The Next very strong Engine will be rofChade NNUE

[Paloma]

Weird quoting what you are doing all the time.
you quoting yourself

[AndrewGrant]

you really must work on the way you use the forums. Quoting yourself means I just skip the post, since I assume I've already read it. Plus you can delete part of the quote history that is not revelent. Otherwise in a long enough forum, the quote takes up an entire page per posting.

[Paloma]

Andrew,
I mean Chessqueen.

by Chessqueen » 07 Jan 2021, 15:46
by Chessqueen » 07 Jan 2021, 15:50

the interleaving is strange

is a pain in the ass
gets on my nerves

[AndrewGrant]

Andrew,
I mean Chessqueen.

by Chessqueen » 07 Jan 2021, 15:46
by Chessqueen » 07 Jan 2021, 15:50

the interleaving is strange

is a pain in the ass
gets on my nerves

Yes yes, I know who you meant.

[Frank Quisinsky]

Hi there,

all the discussion about Elo gain from NN ... 160-200 Elo or more!
Totaly nonsense if many engines in a group of a tourney or test field.

I saw that in test games with Stockfish 12, maximal 85 Elo more with NN support.
But of course, if I test Stockfish vs. a small group of engines only, Elo goes higher.

More sensational are messages ... Elo will go 160 Elo higher.
People, the market or chess server like that because more people will looking.

Same for all the other NN engines.

In case of rofChade!
Engine play the game phases (late mid-game, early endgame and endgame) very stable.
Evals are strong!

I am quite sure that rofChade will not get 160 Elo more with NN support.
Maximal possible 80-100 Elo.

In case of Pedone ...
Maximal 100 Elo more for Pedone 3, I believe 80 Elo more.

NN is great but what I read here for Elo-gain wishes isn't realistic.
If rating list systems test stronger engines vs. a small group of engines, without to added weaker engines, Elo goes higher and higher.

Regards to Elo inflation!

Best
Frank

[Chessqueen]

Hi there,

all the discussion about Elo gain from NN ... 160-200 Elo or more!
Totaly nonsense if many engines in a group of a tourney or test field.

I saw that in test games with Stockfish 12, maximal 85 Elo more with NN support.
But of course, if I test Stockfish vs. a small group of engines only, Elo goes higher.

More sensational are messages ... Elo will go 160 Elo higher.
People, the market or chess server like that because more people will looking.

Same for all the other NN engines.

In case of rofChade!
Engine play the game phases (late mid-game, early endgame and endgame) very stable.
Evals are strong!

I am quite sure that rofChade will not get 160 Elo more with NN support.
Maximal possible 80-100 Elo.

In case of Pedone ...
Maximal 100 Elo more for Pedone 3, I believe 80 Elo more.

NN is great but what I read here for Elo-gain wishes isn't realistic.
If rating list systems test stronger engines vs. a small group of engines, without to added weaker engines, Elo goes higher and higher.

Regards to Elo inflation!

Best
Frank

Here is a game versus Houdini 6.03 using an intel i9 8800K Versus my Samsung S20 with Stockfish

[pgn][Event "Samsung Galaxy S20 Vs Intel i9 9900K"]
[Date "2021.01.07"]
[Round "3"]
[White "Stockfish 280920 "]
[Black "Houdini 6.03"]
[Result "1-0"]
[BlackElo "3060"]
[ECO "E87"]
[Opening "King's Indian"]
[Time "16:17:26"]
[Variation "Sämisch, Orthodox, Bronstein, 9.g3"]
[WhiteElo "3480"]
[TimeControl "900+3"]
[Termination "normal"]
[PlyCount "88"]


1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. f3 O-O 6. Be3 e5 7. d5 Nh5 8. Qd2
Qh4+ 9. g3 Nxg3 10. Qf2 Nxf1 11. Qxh4 Nxe3 12. Qf2 Nxc4 13. h4 h5 14. Qe2
Nb6 15. Nb5 Na6 16. Nh3 Bh6 17. a4 Bxh3 18. Rxh3 Nd7 19. a5 Ndc5 20. Qc4 f5
21. Kf1 fxe4 22. b4 Nd7 23. Nc3 exf3 24. Ne4 Nf6 25. b5 Nc5 26. Nxc5 dxc5
27. d6+ Kh7 28. dxc7 e4 29. Re1 Rae8 30. Rxf3 Rc8 31. Rf2 Rxc7 32. Ke2 e3
33. Rf3 Bg7 34. Ref1 Rd7 35. Rd1 Re7 36. Rf4 Ne4 37. Rxe4 Rf2+ 38. Kxe3
Ref7 39. Re8 R7f3+ 40. Ke4 Rf4+ 41. Kd5 Rxc4 42. Kxc4 Rc2+ 43. Kd3 Ra2 44.
Re7 Kh6 45. Rf1 c4+ 46. Kxc4 Ra4+ 47. Kc5 Bd4+ 48. Kd5 Bc3 49. Rxb7 Rxa5
50. Kc4 Be5 51. Rff7 g5 52. b6 Ra4+ 53. Kb5 Ra1 54. bxa7 Kg6 {Black
resigns} *[/pgn]

[voffka]

[...]
after playing around ~40,000,000 games, training on a few billion positions, using a machine with 104 threads. Noob basically gave me short term access to more computational power than most people can even begin to muster for such a trial and error task like this.

Yeah, 104 threads machine will speed up things significantly (probably converts multiple months into weeks of training). Elo gains by strong HCE engines like Ethereal and Rofchade will be hard to achieve (but not impossible) as they require a lot of computational power and training experience just to reach the level of HCE and then bypass it.

[Chessqueen]

Weird quoting what you are doing all the time.
you quoting yourself

Why did you said that Paloma voladora, in my preview post I did NOT mentioned that I believe that Komodo also gained at least 73 using Komodo NET. At the moment Ethereal is only about 59+ yp about 65+ Elo stronger than rofChade, so roChade with a NET could easily reach the same rating of Ethereal
https://tcec-chess.com/live.html


7
Ethereal
TCEC S20 DivP
40 14.5 [2.5] 36.25% 4 [4/0] 15 [0/15] 21 [16/5] 14.596 [0.096] 273.25 3482 -11 [3471]
8
rofChade
2.309
39 11.5 29.49% 2 [2/0] 18 [1/17] 19 [17/2] 11.656 [0.156] 233.75 3423 -10 [3413]