a troll is utterly worthless information. You couldn't butter the right side of a crumpet, much less make reasonable judgments about the world around you.
Regarding the way engines work, Leela uses a fairly large NN (about 64 million parameters) and needs a GPU to run efficiently (the same type of hardware used in most modern AI, because they use quite large neutral networks that need to do the sort of large matrix calculations where GPUs are way faster than CPUs.
Stockfish has used a smaller neural network for positional evaluation since a few years back (introducing it resulted in more than 100 points of rating jump), and runs ok on a CPU alone. The first layer of the network is medium sized (about 10 million parameters) but later ones are much smaller (< 20k). Leela chess is heavily influenced by AlphaZero's architecture and has 20 to 80 layers with 128 or 256 outputs each. So it is deep but rather narrow. I have seen an estimate that it has 65,000,000 trainable parameters in total.
Amazingly, Stockfish manages 100 million nodes per second using alpha-beta tree search, while Leela manages with 40,000 (2,500 times fewer!) using Monte Carlo tree search. The first is like trying as hard as possible to find the best move by thoroughly analysing and then assuming the players play optimally, while the latter is like playing lots of somewhat random games starting with candidate moves, and using the results to come to a conclusion which move is best. Or something approximating those.
thank you Big Cheeze for #11815. very thoughtful . L♥
First, nobody here knows EXACTLY how the top chess engines are coded. if we did, then we would be out there coding the engines instead of wasting time on this forum. but I can probably give you a general idea. My strength is in mathematical analysis, so I might be a little off/vague in the description of the chess engines.
in general, a chess engine works by:
1. Choosing which paths down the game tree to look, and how deep to look.
2. Doing a "small" calculation with each individual position looked at (these are often formally referred to as "nodes")
3. Compiling each of the "small" calculation values into "large" calculations
4. Choosing the best move based on the results of the "large" calculations.
Engines will often do these steps multiple times, in a variety of orders. What differentiates different engines is: the order and computing power that is put into each step, the type of "evaluation"/"calculation" used, and how the 'calculator' was trained/created.
"and also any comments on how NN'ing/AI & adv memory & stronger processors will change things." AI and NN'ing allows for greater/more effective processing on the individual node calculations and a better evaluation of those values. Engines like Leela use less nodes/look at less positions than a general engine, but they have a much larger and complex system of evaluation. This evaluation system has been trained on neural networks extensively. Stronger processors allow us to host larger evaluation matrices, in addition to allowing us to calculate over a much larger pool of nodes.
"also do ppl feel that a unbeatable engine solves chess ?...or does it need to win every game ? iows plz define 'solved'. and is this def agreed upon. thx...i hope."
By game theory, there are multiple ways to "solve" a game like chess. Wikipedia has great base explanations, so I'll borrow those, and add notes on the current state of this 'discussion' and chess with regards to those definitions. There are three ways to consider a game "solved" in game theory.
Strong Solution: "Provide an algorithm that can produce perfect play for both players from any position, even if imperfect play has already occurred on one or both sides." it is universally agreed upon (even by the misinformation spreaders) that humanity is nowhere close to strongly solving chess with current technology and chess knowledge. An 'algorithm' could either be the game tree in its entirety, a shortcut strategy with an accompanying mathematical proof that the strategy leads to perfect play, or a combination of the two.
Weak Solution: "Provide an algorithm that secures a win for one player, or a draw for either, against any possible play by the opponent, from the beginning of the game." This is the main type to be considered in a conversation of "solving" chess. The current consensus is that humanity will likely achieve a weak solution some time in the future, but that it is beyond current technology. tygxc claims that humanity is very close to a weak solution, however his claims rely on faulty calculations that have been shown to be off by a factor of over 1 million.
Ultra-Weak Solution: "Prove whether the first player will win, lose or draw from the initial position, given perfect play on both side" - Note that we do not have to know what perfect play looks like to Ultra-Weakly solve a game. A great example of this is Chomp. Tygxc has claimed that chess is ultra-weakly solved (he will sometimes add "for all practical purposes" to cover his tracks, but the phrase "for all practical purposes" is an oxymoron in mathematical analysis and game theory, a solution is either rigorously proved or it isn't). There have been many flaws pointed out in tygxc's 'proofs' of these claims by many people.
