Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Opening Books / 250 MB - The End Is Near
1 2 Previous Next  
- - By Auryn (***) [it] Date 2015-11-01 21:11 Edited 2015-11-01 21:15
So... my manual book (no imported games in it) in these days celebrates the size of 250 MB.
My latest post was when it reached 200 MB, 2 or 3 years ago, not sure.
I am losing interest in book work, the days when every game will end in a draw is getting closer and closer.
It was hard to realize that my battle to get an advantage for White with 1.e4 is lost. Not only I can't find any advantage anymore against 1...c5 or 1...e5, but even against 1...c6 and 1...e6.
In one way I am happy, I contributed in openings development, and it took me very long to achieve these 50 more MB, as my daily work on book took long pauses; during some months I even stayed 3 or 4 days without touching it.
A desperate 1.d4 try also failed miserably.
I will probably never "celebrate" the 300 MB, as it seems to me meaningless trying so hard anymore when seeing that the openings that easily equalize for Black are so many.
So... this is a sad probably last celebration for my book.
I had fun in all these years, I created the book in the far 2002, but now... the time has come for me to find a new hobby.
I am also stopping playing correspondence chess, as the draws there are close to 100% already.

Eros.
Parent - - By darmar (**) [rs] Date 2015-11-01 22:35
I want play draw vs strong opponents and reach 2400 elo in correspondence chess. Is this easy with your book?
Parent - By Auryn (***) [it] Date 2015-11-02 12:01 Edited 2015-11-02 12:10
Top Engines on fast computers have good drawing chances vs over 2600 correspondence rated players even without using a book, if you give them enough thinking time (the more the better of course).
To raise your elo just play tournaments with players higher rated than you, and by simply making all draws vs them you will raise your elo more and more.
Of course avoid lower rated players, as they will also easily draw and take points away from you.
Parent - - By Graham Banks (*****) [nz] Date 2015-11-02 00:37
Any chance that you'll share your work once you stop?

Graham.
Parent - By Auryn (***) [it] Date 2015-11-02 12:05
Yes, there are some chances, even if it will be quite worthless to have a drawish book.
Parent - - By Ozymandias (****) [es] Date 2015-11-02 09:21

>I am also stopping playing correspondence chess, as the draws there are close to 100% already.


And yet, there's still people out there (talkchess in particular) who will deny the draw death of chess.

How little do they know.
Parent - - By Auryn (***) [it] Date 2015-11-02 12:17
Ignorance is good when it keeps the enthusiasm alive in you.
I envy them.
Parent - By Ozymandias (****) [es] Date 2015-11-02 15:02
ignorance is bliss
Parent - - By Venator (Silver) [nl] Date 2015-11-02 18:13
And yet, there's still people out there (talkchess in particular) who will deny the draw death of chess.

Those who analyse deeper and deeper and think it is best to play 30-40 moves out of book, leaving very few pieces behind on the board, will get this impression. They are dead wrong.

The best chess engines still have no clue how to increase their winning chances. They are excellent in finding draws, not in finding creative ideas that can cause errors in the opponent's play. There can be a huge leap forward when an intelligent engine author finds ways to tackle the problem 'how to play for a win'.

The development of chess theory was interesting, but that has come to a dead end. The main lines are analysed up to a draw, but there still is a huge amount of virgin territory that isn't. Then count the many, many eval errors Komodo and Stockfish are still making and it is 100% sure we are still far, far away from chess being a dead draw.
Parent - - By Ozymandias (****) [es] Date 2015-11-02 19:33

>The best chess engines still have no clue how to increase their winning chances. They are excellent in finding draws, not in finding creative ideas that can cause errors in the opponent's play.


Even if you can find better moves, as a centaur, your opponent, with engine aide (or an engine itself) will still avoid those errors, most of the time. The cracks are selling shut, no matter where you may look for them.

>The development of chess theory was interesting, but that has come to a dead end. The main lines are analysed up to a draw, but there still is a huge amount of virgin territory that isn't. Then count the many, many eval errors Komodo and Stockfish are still making and it is 100% sure we are still far, far away from chess being a dead draw.


