correspondence chess in the age of NNUE

[Master Om]

Here's the opinion (3 best moves) of Crystal3.1 about the position after 33... Rfg7 34. Nfg4 hxg4 35. fxg4
It likes Qg6 followed by f3

[d]6rk/pp1bb1r1/3p1n1q/PPpPp1p1/2P1PpPn/R6P/4B1PN/1Q2BRK1 b - - 0 35

Code: Select all

Crystal3.1-sse4-popcnt:
...
 29/48	04:03	1 311 698 851	5 378 724	-2,13	35. ... Qh7 36.Bxh4 Qxh4 37.Rf2 Ra8 38.Rf1 Bd8 39.Bf3 Nh7 40.Rf2 Qh6 41.Rfa2 Nf8 42.Ra1 Kg8 43.R1a2 Ng6 44.Kf2 Rb8 45.Kf1 Qh4 46.Ke2 Qh7 47.Ke1 Nh4 48.Kf1 Rf7 49.Qd3 Qh8 50.Qb1 Kg7 51.Qd3 Ng6 52.Qb1 Be7 53.Kg1 Bf6 54.Kf1 Nh4 55.Kf2 a6 56.b6
 29/52	04:03	1 311 698 851	5 378 724	-2,13	35. ... Ra8 36.Ra2 Bd8 37.Ra3 Ng6 38.Bf2 Qh7 39.Bf3 Nh4 40.Bxh4 Qxh4 41.Rf2 Nh7 42.Rf1 Nf8 43.Rf2 Ng6 44.Rfa2 Kg8 45.Rb2 Qh6 46.Rba2 Rb8 47.Kf1 Qh4 48.Ke2 Qh7 49.Ke1 Nh4 50.Kf1 Rf7 51.Qd3 Qh8 52.Qb1 Kg7 53.Qd3 Ng6 54.Qb1 Be7 55.Kg1 Bf6 56.Kf1 Nh4 57.Kf2 a6 58.b6
 29/47	04:03	1 311 698 851	5 378 724	-2,35	35. ... Qg6 36.Ra2 f3 37.Nxf3 Qxe4 38.Qxe4 Nxe4 39.Nh2 Ng6 40.g3 Nf6 41.Bc3 Rf8 42.Nf3 Kg8 43.Kg2 Bd8 44.Bb2 Nh8 45.a6 b6 46.Nd2 e4 47.Ra3 Rgf7 48.Rb3 Re8 49.Re3 Rff8 50.Bd1 Ng6 51.Bc2 Ne5 52.Ba1 Re7 53.Bb2 Be8 54.Bxe5 Rxe5

Here's the opinion (3 best moves) of Crystal3.1 about the position after 33... Rfg7 34. Nfg4 hxg4 35. fxg4
It likes Qg6 followed by f3

After Qg6 there will be Bxh4!. Now gxh4 seals the position.

After 35... Qg6 36. Bxh4, Crystal3.1 wants to play : f3 37. Nxf3 gxh4 38. Nxh4 Qxe4 39. Qxe4 Nxe4 with an eval around -2

After

37. Nxf3 gxh4 38. Nxh4 Qxe4 39. Qxe4 Nxe4 40. Nf5 Bxf5 41. Rxf5 Bg5 42. g3
I get a eval less then 1.00 after deep analysis.

After your line, Crystal got -3.00. Which engine(s) do you use ?

