TCEC S15, END of an ERA event is much more Brutal than I thought!

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Ozymandias » Wed May 29, 2019 5:14 pm

Kanizsa wrote:
Wed May 29, 2019 1:32 pm
2 I suggest that Centaur player should read the first lines of multiple engines but not explore them back and forth.
3 the possibility to withdraw the move after LC0 reply is another advantage for human player.

Regardless of the rules the aim of similar experiments is as follows:

despite having huge advantage thanks to the best AB engines, human intuition would not be able to beat the positional insight of LCO, not even withdrawing the moves played. If this is true, Kasparov ipothesis should be rejected.
Rule 3 is only of any use if you introduce rule 2, which goes against the most basic centaur technique there is.

Kasparov was talking about all this when Fritz was top dog, I doubt that whatever were to happen today would invalidate his hypothesis, in any case it would point to it being outdated. Furthermore, he was talking about FIDE pace of play, not correspondence play. With so much time, there's no doubt the centaur still has the edge.

Kanizsa
Posts: 32
Joined: Mon Feb 20, 2017 7:29 am
Location: Rialto, Venice

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Kanizsa » Thu May 30, 2019 12:04 pm

Ozymandias wrote:
Wed May 29, 2019 5:14 pm
Kanizsa wrote:
Wed May 29, 2019 1:32 pm
2 I suggest that Centaur player should read the first lines of multiple engines but not explore them back and forth.
3 the possibility to withdraw the move after LC0 reply is another advantage for human player.

Regardless of the rules the aim of similar experiments is as follows:

despite having huge advantage thanks to the best AB engines, human intuition would not be able to beat the positional insight of LCO, not even withdrawing the moves played. If this is true, Kasparov ipothesis should be rejected.
Rule 3 is only of any use if you introduce rule 2, which goes against the most basic centaur technique there is.

Kasparov was talking about all this when Fritz was top dog, I doubt that whatever were to happen today would invalidate his hypothesis, in any case it would point to it being outdated. Furthermore, he was talking about FIDE pace of play, not correspondence play. With so much time, there's no doubt the centaur still has the edge.
Ok, Kasparov was not a phenomenon even in advanced chess, as he drew against Topalov.

But what do you think about a modern Top GM plus Stockfish in advanced chess against Leela in FIDE modern pace of play (2 hours) ?

I think that Anand or Nakamura should lose by a narrow edge.

So I suggested to give some possibilty to withdraw one or more moves, since Leela often plays extraordinary moves leading to material imbalances, that are not first lines by Stockfish, with difficulties that are far beyond the horizon of a strong centaur.

jp
Posts: 840
Joined: Mon Apr 23, 2018 5:54 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by jp » Thu May 30, 2019 1:19 pm

Kanizsa wrote:
Thu May 30, 2019 12:04 pm
Leela often plays extraordinary moves leading to material imbalances, that are not first lines by Stockfish, with difficulties that are far beyond the horizon of a strong centaur.
Can you give a few examples?

User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Ozymandias » Thu May 30, 2019 1:38 pm

Kanizsa wrote:
Thu May 30, 2019 12:04 pm
what do you think about a modern Top GM plus Stockfish in advanced chess against Leela in FIDE modern pace of play (2 hours) ?
I think they'd do worse than unassisted SF (openings aside, as I said the book situation needs to be addressed before venturing an answer). GMs, top or otherwise, train to play well alone, and may at most use engines, but that doesn't mean they have the knowledge, the experience or the skills to play alongside engines.

jp
Posts: 840
Joined: Mon Apr 23, 2018 5:54 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by jp » Thu May 30, 2019 1:50 pm

Ozymandias wrote:
Thu May 30, 2019 1:38 pm
...
I think they'd do worse than unassisted SF (openings aside, as I said the book situation needs to be addressed before venturing an answer). GMs, top or otherwise, train to play well alone, and may at most use engines, but that doesn't mean they have the knowledge, the experience or the skills to play alongside engines.
But if you instructed them to be very disciplined, playing the opening and gradually letting SF take over, that opening book would be a huge advantage that would nullify the NN engines' main advantage.

User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Ozymandias » Thu May 30, 2019 2:40 pm

jp wrote:
Thu May 30, 2019 1:50 pm
Ozymandias wrote:
Thu May 30, 2019 1:38 pm
...
I think they'd do worse than unassisted SF (openings aside, as I said the book situation needs to be addressed before venturing an answer). GMs, top or otherwise, train to play well alone, and may at most use engines, but that doesn't mean they have the knowledge, the experience or the skills to play alongside engines.
But if you instructed them to be very disciplined, playing the opening and gradually letting SF take over, that opening book would be a huge advantage that would nullify the NN engines' main advantage.
No book was actually proposed, I just said it was a pre-requisite to talking about what we're discussing, but the discussion is going along anyway, so there's that.

jp
Posts: 840
Joined: Mon Apr 23, 2018 5:54 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by jp » Thu May 30, 2019 3:03 pm

Ozymandias wrote:
Thu May 30, 2019 2:40 pm
...
No book was actually proposed, I just said it was a pre-requisite to talking about what we're discussing, but the discussion is going along anyway, so there's that.
There's been frequent discussion before about SF+book (vs. Lc). If you have SF+superGM, then you get superGM's book, though of course not perfectly remembered. Then it's become at least as good as SF+book, which Lc would struggle to get many wins against (though that depends on exactly what book).

User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Ozymandias » Thu May 30, 2019 3:12 pm

jp wrote:
Thu May 30, 2019 3:03 pm
Ozymandias wrote:
Thu May 30, 2019 2:40 pm
...
No book was actually proposed, I just said it was a pre-requisite to talking about what we're discussing, but the discussion is going along anyway, so there's that.
There's been frequent discussion before about SF+book (vs. Lc). If you have SF+superGM, then you get superGM's book, though of course not perfectly remembered. Then it's become at least as good as SF+book, which Lc would struggle to get many wins against (though that depends on exactly what book).
GM's have an opening repertoire and the related analysis. I don't see how their memories of such a thing could even come close to an updated book.

jp
Posts: 840
Joined: Mon Apr 23, 2018 5:54 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by jp » Thu May 30, 2019 3:28 pm

Ozymandias wrote:
Thu May 30, 2019 3:12 pm
...
GM's have an opening repertoire and the related analysis. I don't see how their memories of such a thing could even come close to an updated book.
I assume this hypothetical match is a standard chess match without any forced openings, so it doesn't matter that superGM does not play every single opening. What matters is he has the deepest preparation in his own repertoire including all the novelties saved up that have not appeared OTB yet. I'd certainly take Carlsen's and Anand's secret WC prep over Cerebellum, etc.

User avatar
Ozymandias
Posts: 1137
Joined: Sun Oct 25, 2009 12:30 am

Re: TCEC S15, END of an ERA event is much more Brutal than I thought!

Post by Ozymandias » Thu May 30, 2019 4:26 pm

jp wrote:
Thu May 30, 2019 3:28 pm
Ozymandias wrote:
Thu May 30, 2019 3:12 pm
...
GM's have an opening repertoire and the related analysis. I don't see how their memories of such a thing could even come close to an updated book.
I assume this hypothetical match is a standard chess match without any forced openings, so it doesn't matter that superGM does not play every single opening. What matters is he has the deepest preparation in his own repertoire including all the novelties saved up that have not appeared OTB yet. I'd certainly take Carlsen's and Anand's secret WC prep over Cerebellum, etc.
Their preparation is aimed at OTB play, I don't expect that to do well at all against (for example) Cerebellum, mainly (but no just) because their repertoire is not as wide as the range of openings covered in a good engine book.

Post Reply