I'm afraid the virgin territory hasn't been claimed, because nobody wanted to claim it. As for eval flaws, they're made up for, with increasingly better and deeper search.
Parent - - By Venator (Silver) [nl] Date 2015-11-03 16:12
You are not replying to my main point: that engines are bad at increasing winning chances. So far they have shown that they are excellent in finding and making draws. But there is not one program that is instructed in such a way that it seeks to create better chances to win. On the contrary, they are all doing the same, namely 'stay within a safe draw margin'.

In 2015 we are laughing about the playing strength of engines in 2010. In 2010 we did the same about 2005 and in 2020 we will do the same about 2015. The engine author who is first to make a big jump in tackling the problem 'creating winning chances' will make a big step forward. Searching deeper with the current search behaviour will NOT do the trick.
Parent - - By Ozymandias (****) [es] Date 2015-11-03 18:42

>You are not replying to my main point: that engines are bad at increasing winning chances.


I did, by stating that, in a game that leans towards a draw, being good at doing just that, is the most important thing to do.
Parent - - By Venator (Silver) [nl] Date 2015-11-04 07:00 Edited 2015-11-04 07:03
You based your comment on the fact that top engines lean towards making draws. And as the purpose of the game is 'winning', steering the game towards a draw is NOT the best thing to do. If you have no clue about how to increase your chances of winning (which none of the top engines has any clue about), then you can't say anything about the drawishness of the game of chess.

The mistake you are making, is that you are accepting what the engines are doing NOW, in 2015, is the best thing to do. Without knowing how many great winning chances there are in the rest of the 99.99999999% positions that current engines skip. A simple example is that engines just take the information from endgame bases for granted. There is not one engine having the intelligence trying to figure out if the opponent also has this knowledge and tries for lines in the endgame bases that has the largest possibility of the opponent going wrong.

Sorry, you are dead wrong and 'believing' the current top engines too much. Same mistake they made with Shredder/Fritz, thinking that huge jumps forward were not possible anymore.
Parent - - By Ozymandias (****) [es] Date 2015-11-04 07:56

>You based your comment on the fact that top engines lean towards making draws.


Not solely on that fact. I also see what happens when you analyze an opening line, and what others find when doing the same. I may not be a good chess player, so I might be inclined to think that I "have no clue about how to increase your chances of winning", but then I beat better OTB players than me, at Freestyle, and reconsider. Furthermore, both Eros and O'Neill say the same thing, and I trust their analysis, so there's also that.

>A simple example is that engines just take the information from endgame bases for granted. There is not one engine having the intelligence trying to figure out if the opponent also has this knowledge and tries for lines in the endgame bases that has the largest possibility of the opponent going wrong.


Hoping for your opponent to go wrong is a way of winning, but it doesn't change the fact that, if played correctly, the game is a draw. We're still in a place where finding the correct moves is sometimes difficult, but it's getting easier.

>Sorry, you are dead wrong and 'believing' the current top engines too much. Same mistake they made with Shredder/Fritz, thinking that huge jumps forward were not possible anymore.


When I began playing Freestyle, we already had Rybka and Zappa, which were much better than Shredder/Fritz, and yet, I could clearly see big holes in their play. I don't see them anymore, not since the Stockfish/Komodo race started.
Parent - - By Venator (Silver) [nl] Date 2015-11-04 08:22
Hoping for your opponent to go wrong is a way of winning,

I am not talking about 'hoping'.

both Eros and O'Neill say the same thing

That is because they use today's engines, that will not tell them how to increase winning chances. And playing as many theory moves as possible is not a way to increase winning chances.

I could clearly see big holes in their play. I don't see them anymore, not since the Stockfish/Komodo race started.

And here you are wrong. There are huge holes in their play. They don't know anything about the concept 'increase your chance of winning' and have many wrong evaluations, still. These are only two examples.

As I said earlier, in 2020 we will be laughing about Stockfish 6 and Komodo 9.2 concerning playing strength. Same in 2025 about the programs in 2020.
Parent - - By Ozymandias (****) [es] Date 2015-11-04 08:27

>That is because they use today's engines


And you aren't?
Parent - - By Venator (Silver) [nl] Date 2015-11-04 08:39
Sure (except I don't have Komodo). The difference is, though, that unlike you I don't believe they are telling me the ultimate truth in chess.

