Computer Chess Engine Classical Chess Championship. Stockfish 14 vs. Dragon 2.5.1

mwyoung wrote:

Uri Blass wrote:

mwyoung wrote:

Uri Blass wrote:

mwyoung wrote:

Here is what is funny about the comments I have been seeing about this match. Many would have you believe any engine can do what Dragon 2.5.1 is doing to Stockfish at long time controls. Stopping Stockfish.

This is just false, and from people that do not test chess engines as I do.

The #3 engine in the world can not do this. Or any of the also here engines of a much lower tier.

How do you know that other engines cannot draw against stockfish14 at 120/40 time control in your conditions?
Did you test them?

yes! Have you....

Stockfish wins against everything.

This is the first test with Dragon 2.5.1 at long time controls. If you think another engine can also stop stockfish. Post it here!

And we will test it.

I think that it may be interesting to test stockfish11 against stockfish14 because stockfish11 is the best or close to the best engine that does not use NNUE.

I have many times. SF 11 is crushed.

But if that is your engine for a Long time control match. I will play it under the same conditions.

It will be fun to see!

Your match will be played under the exact conditions as this match. As soon as this match is over.

I know that SF11 is weaker at faster time control based on rating lists.
I do not know if it is the same at long time control considering the fact that it seems SF14 does not earn from additional time.

I will not be surprised if SF11 is going to lose but it may be interesting to see and it will be interesting to see if SF11 can get at least one win against SF14.
Losing 6-1 with 17 draws is of course worse than 24 draws but at least show that it is possible to beat SF14.

Given that, very roughly, LC0 has a lot of knowledge but a slow search, I would expect LC0 to be the biggest beneficiary of very long time controls.

Am I right?

Admin wrote:

Does the GUI not have a draw button? If both engines display 0.00 for 5 moves then draw the game. These silly endings are a pain to watch.

I prefer not to have draw when both engines show 0.00 because I am not sure that every case of it is a draw.
We already have 100% draws and I am afraid this type of adjudication is going to prevent a win in case that there is a chance for a win in some case.

0.00 does not mean draw but means equal position.

Score of Dragon 2.5.1 vs Stockfish 14: 0 - 0 - 24 [0.500]
...      Dragon 2.5.1 playing White: 0 - 0 - 12  [0.500] 12
...      Dragon 2.5.1 playing Black: 0 - 0 - 12  [0.500] 12
...      White vs Black: 0 - 0 - 24  [0.500] 24
Elo difference: 0.0 +/- 0.0, LOS: nan %, DrawRatio: 100.0 %
24 of 24 games finished.

Not funny but thanks for trying.

The top engines facing each other has become an exercise in draw death, not just because of the high strength, but the closeness of the ratings.

A more practical and revealing test would be gauntlets against a multitude of lower rated engines, as much as one to three classes below the top, and then winning will become common place, and the entity that wins the most that way can be the top dog.

(See my other post in the NNUE dead end thread)

mwyoung wrote:

Score of Dragon 2.5.1 vs Stockfish 14: 0 - 0 - 24 [0.500]
...      Dragon 2.5.1 playing White: 0 - 0 - 12  [0.500] 12
...      Dragon 2.5.1 playing Black: 0 - 0 - 12  [0.500] 12
...      White vs Black: 0 - 0 - 24  [0.500] 24
Elo difference: 0.0 +/- 0.0, LOS: nan %, DrawRatio: 100.0 %
24 of 24 games finished.

I wonder if the opening book lead to drawing lines or maybe stockfish improved significantly from stockfish14.

I am surprised to find that Stockfish could beat dragon at 90+30 with many opening lines(1.Nf3 1.d4 and 1.e4(both sides had no books later) so black cannot blame bad lines in the opening book for the loss)
I am also surprised to find that stockfish could hold g4 against Dragon Komodo at 90+30 time control when I considered 1.g4 to be accepted losing move at the high levels based on some correspondence games.

Maybe part of it is because no egtb are used but it seems at least not to be the only reason.

90+30 s faster time control than 120/40 but I remember to see usually draws also in faster time control of mark young and in time handicap games.

http://phcel.cz/