BladezII wrote on 12/09/22 at 16:32:32:


I wrote about that "new" trend already on my blog before see
http://chess-brabo.blogspot.com/2015/07/computers-achieve-autonomy.html
(engines start to develop openings without human input in 2015)
http://chess-brabo.blogspot.com/2019/06/computers-achieve-autonomy-part-2.html
(engines are getting final conclusions about more and more openings without human input in 2019)

cathexis wrote on 12/09/22 at 14:17:53:

Sure. But a qualitative difference is that the engines are so strong now, no human input is needed. Twenty-five years ago Mr. Edwards himself said he hoped his opponents were using an engine for CC! Meaning he expected them to be led astray by the engine's top moves, and he could outplay them by using his own brain. (This was USCF correspondence play which did ban computers then, I have no idea about today.) Probably by ten years ago (at the latest) things had flipped and a human-guided engine was sure to outplay both the human alone and the engine alone. Then there is today where the CC players in this thread say there is a range of openings where the engines play perfectly on their own. We can only expect that range of openings to expand over time.

Steinitz's theory of "accumulation of small advantages", although it predates engines, is similar to today's notion of "average centipawn loss". If this is how chess games are won and lost, what happens to our winning chances when the opponent's average centipawn loss equals zero?

Jonathan Tait wrote on 12/05/22 at 16:24:17:



This post here is getting to the heart of what I am trying communicate. 

If you play a sound opening as black, Stockfish 15 is unbeatable.  It will defend just fine black

You can leave it alone in standard 2. Nf3  queens gambit decline lines as black and it will hold without human help 

Marshall gambit in the Ruy Lopez , same thing

Giucco Piano, same thing 

What he posted above is a strong piece of evidence supporting my argument. -- at this level of play, you are not going to see wins outside of other factors.

At this level , using engines, using the best engines, and how good those engines are now, especially Stockfish 15... 

... at this level, playing solid, reliable, tested openings, and avoiding human mistakes, the games should be drawn.

Case in point?   
Look at the previous comments concerning the latest World Correspondence Championships 

In addition to that

1. e4  e5

Ruy lopez, like I said before, especially Marshall Gambit, that's drawn, so white is dead in the water trying to get anything there

Marshall Gambit... anti Gambit line, there's nothing there either

Giucco Pianno, nothing there.  Drawn. 

1.d4, both Nimzo Indian and QGD Nf3 lines... white has nothing there

English... forget about it, after 1... e5 White again has nothing 

Read again over the comments incase you missed anything 

The WCCC was all drawn except for other factors that had nothing to do with what I am discussing here. 
In those cases, those very exceptional  cases, those were the only decisive games. 

Otherwise, if all things were normal, they were all drawn.

Quote:


Hhmmm,

Consider this:

https://exeterchessclub.org.uk/content/theory-steinitz?msclkid=5cd557e8d14a11ec9...

I know some might quibble about the "Theory of Steinitz". But taking it at face value, I don't find anything in your comment that disproves the statements in the link above. So SF is not destroying anything, merely affirming what was already known.

Jonathan Tait wrote on 12/06/22 at 16:39:52:


Very well said and on point observation.  Ultimately if the engine isn't finding a way to progress as you analyze further it likely means a draw.  

Objectively speaking games have three possible outcomes.  Any evals that are not 0.00 or a concrete win are objectively "wrong" in the long run (although meaningful in a practical sense over the board).  As more analysis is done the evals should either converge on 0.00 or increase as progress is made toward a win.  If they don't change then it means the engine isn't finding a way to progress and likely has not realized that the game is drawn yet.  ICCF players spend a lot of time guiding their engines down lines to find the truth in these positions.

Edit:  I also think this highlights an interesting conundrum as engines get stronger.  As we get to a point where engines start accurately evaluating more and more positions as 0.00 much sooner the average club player OTB is faced with evals that don't really help them prepare as the path to 0.00 may be narrow.  Yes, the positions may be 0.00 but from a practical perspective much easier for one side or the other.  OTB one side does have "an advantage" even if it isn't objectively convertible with perfect play.  So objective evals we seek for ICCF play are not really of much help for those preparing lines OTB or analyzing OTB games.  People are going to need to pay attention to more metrics than just the centipawn evaluation.

@ bragesjo, just a thanks for your reply!

About Bogo I played so badly I am ashamed over those games. I became demoralized by both Stockfish 13 and Fat Fritz 2 thought pawn better for opponent and suggestiv planless heavy defence moves so is like human vs human+computer. Stockfish 15 is more optimistisk 0.71  better for white depth 50.

I saw a later camp at ModernChess and other author wrote that engines hates this line but we are playing againt humans not computers so its nothing new. Many opening books on bases are written for over the board play and are not corretly computer checked in every line so nothing new. There was a simular mistakes a Nimzo bases as well but I had preanalysed and played right diversions and easy draws.

I can also add that books can quate human games with both 2700 claim equal when modern computer can claim one side is winning after a impossible to see computer move