Code: Select all

FEN: 6rk/pp4r1/3p4/PPpPpRb1/2P1n1P1/R5PP/4B3/6K1 b - - 0 42

Crystal3.1-sse4-popcnt:
...
 41/70	4:16:05	82 149 166 199	5 346 450	-2,22	42. ... Re8 43.h4 Bc1 44.Rb3 Nd2 45.Rc3 Bb2 46.Rc2 Bd4+ 47.Kg2 Nb3 48.g5 e4 49.Bh5 Ree7 50.Rf6 Rh7 51.Re6 e3 52.a6 bxa6 53.bxa6 Nd2 54.Rc1 Rxe6 55.dxe6 Re7 56.Bg4 Kg8 57.h5 e2 58.Re1 Be3 59.g6 Bg5 60.Bf5 Nxc4 61.Kf3 Ne3 62.g4 Nd5 63.Rxe2 Nc7 64.Rh2 Bf6 65.h6 Nxe6 66.h7+ Kh8 67.Rh6 Nd4+ 68.Kg2 c4 69.g5 Bxg5
 41/53	4:16:05	82 149 166 199	5 346 450	-2,85	42. ... Rb8 43.Kg2 Bd2 44.Rd3 Bxa5 45.Re3 Nc3 46.Bd3 e4 47.Bxe4 Nxe4 48.Rxe4 Bc3 49.Re3 Bd4 50.Ra3 a6 51.b6 Re8 52.g5 Re5 53.Rxe5 dxe5 54.Rf3 Kg8 55.Rf6 Rxg5 56.d6 Rg7 57.Re6 Kf8 58.h4 a5 59.h5 Rd7 60.h6 Kf7 61.h7
 41/83	4:16:05	82 149 166 199	5 346 450	-3,00	42. ... Rd8 43.Kg2 Bd2 44.Ra2 Be3 45.Rc2 Bd4 46.Bd3 Nc3 47.Rh5+ Kg8 48.Bf5 e4 49.Rc1 e3 50.Re1 a6 51.b6 e2 52.Be6+ Kf8 53.h4 Ke7 54.Bf5 Be3 55.g5 Bd2 56.Kf2 Re8 57.Rh6 Kf8 58.Be6 Rf7+ 59.Bxf7 Nd1+ 60.Kg2 Bxe1 61.Bxe8 Bc3 62.Re6 e1Q 63.Rxe1 Bxe1 64.Bb5 Bxa5 65.Bxa6 bxa6 66.b7 Bc7 67.Kf3 Nb2 68.Kg4 Nxc4 69.h5 Ne5+ 70.Kf5 Kg8 71.h6 c4 72.g4 c3

Almost all engines i had specially SF NNUE. Although I use IDeA in aquarium i didnt find any way to penetrate the position.

[dkappe]

Almost all engines i had specially SF NNUE. Although I use IDeA in aquarium i didnt find any way to penetrate the position.

Try the night nurse or dark horse nets. Though not yet as strong as the SF nets, they definitely have a different opinion on many positions.

[Master Om]

Almost all engines i had specially SF NNUE. Although I use IDeA in aquarium i didnt find any way to penetrate the position.

Try the night nurse or dark horse nets. Though not yet as strong as the SF nets, they definitely have a different opinion on many positions.

Sure....

[Ozymandias]

Taking out really weak players (probably not even using a computer) the draw rate for the past three years is already close to 60%. If you take out weak players, like those who only use a computer to check some positions, it raises over 75%. We're still talking over 140,000 games for just three years, and the best ones from this period are likely to finish yet.

To get to 90%, you need to increase the strength of players to a point were only 65,000 games are left. To give you an idea, you still have close to 2,500 players at this level. These are serious players who care about results and have both access and willingness to use HW. This doesn't mean they use the latest, but that's not necessary to face a draw problem, and if you select the cream of the crop (more than 14,000 games and close to a thousand players) you surpass the 95% draw rate. This is why Arno wrote a proposal I submitted to you some months ago, you didn't need the advent of NNUEs to have a real problem in your hands.

I've also updated the playchess part of my DB and these are the stats since summer (about 8 months):

• 743,000 games

• 677 players

• 91.8% draw rate

While you didn't need SF-nnue to get draws everywhere in correspondence chess, it does have made them ubiquitous in the engine room.

[Milos]

Taking out really weak players (probably not even using a computer) the draw rate for the past three years is already close to 60%. If you take out weak players, like those who only use a computer to check some positions, it raises over 75%. We're still talking over 140,000 games for just three years, and the best ones from this period are likely to finish yet.