If Stockfish 16 in 2025 is beating Stockfish 6 by 75-80%, it is clear that Stockfish 6 isn't perfect in playing chess.
Parent - - By Ozymandias (****) [es] Date 2015-11-04 08:42
I don't believe they are telling me the ultimate truth in chess, neither does O'Neill and I'd be surprised to read Eros writing something along those lines.
Parent - By Venator (Silver) [nl] Date 2015-11-04 08:48
They give up because they think chess is a draw. This is based on analysis using today's engines/analysis. This analysis is not the ultimate truth.

The next leap forward in computer chess will be the point where an engine author manages to tell its engine how to maximise winning chances without increasing risk of losing. That's still quite a bit away from now, is my estimation.
Parent - - By Morpheus (****) [pl] Date 2015-11-02 15:16 Edited 2015-11-02 20:33
It's sad news. So many years and STOP playing now ? Try to find new view.
Have a break and come back stronger! :wink:
Parent - By Auryn (***) [it] Date 2015-11-02 16:40
It's sad in one way, but in another way it's good that I am finally dedicating less time to this hobby.
Parent - - By Fulcrum2000 (****) [nl] Date 2015-11-02 16:43
Hi Eros, congratulations on the hand-tuned 250MB book. Must be one of the best, and very possibly, THE best book out there!.

Sorry to hear you are loosing interest, but I can sort of relate to it. Spending many hours per week on a book, only to find out everything you try is refutable, is kind of discouraging. Having played computer-chess for quite a number of years myself, I'm getting the feeling current chess is to 'simple' for the top engines we have at our disposal nowadays. Sometimes I wonder what will happen if we all agree to move from a 8x8 board to a 10x10 board. The whole revolution would start again, book making, tablebase generation, chess program developments etc. But then again, I'm getting too old to do the whole cycle again ;-)
Parent - - By Auryn (***) [it] Date 2015-11-02 16:58
That's evolution, Werner :-)
Some would say things were better when things were worse, and I may agree with them. I recall with nostalgy the old days when deepfritz and shredder were fighting on dual cores, and opening books were still so behind in knowledge... Back then the hours of book work were really productive, and the fun and satisfaction much bigger.
Like in all things, when they are "investigated" too far, they lose interest to our eyes.
Anyway maybe I was too pessimistic, there is still some room for improvement, even if it's harder and harder, and we know that the day when over 90% of the games will end up in draws is not too far.
I think I wouldn't like a new version of chess, like I don't like chess 960.
Anyway it might be interesting to make the board bigger and create one or two new pieces.
Parent - By Ozymandias (****) [es] Date 2015-11-02 20:00

>I think I wouldn't like a new version of chess, like I don't like chess 960.


Me neither, the only one that's caught my eye, is Fics wild style 0 (White has the typical set-up at the start.  Black's pieces are the
same, except that the King and Queen are reversed). Arno seemed to like it, too.
Parent - - By Venator (Silver) [nl] Date 2015-11-02 18:05
I'm getting the feeling current chess is to 'simple' for the top engines we have at our disposal nowadays.

I completely disagree. There are still many, many evaluation errors made by the top engines. And there are still quite a lot of positions they don't understand, where a 2200 elo human player easily sees through.

On top of that, one of the bad character traits of modern engines is, that they completely fail in creating the best winning chances. They are great when it comes to finding drawing lines, but they are bad in increasing their chances to win.
Parent - - By The Wizard (***) Date 2015-11-10 09:01
Hi,
     Sadly I came to the same conclusion as Auryn a few years ago and although I continue to grow my database more out of habit than enjoyment I have gone back to playing 5 minute blitz in the cafe on chessbase with all the bad moves and blunders lol which you find when going through the games with an engine but you still get to try opening ideas you come up with against human players which may or not work :grin:

Regards
Parent - - By Venator (Silver) [nl] Date 2015-11-10 15:05 Edited 2015-11-10 15:08
Hi,

You didn't address the point I was making: current top engines strive for draws. They don't have a clue how to maximize winning chances. As soon as we have an engine author succesfully addressing this issue, computer chess will become interesting again.

