Stockfish builds

2022-01-21 06:53:01
ShashChess 20.2 vs Lichess
35 Ply fixed Depth
512 Hash
8 threads

UCI Options:
Tal (Attack position/algorithm)

1. e4 c5
2. Nf3 d6
3. d4 cxd4
4. Nxd4 Nf6
5. Nc3 a6
6. Be3 e5
7. Nb3 Be7
8. h3 O-O (Bc8-e6)
9. g4 h6
10. g5 hxg5
11. Bxg5 Be6
12. Qf3 Nbd7
13. Rg1 Re8
14. O-O-O b5
15. Kb1 Qc7
16. h4 Rec8
17. Bd3 Bc4
18. Bh6 Bf8
19. Rxg7+ Bxg7
20. Rg1 b4
21. Bxg7 Bxd3
22. Bh6+ Kh8
23. Qg2 Bxc2+
24. Ka1 Nh5
25. Bg7+ Kg8
26. Bxe5+ Kh7
27. Qg5 Ndf6
28. Qf5+ Kh6
29. Bf4+ Nxf4
30. Qg5+ Kh7
31. Qg7# 1-0

I like how ShashChess just pierces through defenses. Gives me the feeling that no defense can stop it, which is not the case ofc. :D
2022-01-20 17:59:31
Past 4 builds are either no functional change, or only affect timed game performance. No new improvements for infinite analysis since Jan 13th release.
2022-01-20 15:18:58
Another example

[Event "Alekhine-Botvinnik match"]
[Site "?"]
[Date "2021.11.29"]
[Round "1"]
[White "Alekhine, Comp"]
[Black "Botvinnik, Comp"]
[Result "1-0"]
[ECO "E34"]
[PlyCount "103"]

1. d4 Nf6 2. c4 e6 3. Nc3 Bb4 4. Qc2 d5
5. cxd5 Qxd5 6. e3 c5 7. a3 Bxc3+ 8. bxc3 O-O 9. Nf3 cxd4 10. cxd4 b6 11. Bc4
Qc6 12. Bd2 Ba6 13. Rc1 Rc8 14. Bxa6 Qxc2 15. Bxc8 Qe4 16. O-O Nbd7 17. Rc7 Nd5
18. Bb7 Rd8 19. Rfc1 f6 20. Bb4 Nf8 21. Nd2 Qd3 22. Bxd5 exd5 23. Rxa7 Re8 24.
h3 h5 25. Nf3 Ne6 26. Be7 b5 27. h4 Qe2 28. Rc6 Nf4 29. Bxf6 gxf6 30. exf4 Re7
31. Raa6 Rf7 32. Rxf6 Rxf6 33. Rxf6 Qd1+ 34. Kh2 Qc1 35. Rf5 Qxa3 36. Rg5+ Kf7
37. Rxh5 Qc1 38. Ne5+ Kf6 39. Ng4+ Kf7 40. Rf5+ Kg8 41. Rg5+ Kh8 42. f5 b4 43.
f6 Qf4+ 44. g3 Qf3 45. Kg1 b3 46. Rh5+ Kg8 47. f7+ Kg7 48. Rg5+ Kh7 49. Ne5 Qf6
50. Rg6 b2 51. Rxf6 b1=Q+ 52. Kg2 1-0
33270dave F
2022-01-19 20:00:43
If anyone wants to just see the ending combo, here is the Forsyth-Edwards Notion of the position where the combo begins.


34. Qg4!! Bh2+
35. Kh1 Rdd8
36. Re8! Rdxe8
37. Qg7+!! Rxg7
38. Rxe8+ Rg8
39. Rxg8#
2022-01-19 19:32:26
I love the ending combination in this game. White offers his Queen twice and the 2nd time it HAS to be taken. White: Stockfish 14.1 depth about 30. Millenium "Genius" 2 minutes a move. Hungarian Defense Giuoco Pianissimo C50.

1. e4 e5
2. Bc4 Nc6
3. Nf3 Be7
4. d4 exd4
5. Nxd4 d6
6. O-O Nf6
7. Nc3 O-O
8. Bf4 Ne5 (last book move)
9. Bb3 Bd7
10. a4 Qc8

11. Qe2 a6
12. h3 c5
13. Bxe5 dxe5
14. Nf3 Be6
15. Nxe5 Bxb3
16. cxb3 Qe6
17. f4 Rad8
18. Nc4 Rd4?! (looks good but is actually weak. Instead, ...b5 and then Qd7)
19. e5 Nd5
20. f5 Qc6

21. Nxd5 Rxd5
22. a5 Rfd8
23. f6 gxf6
24. Qf2 Kg7? (better is Kh8)
25. Rae1 Rg8
26. Ne3 Rd7
27. Ng4 Kh8?! (It's hard to suggest anything but SF recommends ...f5, Qxf5 Rd2)
28. exf6 Bd6
29. Nh6 Rc8
30. Qh4 Bf8? (Qd5)

31. Rf4 Qd5
32. Rg4 Bd6
33. Rge4 Rg8 Now comes the real fun!
34. Qg4!! Bh2+ (he has to do this to clear the d file. This helps defend the back rank.
35. Kh1 Rdd8
36. Re8! Rdxe8
37. Qg7+!! Rxg7
38. Rxe8+ Rg8
39. Rxg8#
33268dave F
2022-01-19 18:46:31

Poor Black got behind in development. When the pawn shield in front of his King was gone and White had all of his heavy pieces bearing down, the outcome was no longer in doubt. Nice King side attack too.