To get to 90%, you need to increase the strength of players to a point were only 65,000 games are left. To give you an idea, you still have close to 2,500 players at this level. These are serious players who care about results and have both access and willingness to use HW. This doesn't mean they use the latest, but that's not necessary to face a draw problem, and if you select the cream of the crop (more than 14,000 games and close to a thousand players) you surpass the 95% draw rate. This is why Arno wrote a proposal I submitted to you some months ago, you didn't need the advent of NNUEs to have a real problem in your hands.

I've also updated the playchess part of my DB and these are the stats since summer (about 8 months):

• 743,000 games

• 677 players

• 91.8% draw rate

While you didn't need SF-nnue to get draws everywhere in correspondence chess, it does have made them ubiquitous in the engine room.

If you take ICCF results from the past 9 months and exclude those with Elo below 2100, you'd get 97% draw rate.
And it is virtually impossible to score a win vs. unassisted SF-NNUE on a 300$ worth 8-core machine running 24/7 at correspondence TC.

[Ozymandias]

In the engine room, the vast majority of games are blitz (most 5m), that's the point we're at.

[Alayan]

Engine chess from the start position is pretty much dead.

[towforce]

I have lived through several era's ending: this feels like the beginning of the end of the computer chess era to me.

[MikeB]

Check this position.

[d]6rk/pp1bbr2/3p1n1q/PPpPp1pp/2P1Pp1n/R4P1P/4BNPN/1Q2BRK1 b - - 0 33

This is from my own game. This came out from a KID opening.
Just Check the eval here. Here my opponent who is an OTB GM played 33.Qb1
This is a corr game and u can see how helpless engines are here.

Now after deep analysis I found 33...Nxg2! can be played and then Bxh3!
But check the eval and the moves suggested...

[d]6rk/pp1bbr2/3p1n1q/PPpPp1pp/2P1Pp1n/R4P1P/4BNPN/1Q2BRK1 b - - 0 33

33... Nxg2 34. Kxg2 Bxh3+ 35. Nxh3 g4 36. fxg4 hxg4 37. Bxg4 Nxg4 38. Kh1 Rfg7
39. Nxg4 Rxg4 40. Qd1 Qh7 41. Qe2 Bh4 42. Bxh4 Rxh4 43. Rff3 Qg6 44. Ra1 Qh6
45. Raa3 Qg6 46. Ra1 Rg7 47. Qf2 Qh7 48. Raa3 Kg8 49. Qe1 Kf7 50. Qf2 Rhg4 51.
Qe3 Ke7 52. Qe1 *

If Black gives up 2 rooks for the queen Nxf4!! after doubling of rooks in f file makes the game draw.
But engines dont get it.

Now after 33...Nfg7 34.Nfg4!! hxg4 35. fxg4 the position is interesting.
See how no engines show a valid way to play this position.
After Bxh4 and Bf3 there is no way for black to penetrate white's camp.

Very nice position , my current net in process.
NNUE evaluation using C:\nnue-gui.1.5\reinforce-network\final\nn-2c35355d9b61.nnue enabled