On the "Unbeatable engine" - Something important to consider is that an algorithm which guarantees white at least a draw against black no matter what black plays isn't a Weak solution. Such an algorithm does not prove that white cannot win from the starting position with better play.
However, an algorithm that forces at least a draw for one side would still be a massive accomplishment, and could be referred to (casually) as a "solve" of chess.
"btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd."
Simple. The 7-table is calculated off of the entire game tree of each position in the table. it is a "Strong Solution" of that position, acquired from many years of work. Current chess engines are not able to calculate every single possible position at that depth, so they can occasionally miss sequences of moves that could change the game.
Firstly, here you are not talking like that which you have represented yourself to be. You understand programming and programming logic even if you do not understand people and you cannot make a judgement regarding people and expect to get it right. That's par for the course.
You should have explained the basic differences in positional evals between engines that are considered more "human-like" and more "computer-like". Also basic strategies involved, such as prioritising moves that place pieces near an opponent's king (even if there's no obvious threat) and prioritising searches where multiple pieces can be swapped in various orders, since that kind of thing is comparitively difficult for humans. What you did describe with "small calculations vs large ones" was the creation of a list of candidate moves, beginning with assessments of many possible candidates using a short horizon.
This thread is separated into two factions: basically mathematicians and scientists. Unfortunately the mathematicians have chosen to mock the person who is at the forefront of the scientific approach. Each person contributing seems to have some programming experience. Even Diogenes, who spends his time trying to annoy people and offers no productive input. He has no opinions at all which may be trusted as either honest or well thought through. That, of course, is why he pretends that about me, since we're on opposite sides and he could never hope to win an argument with me, probably on any subject. I'm the most "lay-person-like", discounting playerafar, of course, who has no computing credentials. I have programmed at a level much higher than basic levels but have no experience of programming techniques developed in the past 30 years.
I stand with tygxc because I know that his approach is correct and yours isn't, since it practice it cannot be achieved. I do not spread misinformation. I am, in fact, the person most capable of correctly identifying and explaining your own mistakes. That is why you are pretending that I spread misinformation. Your dismissal of tygxc "due to a factor of 1 million" isn't necessarily a good argument in circumstances where the figures may represent probabilities with factors of trillions greater than one million, so your blanket dismissal is incorrect. Elroch was similarly incorrect when he dismissed my explanation of the odds against 106 draws. When you demand perfection from others and see fit to ignore it for yourself, it doesn't sit well. I won that argument with him because he had no firm theory on which to dismiss it. It means that chess has probably been proven to be drawn with best play. I have come to expect no basic honesty at all, from your side in the discussion.
There are some weaknesses in your description to Lola. What you wrote was quite good but the significant weaknesses it contains include the impossibility of proving that one of the solutions you describe is accurate. You also show a fundamental misunderstanding of the role of mathematics in solving chess. You hide from the fact that chess cannot be represented mathematically with real accuracy. It would always be an approximation and therefore you can't use it, even though you keep talking about it. Unfortunately, you and others don't apply the accuracy you require from the scientific side to your own ideas. It's typical of weak thinkers but it's fairly normal and to be expected.
I don't agree with tygxc that a "weak solution" will be found fairly soon, in the sense that you probably mean. Nevertheless, I dislike the "strong-weak-ultra-weak" jargon because although it's suitable for simple games, it isn't suitable for chess. The definitions you gave are not fit for purpose, in any practical sense. Also the idea that "there are multiple ways to solve a game like chess" is total nonsense. There are no ways as yet. The people who wrote that nonsense need to take account of that before adding to the misinformation which they and you are spreading.
Firstly, here you are not talking like that which you have represented yourself to be. You understand programming and programming logic even if you do not understand people and you cannot make a judgement regarding people and expect to get it right. That's par for the course.
You should have explained the basic differences in positional evals between engines that are considered more "human-like" and more "computer-like".
Also basic strategies involved, such as prioritising moves that place pieces near an opponent's king (even if there's no obvious threat) and prioritising searches where multiple pieces can be swapped in various orders, since that kind of thing is comparitively difficult for humans.
What you did describe with "small calculations vs large ones" was the creation of a list of candidate moves, beginning with assessments of many possible candidates using a short horizon.
This thread is separated into two factions: basically mathematicians and scientists. Unfortunately the mathematicians have chosen to mock the person who is at the forefront of the scientific approach. Each person contributing seems to have some programming experience. Even Diogenes, who spends his time trying to annoy people and offers no productive input. He has no opinions at all which may be trusted as either honest or well thought through. That, of course, is why he pretends that about me, since we're on opposite sides and he could never hope to win an argument with me, probably on any subject. I'm the most "lay-person-like", discounting playerafar, of course, who has no computing credentials. I have programmed at a level much higher than basic levels but have no experience of programming techniques developed in the past 30 years.
I stand with tygxc because I know that his approach is correct and yours isn't, since it practice it cannot be achieved. I do not spread misinformation. I am, in fact, the person most capable of correctly identifying and explaining your own mistakes. That is why you are pretending that I spread misinformation. Your dismissal of tygxc "due to a factor of 1 million" isn't necessarily a good argument in circumstances where the figures may represent probabilities with factors of trillions greater than one million, so your blanket dismissal is incorrect. Elroch was similarly incorrect when he dismissed my explanation of the odds against 106 draws. When you demand perfection from others and see fit to ignore it for yourself, it doesn't sit well. I won that argument with him because he had no firm theory on which to dismiss it. It means that chess has probably been proven to be drawn with best play. I have come to expect no basic honesty at all, from your side in the discussion.
There are some weaknesses in your description to Lola. What you wrote was quite good but the significant weaknesses it contains include the impossibility of proving that one of the solutions you describe is accurate. You also show a fundamental misunderstanding of the role of mathematics in solving chess. You hide from the fact that chess cannot be represented mathematically with real accuracy. It would always be an approximation and therefore you can't use it, even though you keep talking about it. Unfortunately, you and others don't apply the accuracy you require from the scientific side to your own ideas. It's typical of weak thinkers but it's fairly normal and to be expected.
I don't agree with tygxc that a "weak solution" will be found fairly soon, in the sense that you probably mean. Nevertheless, I dislike the "strong-weak-ultra-weak" jargon because although it's suitable for simple games, it isn't suitable for chess. The definitions you gave are not fit for purpose, in any practical sense. Also the idea that "there are multiple ways to solve a game like chess" is total nonsense. There are no ways as yet. The people who wrote that nonsense need to take account of that before adding to the misinformation which they and you are spreading.
Actual points/content bolded. Hard to find. Ironically, what is there...is wrong. What you meant to say was "prioritizing moves that limit a king's options", which can obviously happen from all the way across the board in the case of Q/R/B, but not with K/N/P Your wording of it actually highlights a human failing, that is, seeing closer pieces and their influence exerted more easily than seeing the same pieces exerting the same influence from way across the board, and seeing "attacking moves" (getting closer) much more readily than "retreating moves" that are actually stronger.
Everything else in your post is either completely irrelevant, unilateral declarations with no substance, or weak aspersions, also lacking substance.
...
"btw they say stockfish blundered a known 7-table draw. lost to leela. s/t i find really wierd."
...
It's not weird at all.
Here's SF blundering a winning 3 man KRvK position into a draw against me (from move 35)
And again in a 5 man winning KNNvKP position.
In fact just pick 10 random positions from the last endgame and it will normally blunder all 10.
Because, of course, you know I can out-argue you easily and that I also understand more about the subject than you. Otherwise, you'd treat me as you do tygxc. He understands more than you also. Clearly the monkey understands nothing but sometimes pretends he does and rambles on infinitely, displaying his existential angst for all to applaud. The hamster understands more but he hasn't a logical mind and doesn't do logical arguments. Never seen it, anyway. His strategy is to stop before he's made his killing point, because he underestimates our ability to understand an argument from zero valid premises. Or maybe overestimates our ability to join him in constructing one. Could be either or both.
I suppose you must be the mouse, considering your name here.
LMFAO LOL.
I had suspected that you were pissy about me ignoring you, but this confirms it. I'm going to go back to ignoring you with impunity now
.
I actually read many of your posts, but theres literally zero substance for me to consider them worth responding. This one I chose to, not because of any substance, but because i found it so funny how easily you revealed the cards in your hand.