Elsewhere Peter Grayson has played a match between Komodo and Stockfish using my sharp gambit testsuite. Even from these wild positions they were turning most games quickly into a draw. This is a dead end. The current top programs fail to tackle the number one problem of chess: how do I create the best winning chances. They explore just 0.00000000001% of the game tree, exchange pieces, exchange pawns, do not create meaningful chances for themselves. They are like two equally strong boxers dancing around each other, giving a careful punch only once every five minutes. That's what they are doing.
Parent - - By The Wizard (***) Date 2015-11-11 09:16

>You didn't address the point I was making: current top engines strive for draws


I would suggest if that is true it is not the engine but the programmer? of course you can change the contempt and piece value (I think Larry has the piece value covered) ... I that get engines sometimes dont understand positions at times that involve manoeuvres but maximising winning chances I am not so sure about? it may also mean maximising losing chances :-)  .... 

> Even from these wild positions they were turning most games quickly into a draw ... I have not checked the games but maybe the alternative was a loss


Regards
Parent - - By Ozymandias (****) [es] Date 2015-11-11 11:51

>but maximising winning chances I am not so sure about? it may also mean maximizing losing chances :-)  …. 


That's what has been shown so far, expecting this to change in the future seems far fetched.
Parent - - By Venator (Silver) [nl] Date 2015-11-13 16:50
So you think that the current level of chess engines cannot be improved much further?

to change in the future seems far fetched.

Electric cars replacing cars running on gas also seemed very far fetched 7 years ago. But it is happening.

You forget that the only constant in life is that it is changing continuously. In 2025 you will be amazed about your own remarks here.
Parent - - By Ozymandias (****) [es] Date 2015-11-13 18:39

>So you think that the current level of chess engines cannot be improved much further?


The current level chess engines are showing at LTC can't be improved much further, because we're already quite close to the point where they draw any position not currently covered by opening theory. That's not to say they can't improve, by finding the moves more quickly, for example. In any case, remember I'm talking about computer chess reaching the draw death soon, not just engines. Whether they're capable of handling the black side of the poisoned pawn on their own, or in need of TB help to win a position like the one SF recently missed, in the TCEC final, I really don't care.
Parent - - By Venator (Silver) [nl] Date 2015-11-13 19:15
The current level chess engines are showing at LTC can't be improved much further, because we're already quite close to the point where they draw any position not currently covered by opening theory.

I totally disagree with you on this one, because of what I have said many times before now.

Put your remarks in this thread in your digital 2025 agenda and then let's discuss it again.
Parent - By Ozymandias (****) [es] Date 2015-11-13 20:26
No problem at all, with that.
Parent - - By Venator (Silver) [nl] Date 2015-11-13 16:47
but maximising winning chances I am not so sure about? it may also mean maximising losing chances :-)  ....

Not if you use logical principles :-). The purpose of the game is winning, so you should make life as difficult as possible for your opponent. Exchanging pieces and pawns and staying close to a 0.00 score without knowing how to create these winning chances, is not the way to go.

Besides, if 'not losing' is becoming the main driver in a game of chess, you ask for many boring games and many draws, just as is happening in the TCEC 8 final right now.
Parent - - By Peter Grayson (****) [gb] Date 2015-11-14 16:51

> Not if you use logical principles :-). The purpose of the game is winning, so you should make life as difficult as possible for your >opponent. Exchanging pieces and pawns and staying close to a 0.00 score without knowing how to create these winning chances, is >not the way to go.


>Besides, if 'not losing' is becoming the main driver in a game of chess, you ask for many boring games and many draws, just as is >happening in the TCEC 8 final right now.


I agree with much of what you say. Despite all of the advances in computer chess engines up to today's current offerings, what I can never remember seeing is an engine that gives a suggestion that it can reproduce the beauty of human creativity. Some of the best top human players took risks pushing for the win looking for ways to create instability in the game that would likely give rise to their opponent making a mistake before they did. Sometimes that approach backfired but as a spectacle, seeing a closely contested 100 game finish 51-49 with 51 wins and 49 losses would be far more interesting than 2 wins, 1 loss and 97 draws. I do not see the answer is giving 3 points for a win because the correct question must be asked and that is simple enough. How can true creativity be programmed into artifical intelligence and remain competitive?