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 46	+0.05 	2.03G	5:10.29	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rg3+ Kf7 Rh3 Qe6 Bh4 Bf8 a6 bxa6 Bf2 Qa2 Be1 Qa4 Rh4 Qxb5+ 
 45	+0.05 	1.80G	4:34.00	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rg3+ Kf7 Rh3 Qe6 Bh4 Bf8 a6 bxa6 Bf2 Qa2 Be1 Qa4 Rh4 Qxb5+ 
 44	+0.05 	1.78G	4:30.97	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rf3 Qe6 Re3 Kf7 Re2 Ke8 a6 b6 Kg2 Qg4+ Kh1 Qa4 Rg2 Qxb5
 43	+0.05 	1.78G	4:30.12	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rf3 Qe6 Re3 Kf7 Re2 Ke8 b6 axb6 axb6 Qg6 Ndf3 Qd3
 42	+0.05 	1.77G	4:29.63	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rf3 Qe6 Re3 Qa2 Ne2 Kf6 b6 Ke6 bxa7 Qxa5 a8=Q
 41	+0.05 	1.77G	4:29.26	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qh6 Rf3 Qe6 Re3 Kf7 Ngf3 Qa2 Ke2 Qa4 Kf1 Ke6 Re4 Qxb5+ Kg2 Qxa5
 40	+0.05 	1.76G	4:27.27	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Kg7 Nf3 Kf7 Nc3 Qc4+ Kf2 Qe6 b6 axb6 a6 bxa6 Re2 Kf8 Ra2 Qc4
 39	+0.04 	1.76G	4:26.90	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Kg7 Nf3 Qa2 Nd2 Qa1 Nc3 Bg5 Nb3 Qb2 Rg3 Qxb3 Rxg5+ Kf6
 38	+0.05 	1.75G	4:26.44	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Kg7 Nf3 Qa2 Nd2 Qe6 b6 a6 Re4 Kf7 Kg2 Qa2 Nc4 Qc2
 37	+0.05 	1.75G	4:26.31	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Qa2 Ne3 Rxe3 fxe3 Rf3 Rhg7 Qe2 Rg2 Qxg2 Rxg2 Kxg2 Qg6+ Kf1 Qxe4 Ng1 Qxc4+ Ne2 Qxd5 Rxe3 Qh1+ Ng1 Kg7 Nf3 Qh7 Nd2 Qc2 b6 axb6 a6 bxa6 Kg2 Kf7 Rg3 Ke6
 36	+0.04 	402.4M	1:02.44	Nxf3+ gxf3 Bxh3 Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Kh1 Rg3 Raf3 Qxh3+ Qh2 Qxh2+ Kxh2 Rg2+ Kh1 R8g5 Rh3+ Kg7 Rd1 R2g4 Kh2 Kg8 Re1 Rg2+ Kh1 Rg7 Rf1 b6 Rh2 R2g3
 35	+0.04 	330.3M	0:52.42	Nxf3+ gxf3 Bxh3 Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Kh1 Rg3 Raf3 Qxh3+ Qh2 Qxh2+ Kxh2 Rg2+ Kh1 R8g5 Rh3+ Kg7 Rd1 R2g4 Kh2 Kg8 Re1 Rg2+ Kh1 b6 axb6 axb6 Ra3
 34	+0.04 	305.4M	0:49.03	Nxf3+ gxf3 Bxh3 Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Kh1 Rg3 Raf3 Qxh3+ Qh2 Qxh2+ Kxh2 Rg2+ Kh1 R8g5 Rh3+ Kg7 Rd1 R2g4 Kh2 Kg8 Re1 Rg2+ Kh1 b6 axb6 axb6 Ra3
 33	+0.04 	172.8M	0:30.45	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Rff3 Qh7 Qb1 Rg4 Ra2 Qh5 Qf1 Rg1 Qf2 R8g7 Rxf4 exf4 Qxf4 Qg4 Qxg4 R1xg4 Nf2 Rg2+ Kh1 a6
