Is it useful to use more than one core for stockfish?

[Uri Blass]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

[lkaufman]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

[Graham Banks]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Dragon by Komodo 3 v SF15 for CCRL 40/15

4CPU - 32 draws (no decisive results)
1CPU - 54 draws (no decisive results.

[Uri Blass]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Only to make things clear What I suggest is not a correspondence match between dragon3 on 4 cpu and humans from CCRL starting positions.
The idea is that Dragon is going to use 1 minute per move with 4 cores when the human opponent has correspondence time control in order to find practical tries to win.

I do not claim that humans can get 100 elo advantage but the task is easier than the task of getting 100 elo advantage when both sides use correspondence time control.

[lkaufman]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Only to make things clear What I suggest is not a correspondence match between dragon3 on 4 cpu and humans from CCRL starting positions.
The idea is that Dragon is going to use 1 minute per move with 4 cores when the human opponent has correspondence time control in order to find practical tries to win.

I do not claim that humans can get 100 elo advantage but the task is easier than the task of getting 100 elo advantage when both sides use correspondence time control.

I see. That is at least possible, if not likely. So you would start the games from the positions where CCRL games are started, after a few book moves, but with no book for Dragon after that, right? CCRL uses various books, so the results might be quite dependent on exactly which book was used. If the openings are taken from current practice of top GMs, it will be very hard for anyone to beat Dragon, but if they include some random dubious openings played by amateurs then Dragon may lose some games as Black under these conditions.

[lkaufman]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Dragon by Komodo 3 v SF15 for CCRL 40/15

4CPU - 32 draws (no decisive results)
1CPU - 54 draws (no decisive results.

All draws on just 1 CPU is pretty remarkable. What opening book was used, and can you characterize the book (i.e. openings played by grandmasters, amateur openings, openings selected by some criteria, etc.)? The influence of the opening book on results is becoming critical now; it didn't used to matter much when the level of play was not so incredibly high.

[Graham Banks]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Dragon by Komodo 3 v SF15 for CCRL 40/15

4CPU - 32 draws (no decisive results)
1CPU - 54 draws (no decisive results.

All draws on just 1 CPU is pretty remarkable. What opening book was used, and can you characterize the book (i.e. openings played by grandmasters, amateur openings, openings selected by some criteria, etc.)? The influence of the opening book on results is becoming critical now; it didn't used to matter much when the level of play was not so incredibly high.

IECGMasters,cgb book.
Balanced lines from games between players 2500+ from memory.

[Uri Blass]

The first question is what is possible today to get.

Can humans with the help of engines and correspondence time control get 100 elo advantage or more than it against
Dragon 3 on 4 cpu?

That is at least something that could be tested practically with little cost. My guess would be "no". The answer of course is somewhat dependent on the opening book (if any) used by Dragon 3. A more practical version would be to play the games at some normal classical time limit for a one day event.

Only to make things clear What I suggest is not a correspondence match between dragon3 on 4 cpu and humans from CCRL starting positions.
The idea is that Dragon is going to use 1 minute per move with 4 cores when the human opponent has correspondence time control in order to find practical tries to win.

I do not claim that humans can get 100 elo advantage but the task is easier than the task of getting 100 elo advantage when both sides use correspondence time control.

I see. That is at least possible, if not likely. So you would start the games from the positions where CCRL games are started, after a few book moves, but with no book for Dragon after that, right? CCRL uses various books, so the results might be quite dependent on exactly which book was used. If the openings are taken from current practice of top GMs, it will be very hard for anyone to beat Dragon, but if they include some random dubious openings played by amateurs then Dragon may lose some games as Black under these conditions.

I think that the book include few dubious openings that GM played
This is an example(stockfish won with white and lost with black against ShashChess from this opening so maybe the opening is winning for white but I am not sure but I am sure stockfish shows a clear advantage for white after 8...e4 that is the last book move)

[pgn][Event "CCRL 40/15"]
[Site "CCRL"]
[Date "2022.05.05"]
[Round "822.4.594"]
[White "Stockfish 15 64-bit 4CPU"]
[Black "ShashChess 21.1 64-bit 4CPU"]
[Result "1-0"]
[BlackElo "3531"]
[ECO "A21"]
[Opening "English"]
[Variation "Kramnik-Shirov, 3.Nd5 Bc5"]
[WhiteElo "3531"]
[TimeControl "5400+30"]
[Termination "normal"]
[PlyCount "168"]
[WhiteType "human"]
[BlackType "human"]

1. c4 {0.00/1 0s} e5 {0.00/1 0s} 2. Nc3 {0.00/1 0s} Bb4 {0.00/1 0s} 3. Nd5
{0.00/1 0s} Bc5 {0.00/1 0s} 4. e3 {0.00/1 0s} Nf6 {0.00/1 0s} 5. b4 {0.00/1
0s} Nxd5 {0.00/1 0s} 6. bxc5 {0.00/1 0s} Nf6 {0.00/1 0s} 7. Nf3 {0.00/1 0s}
Qe7 {0.00/1 0s} 8. Be2 {0.00/1 0s} e4 {0.00/1 0s} 1-0
[/pgn]

[Jouni]

I did limited 4th-1th test with SF15. 8 moves book. With 10+0,1 level +70 ELO, but with 60+0,6 level +30 ELO. With 240+2,4 no gain at all? Too time consuming to test with one PC.