EN-Test 2022 - new testsuite

[DrEinstein]

Nagging?
OK, I'll ask elsewhere.
Btw: Now I have an idea why no SFdev is writing here

[Eduard]

You want to know everything in detail. Why so and so... I do my stuff for my fun! Not for science. I don't care what the programmers think. Can't you just say something about the test? You can download the test and all TXT files. you can compare You can study the 3 posted games. And say something about it. You do not do that. You overwhelm me when I have to look for a scientific explanation for everything. But you don't understand that. You don't care about having fun. That's your problem. I have fun and that's the most important thing to me. If you are looking for science then go to Discord and discuss with the professionals there.

[DrEinstein]

I AM a scientist, thus the nick, and I can easily see from your posts that you have a lot of fun with these engines. Me too

[Eduard]

It's nice that everyone can do what they like and have fun with the Stockfish code.

[Eduard]

One more word. In addition to the EN test 2022, the engine is also checked with other known test suites. In addition, a friend of mine tests each engine live on PlayChess for 5 days non-stop with 64 cores. Everything is precisely documented. I don't know anyone else who tests every engine with 64 cores so intensively.

[DrEinstein]

Once more
The engine 'of' a novice programmer in first place and no reaction of any professional SF developer.
WHY?

[Eduard]

In fact, almost all clones score better than Stockfish dev in my test. This is probably because the developers only test with bullet games. The rest is unimportant for the developers. The Fishtest is like a program. Everything runs automatically. Every new code and every parameter change is automatically checked with the Bullet Fishtest.

Example: new parameter value 2, 4, 8, 16, 32. Each value is checked with 40000 bullet games. The best value is the new king.

[Eduard]

Haha, that's funny. A few days ago I saw this code on Fishcooking:

// Adjust full depth search based on LMR results - if result
// was good enough search deeper, if it was bad enough search shallower
const bool doDeeperSearch = value > (alpha + 64 + 11 * (newDepth - d));
- value = -search<NonPV>(pos, ss+1, -(alpha+1), -alpha, newDepth + doDeeperSearch, !cutNode);
+ const bool doShallowerSearch = value < bestValue + newDepth;
+ value = -search<NonPV>(pos, ss+1, -(alpha+1), -alpha, newDepth + doDeeperSearch - doShallowerSearch, !cutNode);

Today: NEW Master - Stockfish Version!

The new code has been partly implemented in Leptir 4 for days. I also document all changes.

Leptir 4:
(---)
+ value = -search<NonPV>(pos, ss+1, -(alpha+1), -alpha, std::max(1, newDepth + doDeeperSearch - doShallowerSearch), !cutNode);

I tested this code in my test, and on PlayChess, and Leptir 4 is the new #1 in my test (115 soluations at 60s) - and before the code was implemented in Stockfish Master.
https://solistachess.jimdosite.com/testing/

[DrEinstein]

In fact, almost all clones score better than Stockfish dev in my test. This is probably because the developers only test with bullet games. The rest is unimportant for the developers. The Fishtest is like a program. Everything runs automatically. Every new code and every parameter change is automatically checked with the Bullet Fishtest.

Example: new parameter value 2, 4, 8, 16, 32. Each value is checked with 40000 bullet games. The best value is the new king.

You have a lot of fun and you know what's going on on fishtest. But assume that the 'killer value' is 25. It will never be found. And now?
And even worse, assume you have several such parameters and all are independent or dependent of each other... 10^12 games or what?

[Eduard]

Just tested, Stockfish dev-20221119:

AMD Ryzen 3900X. 20 Threads, Hash 4 GB, all 3456men Syzygy. Stockfish dev-20221119:

Solving: E:\Database\EN-Test 2022.cbh
Maximum time = 60s.

Stockfish dev-20221119, Result: 99 out of 120 = 82.5%. Average time = 3.93s / 20.57

(All text files on my homepage)

Every newer clone is now stronger! Stockfish from 05.11.22 solves 101 positions.

live:



Which positions have not been solved by Stockfish dev 20221119?


[fen]r1n2N1k/2n2K1p/3pp3/5Pp1/b5R1/8/1PPP4/8 w - - 0 1[/fen]
2. EN 002 (ERET 002) - Zugzwang, Ng6+, Akerblom 1961 > 60s.



[fen]1nb3rr/2pp3p/p1k2P2/n1q1PN2/P1pR3N/b4PBp/1R2P3/1KQ5 w - - 0 1[/fen]
14. EN 014 (TacticalInsanityNr.9), Rd6 > 60s.