 32	+0.04 	136.9M	0:25.45	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Rff3 Qh7 Qb1 Rg4 Ra2 R8g7 Qf1 Rg1 Qxg1 Rxg1 Kxg1 Qxe4 Raa3 Qd4+ Kg2 Kg8 Ng5 e4 Rxf4 Qb2+ Kf1 Qxa3
 31	+0.04 	121.2M	0:23.24	Nxg2 Kxg2 Bxh3+ Nxh3 g4 Nxf4 exf4 Rh1 gxf3+ Kxf3 Rfg7 Kf2 Rg2+ Kf1 Qg6 Bf3 Rxh2 Rxh2 Qg1+ Ke2 Qxh2+ Bf2 Ng4 Qg1 Qxg1 Bxg1 Kh7 Bxg4 hxg4 Kf2 a6 bxa6 Bh4+ Kf1 bxa6 Ke2 Bg3 Kf1
 31	+0.03!	113.4M	0:22.18	Nxg2!
 31	 -0.04?	103.9M	0:20.80	Nxg2 Kxg2?
 30	+0.03 	98.6M  	0:19.90	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Qh7 Nxg4 Rxg4 Qd3 Rfg7 Ra2 Bh4 Rh2 Bg3 Bxg3 fxg3 Rf8+ Rg8 Qf3 gxh2 Rxg8+ Qxg8 Nf2 Rg1+ Kxh2 Ra1 Qh3+ Qh7 a6 b6 Qxh7+ Kxh7
 29	+0.03 	83.7M  	0:17.44	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Qh7 Nxg4 Rxg4 Qd3 Rfg7 Ra2 Bh4 Rh2 Bg3 Bxg3 fxg3 Rf8+ Rg8 Qf3 gxh2 Rxg8+ Qxg8 Nf2 Rg1+ Kxh2 Ra1 Qh3+ Qh7 Qxh7+ Kxh7 Kh3 Ra4
 28	+0.02 	74.2M  	0:15.89	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Qh7 Nxg4 Rxg4 Qd3 Rfg7 Ra2 Bh4 Rh2 Bg3 Bxg3 Rxg3 Qc2 Qxh3 Rxh3+ Rxh3+ Qh2 Rxh2+ Kxh2 Rg3 Rg1 Rd3 Rg6 Re3 a6 b6
 27	+0.03 	71.0M  	0:15.36	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Rh7 Nxg4 Rxg4 Qc2 Bh4 Qe2 Rhg7 Bxh4 Rxh4 Rg1 Rxh3+ Rxh3 Qxh3+ Qh2 Rh7 Qxh3 Rxh3+ Kg2 Re3 Kf2 Kh7 b6 a6 Rg4 Rxe4 Rh4+ Kg7 Rg4+ Kh8
 26	+0.06 	64.8M  	0:14.46	Nxg2 Kxg2 Bxh3+ Nxh3 g4 fxg4 hxg4 Bxg4 Nxg4 Kh1 Nxh2 Kxh2 Rfg7 Qc2 Bh4 Bxh4 Qxh4 Rff3 Rg4 Qf2 Qh7 Qf1 Rg2+ Qxg2 Rxg2+ Kxg2 Qg6+ Kf2 Qxe4
 25	+0.07!	52.9M  	0:12.78	Nxg2!
 25	  0.00!	45.5M  	0:11.55	g4!
 24	 -0.09 	27.5M  	0:07.99	g4 fxg4 hxg4 hxg4 Rh7 Rh3 Nh5 Bd2 Qf8 Qd3 Ng3 Rxg3 fxg3 Qxg3 Ng6 Nh3 Rf7 Rxf7 Qxf7 Nf3 Kg7 Kh1 Bh4 Nxh4
 23	 -0.09 	22.6M  	0:06.98	g4 fxg4 hxg4 hxg4 Rh7 Rh3 Nh5 Bd2 Qf8 Re1 Ng3 Qd3 Nxe2+ Qxe2 Ng6 Rxh7+ Kxh7 Rf1 Bh4 Be1 Bxf2+ Qxf2 b6 Kh1 Ne7 Qa2
 22	 -0.09 	21.3M  	0:06.65	g4 fxg4 hxg4 Bxg4 Bxg4 hxg4 f3 Nxf3 Nxf3+ Rxf3 Rh7 Rh3 Qg5 Qb3 Rxh3 Qxh3+ Kg7 b6 axb6 Qb3 Nxg4 Nd1 Nf6 Qxb6 Qc1 Qxb7
 21	 -0.08 	12.8M  	0:04.43	g4 fxg4 hxg4 Nfxg4 Qg6 Bxh4 Nxg4 Nxg4 Bxh4 Bd1 f3 Bxf3 Qg5 Be2 Rxf1+ Qxf1 Qd2 Nh2 Bd8 a6 bxa6 Ng4 Bxg4
 20	 -0.09 	12.3M  	0:04.28	g4 fxg4 Rh7 Qb2 hxg4 Nhxg4 Nxg4 Nxg4 Bxg4 Bxg4 f3 Bd2 fxg2 Rf2 Qg7 Qb1 Bg5 Be6 Bxd2 Bxg8 Qxg8
 19	 -0.10 	6.57M  	0:02.34	g4 fxg4 hxg4 Nfxg4 Nxg4 Bxg4 Bxg4 Nxg4 Qh5 Rf2 Rxg4 hxg4 Qxg4 Qc2 f3 Bd2 Kg8 Qd3 a6 bxa6
