Testing LazySMP

[LazySMP]

Concluding, this all may seem quite critical and negative to you, but i'm only
trying to tell you to rethink your approach for
example, if you have a better/improved move generation algorithm, then
why not eg. mail Lombardo, the programmer of Obsidian, he then might
get ahead of SF, for a change; ofcourse it wouldn't be kept secret, like a
private engine but who cares;

Thanks for your honest feedback. This all doesn't sound entirely critical or negative to me. Once we become predictable, we become vulnerable.
I just tried to develop an engine and it doesn't matter that they don't want to test my engine. I continue on my way... Good Luck

P.S. If you are interested in the topic you may also want to see this: http://talkchess.com/viewtopic.php?t=84382

[LazySMP]

So, let's see what we have in the meantime.

LazySmp 3.0 - NPS 8.3 Million from the start position, elo 1500-1600 (understandable for a PST engine only)
LazySmp 5.0 - NPS 4.1 Million from the start position, elo 2085 CCRL (a 500 elo improvement in a short time)
LazySmp 9.0 - NPS 2.9 Million from the start position, elo 2700+ based on my testing.

Code: Select all

No. Name Win Draw Loss Unf. Score Games %
--------------------------------------------------------
1 LazySMP-9.0 +419 =224 -357 *0 531.0 1000 53.1%
2 CT800-1.46  +357 =224 -419 *0 469.0 1000 46.9%

CT800-1.46 is 2700+ CCRL rated.

Hi Ed! Thank you for testing my engine, you probably haven't read the change of version 9.0.

Code: Select all

[b]Changes of 9.0 compared to 8.0:[/b]

Remove LazySMP Algorithm
Add Quiescence Search
Improve Evaluation For Pawns, Mobility and King safety
Fix Some UCI Bugs

Also against Fruit 2.1 (2700 elo) scores 52.9%

Meaning in a couple of weeks you gained 600+ elo, unheard for HCE programming.

How does this all look Daniel?

As you probably know, The horizon effect is a problem in many games, the number of positions is immense and engine can only feasibly search a few plies down the game tree. The horizon effect can be mitigated by extending the search algorithm with a quiescence search.

The main reason for gaining 500 Elo rating is the addition of Quiescence Search.

It looks like you took (or maybe you are the author yourself of) a strong UCI engine, somewhere in the range of 3000-3600 elo, locked all the elo stuff till it was PST only (v3), then unlocked elo stuff, of course at the cost of the NPS, and gained 500 elo (v5), tested by CCRL. And with v9 you unlocked more elo stuff, gaining 600 elo, at the cost of a further dropping NPS.

No, I didn't lock anything, I added Mobility in version 5.0 and king safely in version 7.0. This is exactly what has caused the NPS to slow down.

Next step elo 3000? Unlock Hash tables? You have said the needed Zobrist tables are already in LazySMP.

No, Zobrist key are not only used for Transposition Tables. Please read HGM comment:

Why, since you have stated not to make use of a hash table.

TSCP updates a Zobrist key, even though it doesn't have a transposition table. It uses it for the purpose of detecting repetitions.

[LazySMP]

As I had suggested earlier, if you make a linux executable, I am willing to run a bunch of tests in February. Are you using some specific technologies or frameworks or a language that you cannot make it more portable?

Honestly, I use Visual Studio which has MSVC compiler. I'm not sure if my compilation can work properly on Linux.

JikChess 0.02 64-bit has 2520 Elo points in CCRL Blitz rating list. Again, LazySMP 9.0 is a good chess engine.

A question : a bmi2 compile of this version (if possible) ?

You can download LazySMP 9.1 from this link:



I hope my competitors don't want me to report this link

[LazySMP]

So, let's see what we have in the meantime.

LazySmp 3.0 - NPS 8.3 Million from the start position, elo 1500-1600 (understandable for a PST engine only)
LazySmp 5.0 - NPS 4.1 Million from the start position, elo 2085 CCRL (a 500 elo improvement in a short time)
LazySmp 9.0 - NPS 2.9 Million from the start position, elo 2700+ based on my testing.