[fen]1r3rk1/6p1/p1pb1qPp/3p4/4nPR1/2N4Q/PPP4P/2K1BR2 b - - 0 24[/fen]
17. EN 017 (ERET 017) - Koenigsangriff, Rxb2, Boudy-Szabo, Varna 1979 > 60s.



[fen]1r1q1rk1/1p2b2p/pN6/P1pP1bp1/5p2/2B2Q2/1P4PP/R4RK1 w - - 0 1[/fen]
18. EN 018 (HTC10) - Caruana vs Carlsen, g4 > 60s.



[fen]1rb1qrk1/2b2pp1/p3pBn1/3pP1Pp/1ppP4/2P1QN2/PP3P1P/R2BR1K1 w - - 0 1[/fen]
32. EN 032 (CCC._SF CSS._HTC108), Nh4 > 60s.



[fen]2q2r1k/p1p1b1pp/br1p1n2/2p1N1BP/Q1P1P3/8/PP1N1PP1/2KR3R w - - 0 1
[/fen]36. EN 036 (Gashimov), Ng6+ > 60s.



[fen]2r2r1k/qb3p1p/p2p1Np1/np1Pp1P1/2P5/3RP2P/5R2/4Q1K1 w - - 0 1[/fen]
37. EN 037 (CCC+Rybkaf_?_HTC108), Rf4 > 60s.



[fen]2r5/prkpR1p1/2p1ppK1/P1p1N1B1/P1P1P3/8/8/8 w - - 0 1[/fen]
40. EN 040 (Bondarenko&Kuznezov), Bf4 > 60s.



[fen]4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - 0 1[/fen]
44. EN 044 (Gusev), Qxe5 > 60s.



[fen]6br/1KNp1n1r/2p2p2/P1ppRP2/1kP3pP/3PBB2/PN1P4/8 w - - 0 1[/fen]
45. EN 045 (HTC108), Bxc5+ > 60s.



[fen]8/4nk2/1p3p2/1r1p2pp/1P1R1N1P/6P1/3KPP2/8 w - - 0 50[/fen]
46. EN 046 (ERET 046) - Endspiel T&S vs T&S, Nd3, Grandelius-Raznikov, Albena 2011 > 60s.



[fen]r5rk/1bq1bpp1/1p1ppn1p/pP5P/2P1P3/P1N2P2/1B1Q1P2/1K1R1BR1 w - - 0 1[/fen]
60. EN 060 (MorozevichA._Vitiugov, Nd5 > 60s.



[fen]2b5/1pr4p/3bp1pk/1p6/1PpN2PP/K1P1n3/P3N1R1/3R4 w - - 0 1[/fen]
77. EN 077 (Mihai ENeghina), Nxb5 > 60s.



[fen]3r2k1/p4pP1/1ppr4/5Bp1/P2qPnQP/4R3/5P2/5RK1 w - - 0 1[/fen]
79. EN 79 (Tauber vs Sikorsky), e5 > 60s.



[fen]r1bqkb1r/3n1ppp/p3p3/8/Pp1B1n2/3B1N2/1P1NQPPP/R2R2K1 w kq - 0 1[/fen]
84. EN 084 (Sikorsky Horst), Qe1 > 60s.



[fen]4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - 0 1
[/fen]90. EN 090 (ERET 090) - Endspiel L vs L, f4, Salai 2011 > 60s.



[fen]rnbq3r/1p2b1k1/2pp1p1p/pP3pp1/2BP3N/P3P3/2P2PPP/RN1Q1RK1 w - - 0 1[/fen]
93. EN 093 (TTT2.053), Ng6 > 60s.



[fen]2k2Br1/p6b/Pq1r4/1p2p1b1/1Ppp2p1/Q1P3N1/5RPP/R3N1K1 b - - 0 31[/fen]
94. EN 094 (ERET 094) - Damenopfer, Rf6, Sirobaba-Chaika, CorrGame 2011 > 60s.



[fen]5b2/5k1p/1pb2pp1/1r1p4/3P1B1P/P7/1P3PP1/2R3K1 w - - 0 28[/fen]
114. EN 114 - Lc0 vs Stockfish, Onlinegame, b4! > 60s.



[fen]2kr4/pp1n1pBp/3bp1p1/3p2qP/P7/3B1K2/1PP2P2/R2Q1R2 b - - 0 24[/fen]
117. 117 Bestmove - Online Engine game Solista, f6! > 60s.



[fen]r4rk1/1b4pp/4pq2/PB2n3/3N1Pn1/1PN3K1/1B2Q1P1/R4R2 b - - 0 1[/fen]
119. EN 119 Dorsz - BlueMarlin vs Stockfish, Qh6! > 60s.