 18	 -0.10 	5.05M  	0:01.85	g4 fxg4 hxg4 Bxg4 Bxg4 hxg4 f3 g3 Rh7 Qb3 Ng2 Nxf3 Nxe1 Rxe1 Nxg4 Kg2 Nxf2 Qa4
 17	 -0.11 	4.15M  	0:01.56	g4 fxg4 hxg4 Bxg4 Bxg4 Nfxg4 Nxg4 hxg4 Rh7 Rh3 Qg7 b6 a6 Qb2 f3 Qc2 fxg2 Rf2
 16	 -0.10 	3.98M  	0:01.51	g4 fxg4 hxg4 Nhxg4 Nxg4 Nxg4 Bxg4 hxg4 Rh7 Rh3 f3 Bxf3 Qe3+ Rf2 Nxf3+ Rxf3 Qd4 b6 Qxc4
 15	 -0.12 	2.04M  	0:00.87	g4 fxg4 hxg4 hxg4 Rh7 Rh3 Nh5 Qb3 Qg6 a6 Ng3 Rxg3 bxa6 bxa6 fxg3
 14	 -0.12 	1.44M  	0:00.66	g4 fxg4 hxg4 hxg4 Rh7 Rh3 f3 gxf3 Qe3 Qc2 Ng6 b6 Rxh3 bxa7 Rg3+ Kh1
 13	 -0.12 	1.35M  	0:00.62	g4 fxg4 hxg4 hxg4 Rh7 Rh3 Nh5 Qb3 Ng3 Rxg3 fxg3 Qxg3
 12	 -0.12 	939498	0:00.43	g4 fxg4 hxg4 Nhxg4 Nxg4 hxg4 a6 Qa2 Kh7 Bd2 axb5 cxb5
 11	 -0.12 	335788	0:00.16	g4 fxg4 hxg4 Nhxg4 Nxg4 hxg4 Qg6 a6 bxa6 b6 axb6 Rxa6
 10	 -0.12 	130104	0:00.06	g4 Nhxg4 hxg4 hxg4 Nxg4 Nxg4 Bxg4 a6 Rh7 Rb3
  9	 -0.11 	57418  	0:00.03	g4 fxg4 hxg4 Nhxg4 Nxg4 Bxg4 Qg5 a6 bxa6 bxa6
  8	 -0.12 	37143  	0:00.02	g4 fxg4 Nxg4 Bxg4 hxg4 Nhxg4 Bxg4 a6
  7	+0.13 	7124    	0:00.01	a6 b6 Bd8 Bd1 Ng6
  6	 -0.13 	5342    	0:00.01	a6 b6 Bd8 Bd1 Ng6 Ra1
  5	+0.13 	1648    	0:00.00	a6 g4 Ra8 gxh5
  4	 -0.13 	1028    	0:00.00	Nxg2 Kxg2 Rd8 b6
  3	+0.14 	319      	0:00.00	a6 bxa6 bxa6
  2	+0.51 	199      	0:00.00	Bd8 a6
  1	+0.70 	57        	0:00.00	Rff8

[Ozymandias]

If you take ICCF results from the past 9 months and exclude those with Elo below 2100, you'd get 97% draw rate.
And it is virtually impossible to score a win vs. unassisted SF-NNUE on a 300$ worth 8-core machine running 24/7 at correspondence TC.

With those games, taking as long as they do to complete, saying that a game finished in a given date doesn't indicate how much a particular engine was used. But as was I originally arguing, SF-nnue wasn't really needed for high draw rates at those TCs. Even Freestyle "died" somewhere between 2014 and 2017 (no tournaments in the interim) and that's played at a normal pace.

The engine room was "alive" a while longer, due to faster TCs, but NNUE did bring death shortly after SF12. Before last summer, draw rate for the top dogs was at 86.6% and the first implementation on SF-nnue only raised it to 88.2%, but since autumn, we're above 94%. Something which has been corroborated by the last two Engine Masters tours.