Thanks for posting!
33267dave F
2022-01-19 18:35:59
Nice Game! Poor Black was teetering on the edge of disaster for about the last dozen moves. A Queen sac is always satisfying to see, even if it is a "simple" mate in 2.

Thanks for taking the time and effort to post that game.
2022-01-18 15:44:30
A weaker program will easily lose to a stronger program. It is interesting to see impressive victories in a game of equal strength opponents. Take a look at this game from the Morphy-Steinitz match simulation. Morphy chooses Evans' gambit. After 20 moves the position seemed balanced and the king's move to h8 safe, but white ingeniously developed his forces and black faced defeat.

[Event "Morphy-Steinitz match"]
[Site "?"]
[Date "2021.11.19"]
[Round "1"]
[White "Morphy, Comp"]
[Black "Steinitz, Comp"]
[Result "1-0"]
[ECO "C52"]
[PlyCount "61"]

1. e4 e5 2. Nf3 Nc6 3. Bc4 Bc5 4. b4
Bxb4 5. c3 Ba5 6. d4 exd4 7. O-O dxc3 8. Qb3 Qf6 9. Bg5 Qg6 10. Nxc3 Bxc3 11.
Qxc3 Nge7 12. Bxe7 Kxe7 13. e5 Re8 14. Bd3 Qh6 15. Rae1 Kf8 16. Re4 g6 17. Rfe1
Qg7 18. Qc1 Kg8 19. h4 h6 20. R1e3 Kh8 21. h5 g5 22. Nxg5 hxg5 23. h6 Qf8 24. Rg3 d5
25. Qxg5 f5 26. Rh4 Nxe5 27. Qg7+ Qxg7 28. hxg7+ Kg8 29. Rh8+ Kf7 30. Rxe8 Kxe8
31. g8=Q+ 1-0
2022-01-18 05:16:07
ShashChess 20.1 vs Lichess (computer difficulty 8)
30ply fixed depth (per move).
512mb Hash
8 threads

UCI options:
1 thread assigned to - Full depth threads.
1 thread assigned to - MCTS. MCTS: Single (Monte-Carlo Tree Search)
6 threads default normal processing.

1. e4 c5
2. Nf3 e6
3. Nc3 Nc6
4. d4 c5xd4
5. Nxd4 Qc7
6. Be3 a6
7. Qd2 Nf6
8. f4 Bb4
9. Bd3 e5
10. Nf5 0-0
11. f4xe5 Ng4
12. Bg5 Ng4xe5
13. Bf6 Nd4
14. Nxg7 d6
15. 0-0 Ne6
16. Nf5 Rd8
17. Qh6 Bxc3
18. b2xc3 Ng6
19. Rf3 Qc5+
20. Kh1 Qxc3
21. Ne7+ Nxe7
22. Bxc3 Ng6
23. Bf6 b5
24. Rh3 Ng6-f8
25. e5 Nf8-g6
26. Qxh7 Kf8
27. Qh8+ Nxh8
28. Rxh8#
2022-01-18 04:26:07
@ dave F, from my experience Honey 14 beats all. Casually on bullet times, not 1000s of games.
I noticed sf-tester has ShashChess on top, but in my games Honey beats even current (new) Shash20.2.

Sure i'll post it. It reminds me of Glaurung game that i had a few years ago. Glaurung was known to be kinda sacrificial. Ofc nowdays NNUE engines are waay stronger now. :)
33263dave F
2022-01-18 00:36:27
NN123 - Cool! I agree that a 28 move win against a top engine is impressive. There are several engines based on Stockfish out there. I've tried some quick games with a few of those clones against SF and there wasn't a discernable gap in strength. I didn't try a large sample of games though. The results were mostly draws!
I know it's a hassle to type out whole games but if you have time and the motivation, I would like to see that game you mentioned.
2022-01-17 21:59:12
dave F
I have a beautiful game, where ShashChess won against Lichess (online computer) in 28 moves.
But that's not Stockfish, it's SF derivative so maybe a bit offtopic.
I mentioned it cause it's a beautiful game.