In that 60 game match using the Sharp Gambit lines there were a number of instances where the game finished as a draw when one of the engines missed a win that was not so difficult for the human eye to see, but as you rightly say the engines seemed far too focused on equalising the position and perhaps in many games the most dynamic possibilities were never explored.

Peter
Parent - - By Venator (Silver) [nl] Date 2015-11-14 17:57
Hi Peter,

the engines seemed far too focused on equalising the position and perhaps in many games the most dynamic possibilities were never explored.

This sums it up nicely. But even in level positions you see the same behaviour. Currently I am following the TCEC season 8 final and I am truly amazed how both SF and Komodo quickly turn to 0.00 evals, with a board full of pieces and many, many possibilities to make life difficult for your opponent.

How can true creativity be programmed into artifical intelligence and remain competitive?

Yes, absolutely. From time to time Deep Junior showed a little bit of this human-like asset, by sacrificing a pawn for strong pressure and the opponent having really no good possibility to get out of the bind. Once I spoke with Amir and he said something similar, i.e. to let Junior be creative while taking not too much risk.
Parent - - By Ozymandias (****) [es] Date 2015-11-14 18:46

>Currently I am following the TCEC season 8 final and I am truly amazed how both SF and Komodo quickly turn to 0.00 evals, with a board full of pieces and many, many possibilities to make life difficult for your opponent.


Apart from the missed win by SF, can you show a position where any of the engines went wrong?
Parent - - By Venator (Silver) [nl] Date 2015-11-14 20:18
'Went wrong'? I spoke about 'quickly going to 0.00 scores, without trying to get more from the position, a board full of pieces'.

You have big difficulties in accepting the fact that there still can be lot of improvement, haven't you?

As I wrote before: let's take up the discussion in 2025. That's more fruitful than me replying that you are wrong.
Parent - - By Ozymandias (****) [es] Date 2015-11-14 20:54

> 'Went wrong'? I spoke about 'quickly going to 0.00 scores, without trying to get more from the position, a board full of pieces'.


Either the 0.00 eval is correct or wrong, if it's correct it doesn't matter what you do with a board full of pieces, if it's wrong, show me an example.

I'm willing to wait whatever amount of time you want, to see if something different comes our way, in computer chess, but meanwhile, I'd like to clarify your position, because it's not clear to me if the wait will be worth it. If 10 years from now, an engine comes that "gets more from the position", without actually changing the outcome of a position evaluated as a draw today, why wait?
Parent - - By Venator (Silver) [nl] Date 2015-11-15 08:27 Edited 2015-11-15 09:28
OK, as you stubbornly keep repeating the same stuff over and over again, here are two clear cut examples:

Example 1) SF-Komodo, game 38 in the TCEC season 8 super final. Given opening moves: 1.d4 Nf6 2.c4 g6 3.Nc3 d5 4.Nf3 Bg7 5.Qb3 dxc4 6.Qxc4 0-0 7.e4 a6 8.Qb3 b5. Now from this position both SF and Komodo play 9.Be2 Nc6 10.e5 Be6 11.Qd1 Nd5 12.0-0 Nxc3 13.bxc3 Na5.

Any black player who plays the Grünfeld will agree that white completely failed to get anything out of the opening, with the squares d5 and c4 nicely blockaded. Of course the only serious try in this line is 9.e5, which any decent 2000-2200 elo player who plays the 5.Qb3 Grünfeld with white, knows. But 3000+ engines on a very fast machine play an inferior line, killing all possible chances before move 15.

Now let's follow this game more closely and go to move 23.Red1. Every decent chess player will easily see that black's position is far more comfortable and he is the only one who can try to win this position. Still, SF insists this is only -0.08, i.e. a very slight edge to black. Black's bishops are superior, his knight is superior and his queen is more active, when moving to f5. And this is supposed to be almost equal? Totally wrong.

A couple of moves later SF simply concedes the a-file to black, improving Komodo's position even further. Still, only a -0.15 eval after a 39 ply search. Within a few moves white's position becomes hopeless, which will come as no surprise to decent chess players.

Example 2) Set up the following position: White Kg2,Rb2,pawns f2,g3,h4. Black Kg7,Ra7, pawns g6,h5. Without tablebases SF claims white has a clear advantage here, on my slow notebook +0.73. Of course a completely wrong eval, as the position is a 100% draw.