I should add some more points to Ed's incorrect analysis:

1. LazySMP 3.0 is stronger than Lazy 5.0 in blitz game
2. You forgot to test version 7.0 which is stronger.
3. My engine rating is current at 2500 not 2700
4. Just playing with an engine doesn't determine my engine rating.

King safety has always been an engine weakness, so losses like this can happen, and if they do, the game is shorter naturally because it doesn't go into an endgame. However, it's not about how quickly the engine loses if it does, but how often, with varying openings of course. If you say LazySMP 9.0 should be around 2500, that's 200 Elo difference, so LazySMP should get about 2500 points in a 10000 games match - including quite some wins, obviously.

The real info is this:

LazySmp 3.0 - NPS 8.3 Million from the start position, elo 1900 (understandable for a PST engine only)
LazySmp 5.0 - NPS 5.1 Million from the start position, elo 2085 (a 185 elo improvement in a short time)
LazySmp 7.0 - NPS 3.8 Million from the start position, elo 2200+ based on my testing.
LazySmp 9.0 - NPS 2.9 Million from the start position, elo 2500+ based on my testing.

There is only +300 Elo at LazySMP 9.0 due to the implementation of the Quiescence function.

Where will it end? Next step elo 3000?

P.S. Are you worried that my engine is over 2500? What are you trying to prove? Go and bring a strong reason, not childish reasons.

[LazySMP]

[pgn]
[Event "Blitz 4.0min+2.0sec"]
[Date "2025.01.06"]
[Round "1"]
[White "LazySMP Version 9.0"]
[Black "Berserk 10"]
[Result "0-1"]
[Variant "Standard"]
[TimeControl "240+2"]
[ECO "B53"]
[Opening "Sicilian Defense: Chekhover Variation"]

1. e4 { [%clk 0:05:14] } c5 2. Nf3 d6 3. d4 cxd4 4. Qxd4 { B53 Sicilian Defense: Chekhover Variation } Nf6 5. e5 Nc6 6. Bb5 Qa5+ 7. Nc3 Qxb5 8. Nxb5 Nxd4 9. Nfxd4 dxe5 10. Nc7+ Kd7 11. Nxa8 exd4 12. Bf4 e5 13. Bxe5 Bd6 14. f4 Kc6 15. O-O-O Be6 16. Bxd6 Kxd6 17. Rxd4+ Kc5 18. Rhd1 Rxa8 19. h3 h5 20. R1d3 a5 21. a3 a4 22. Rd1 g6 23. Kb1 b5 24. h4 Re8 25. Rd8 Rxd8 26. Rxd8 Nd5 27. g3 Kd4 28. Rb8 Ke3 29. Rxb5 Kf3 30. c4 Ne3 31. c5 Kxg3 32. c6 Nd5 33. Rc5 Kxh4 34. c7 Nxc7 35. Rxc7 Kg3 36. Kc1 h4 37. Rc3+ Kxf4 38. Rd3 g5 { White resigns. } 0-1
[/pgn]

This shows that my engine is not that strong. So your intention is that my engine will not be tested. I remember when I released version 5 you tried to show that my engine was weak so that it would not be tested and now that version 9 has become strong, you are trying to make it look bad. Maybe I expected this from the community members but I never expected the moderators to want to slander.

[LazySMP]

So, Due to the accusations and insults that were made against me, I leave Talkchess forever. Good Bye!

[Modern Times]

So, Due to the accusations and insults that were made against me, I leave Talkchess forever. Good Bye!

Hmm.... how many times have you said that before

[lucametehau]

So, Due to the accusations and insults that were made against me, I leave Talkchess forever. Good Bye!

On god, please do this time

[Daniel Anulliero]

So, Due to the accusations and insults that were made against me, I leave Talkchess forever. Good Bye!

Bye bye for good this time ?

[sarona]

So, Due to the accusations and insults that were made against me, I leave Talkchess forever. Good Bye!

Promises, promises. . .