Note that lichess is pretty strong, so to win against it in 28 moves is pretty awesome and really rare. It doesn't stand a chance against top engines, but anything below that it's pretty equal in strength.
33261dave F
2022-01-17 20:38:40
Here is one of the shorter ones. 32 moves. Philidor Defense C41 White; Millenium 2m a move. Black: Stockfish 14.1 on 4 cores, depth generally between 25 and 33 moves.
1. e4 d6
2. d4 Nf6
3. Nc3 e5
4. Nf3 Nbd7
5. Be2 c6
6. a4 (last book move) h6
7. o-o Be7
8. Qd3 (a5) exd4
9. Qxd4 o-o
10. Rd1 Re8

11. a5 Bf8
12. Bc4 Qe7
13. Nh4?! (Re1) Ne5 (nicely timed)
14. Bb3? (Bf1) Nh5 (the last 2 moves illustrates Millenium's tendency towards mistakes)
15. Nf5 Bxf5
16. exf5 d5
17. Be3?! (Bf1) Qd7
18. Ne2 Qxf5
19. h3 a6
20. Qb6 R(a)d8

21. Qxb7? (not a good time to go pawn grabbing!) Nf3+!
22. gxf3 Rd6
23. Kf1 Rxe3!! (a nice sacrifice plotted along with his 21...Nf3+!)
24. fxe3 Qxf3+ (White's king is in deep peril)
25. Ke1 Ng3!
26. Nxg3 (Kd2 holds out a little longer) Qxe3+
27. Ne2 Rg6 (Stockfish is down a rook and a night and he's got an announced mate!)
28. Kf1 Rf6+
29. Ke1 Rf2
30. Rd2 Rh2
31. Kd1 (Millenium isn't programmed to resign. Obviously he should have done this by now) Rh1+
32. Ng1 Rxg1#

I like the way Stockfish slices through Millenium's position. He does it quickly and precisely.
2022-01-17 14:11:36
@dave F
Please show some example
2022-01-14 13:50:47
Elo +460 reached sf dev builds Rui Coelho version in NCM test
33258dave F
2022-01-12 23:56:08
I've been running a series of long games with Stockfish against Millenium's "Chess Genius". Of course, the Millenium program is being consistently overrun by Stockfish BUT, having said that part, some of the games are very beautiful. Many of the games have a human quality and remind me of games from the golden age out of the Russian school from the 1940 to 1970 period. (Tal, Botvinnik, Kotov, Keres, Smyslov, Bronstein, etc.)

The games feature sacrificial attacks, some of which are mind boggling and seem to come conjured out of pure chess sorcery. I think the reason this happens is because Millenium has some small weaknesses and Stockfish always exploits these little mistakes with swift retribution. You don't see long boring games and these contests are usually between 30 and 50 moves.

I mention this here because these games have given me a lot of pleasure to watch. Stockfish never lets me down due to the deep positional and tactical play the program is so wonderfully capable of producing. I've even thought of compiling the best games in a book. If there is any interest in seeing some examples, just post a comment about it and I'll type out some of the best games and put them up here.
2022-01-12 09:24:14
In my case, it seems to overtax my system using more cores and I think while using ponder I should probably not use more than 4 cores since both engines would effectively max my pc.  However, I started using 6 cores for analysis (single) and simply haven't changed since everything seems to be running fine.  Quite possibly, the engine competitions may run better with with just 4 cores ... maybe I will give that a curious try.
2022-01-12 08:23:06
@Kingpawn With 8 cores (16 threads), why not go 8 threads per side when using ponder?
2022-01-11 23:53:30
I mistyped the timestamp ... should be 1640157515 date 21122208
2022-01-11 19:05:30
These were a series of hundreds of blitz+1 games with mostly selected difficult openings often leading to decisive games.  Games played using Aquarium, AMD Ryzen 7 5800X 8-Core Processor, 6 threads, ponder=yes and no book.  Engines include latest Stockfish Development Versions plus latest hybrids Kayra and others.  Openings included, random, Falkbeer Countergambit C31, Latvian gambit, King's Gambit - Keres, French Advance -C02e, Two Knights (Fried Liver) C57e, and C56a, Russian Cochrane Gambit C42h and Bird’s From Gambit Lasker Variation A02.  Some are very wild openings but engine testers.
2022-01-11 08:34:19
@Kingpawn, what are your testing parameters?
2022-01-10 19:41:39
Constantly pitting top engines ... timestamp 1640167354 consistently stronger than any SF engine to date
2022-01-10 18:28:04
>  but also the result of ultra-fast bullet time controls using just 2 cores.
the same ultra bullet TC does not produce "noise losses" if not run in insanely multi-threaded unfair hardware environment with dozens tournaments run in parallel on the same CPU. I cannot reproduce losses even in bullet TC
2022-01-09 13:30:58
Ok, so it was only 50 games...

score 13-9 which looks normal.

Thanks very much SamS! I always forget how to do it !
2022-01-08 23:06:17
What about  stockfish 15? Any news??

next page >