Now let's put 1 and 2 together: apparently one of the best engines in the world claims that position 2 offers more winning chances than position 1 after 9.e5. Which is hogwash. Even worse, if SF sees a chance from a board full of pieces and a +0.50 edge to reach position 2 by force, it will duly exchange all pieces and pawns. But by doing so, it loses all chances of winning.

Conclusion: even the best engines make huge eval errors and throw away winning chances. Hence, there is still a lot to improve. But I already told you that many times before. Naturally, there are still millions of bad evals out there, which means there are lots of calculated lines that do not reflect the position correctly. And, as I showed you, in some positions a 2000 elo player knows better than the top engine of the world.
Parent - By Ozymandias (****) [es] Date 2015-11-15 12:11

>OK, as you stubbornly keep repeating the same stuff over and over again, here are two clear cut examples:


You mean to say you aren't stubborn?

>Any black player who plays the Grünfeld will agree that white completely failed to get anything out of the opening, with the squares d5 and c4 nicely blockaded. Of course the only serious try in this line is 9.e5, which any decent 2000-2200 elo player who plays the 5.Qb3 Grünfeld with white, knows. But 3000+ engines on a very fast machine play an inferior line, killing all possible chances before move 15.


Let me get back to you on that, once I get ahold of a decent 2000-2200 elo player who plays the 5.Qb3 Grünfeld with white.

>Now let's follow this game more closely and go to move 23.Red1. Every decent chess player will easily see that black's position is far more comfortable and he is the only one who can try to win this position. Still, SF insists this is only -0.08, i.e. a very slight edge to black. Black's bishops are superior, his knight is superior and his queen is more active, when moving to f5. And this is supposed to be almost equal? Totally wrong.


Komodo was giving a similarly timid -0.13 eval, and it ended wining, so any engine giving a higher eval for black would not make a big difference here. The only important thing to analyze is whether SF went wrong with 14. Ba3, or later.

>Example 2) Set up the following position: White Kg2,Rb2,pawns f2,g3,h4. Black Kg7,Ra7, pawns g6,h5. Without tablebases


Tablebases are part of computer chess, leaving them out is arbitrary.

Opening books are likewise an integral part of computer chess, looking for engines to solve early opening positions may be fruitful sometimes, but is generally ill-advised.

Conclusion: eval errors are often compensated trough search, and when engines can't reach deep enough, or prune in excess, you get one of those RARE cases, when they can be beaten. There's some room for improvement, but is nothing dramatic and doesn't preclude the fact that we're near the end. A new crop of engines looking to "maximize winning chances" isn't going to do anything differently to what top centaurs do today, and they might already be no better than engines, as their performance in the CWTs show.
Parent - - By Venator (Silver) [nl] Date 2015-11-15 10:36
r1b1k2r/1p1nbppp/1qn1p3/p2pP3/3P4/3B1N2/PP1N1PPP/R1BQR1K1 w kq - 0 11


Game 41 of the TCEC 8 super final. Position after 10... Qb6.

Stockfish evaluation: 0.00
Komodo evaluation: 0.00

So in this interesting position with all pieces but one pawn still present, both SF and Komodo evaluate this as completely equal (= no winning chances, high probability of a draw)!?

This just sounds as silly as it is. I am not surprised that most level positions between the two will end in draws. They are just not interested in creating chances from this position, they are steering it towards a draw as quickly as possible.
Parent - - By Ozymandias (****) [es] Date 2015-11-15 11:39

>So in this interesting position with all pieces but one pawn still present, both SF and Komodo evaluate this as completely equal (= no winning chances, high probability of a draw)!?


We agree on something here, this sounds silly, because it is. An evaluation of 0.00, doesn't imply universally that you should consider the position as a draw. Clearly in this instance (one side not having even castled yet), and even if the game ends in a draw (which is what it looks like), the eval just means that both sides consider the position as equal, no statement about winning chances should be inferred from it.
When I asked you about a 0.00 eval where the engines went wrong, I meant that, one where you can show a forced win despite said eval. Anything else is inconsequential, as it doesn't have practical consequences. Keeping the pieces on the board, complicating things, means nothing unless your opponent makes a mistake. Even if you manage to actually find ONE position, out of the thousands we'll see during the super-final, where engines draw after incorrectly evaluating the position, that won't imply that an engine capable of playing in such a way, would be a big improvement (which seems to be what you think).
Parent - By Venator (Silver) [nl] Date 2015-11-15 11:57 Edited 2015-11-15 12:42
You keep running in circles :-).

I gave you 2 clear examples with explanation, where SF showed evaluations that were totally off the mark. Peter Grayson has posted numerous examples where SF lost after showing a 0.00 score for a VERY long time (15-20 moves, if I recall correctly). One of these positions can be found in this thread:

http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?tid=30679

Furthermore, if SF says -0.15 in a clearly worse (maybe even lost) position *and* spoils a better position by exchanging a 0.5 edge in a middlegame position into a 0.75 endgame position which is drawn, it is clear to everybody that there are still huge evaluation errors in the current top engines.

An evaluation of 0.00, doesn't imply universally that you should consider the position as a draw.

Sure, but in this example (41st Komodo-SF game) most of the main lines end in perpetual check, or a move repetition. Hence the 0.00 score represents a draw.

You still stick to the opinion that the current engines cannot be improved much further and explore all possibilities to win a game. Fine, so be it. I know it is wrong and that the future will prove my point. For me there is no need anymore to continue this discussion.
Parent - - By Ozymandias (****) [es] Date 2015-11-14 18:44

>Some of the best top human players took risks pushing for the win looking for ways to create instability in the game that would likely give rise to their opponent making a mistake before they did.


Apples and oranges, you can translate human playing behavior, aimed at exploiting psychological flaws, to computer chess.

>In that 60 game match using the Sharp Gambit lines there were a number of instances where the game finished as a draw when one of the engines missed a win that was not so difficult for the human eye to see


For example?
Parent - - By Venator (Silver) [nl] Date 2015-11-14 20:23
You keep stubbornly denying what Peter and I observe, namely the urge of both K and SF to draw games instead of trying to extract the maximum from a position.

SF and Komodo are still far, far away from being perfect chess players. They know how to draw, not how to maximise winning chances. But I already told you this on several occasions....
Parent - - By Ozymandias (****) [es] Date 2015-11-14 20:48

>SF and Komodo are still far, far away from being perfect chess players. They know how to draw, not how to maximize winning chances.


If they know how to draw, the game's over, it doesn't matter what you do with your "winning chances", because they're 0.00
Parent - - By Venator (Silver) [nl] Date 2015-11-15 07:59
I have a simple question for you: are you a strong chess player? What is your elo?
Parent - - By Ozymandias (****) [es] Date 2015-11-15 11:22

>are you a strong chess player?


I'm going to translate a previous reply, from me to you, which you clearly didn't understand.
Original quote:
I may not be a good chess player, so I might be inclined to think that I "have no clue about how to increase your chances of winning", but then I beat better OTB players than me, at Freestyle, and reconsider.
Translation:
Even though I'm not a good chess player, the possibility that I might "have no clue about how to increase your chances of winning", gets sidelined ben I consider the fact that I've beaten better OTB players than me, at Freestyle.
What this all means, is that human ELO has a very small role, if any, when playing with computer aid. The first example of this, you could already find in the 1st PAL/CSS, and Anson continues being one of the best operators, regardless of his human chess playing skills.

>What is your elo?


1803, why do you want to know?
Parent - - By Venator (Silver) [nl] Date 2015-11-15 11:38
1803, why do you want to know?

Because I have the experience that a lot of weak(er) players are having firm opinions about computer chess games and evals, while it is obvious to me they often don't have a clue what they are talking about.

But not only weaker players trust engines too much. Even Kramnik lost a World Championship game against Leko in a Marshall gambit, because he trusted Fritz's 'white is clearly better' eval, while in fact white was already in dire straits.

BTW, in my best days I had over 2200 elo and I am in the computer chess business since 1987. I have played in 2 Freestyle finals, winning one of the preliminary rounds together with Flyingfatman.
Up Topic The Rybka Lounge / Opening Books / 250 MB - The End Is Near
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill