Positional-style chess engines

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

matejst
Posts: 368
Joined: Mon May 14, 2007 8:20 pm
Full name: Boban Stanojević

Re: Positional-style chess engines

Post by matejst »

Having really spent a lot of time analyzing with engines, trying on the board their lines, while I am not able to give a definition, I can say that some of them definitely feel more "human" than others. With a Hce, Komodo until K8, HIARCS, the last versions of Frenzee and Zarkov, the first version of Wasp, etc. In video courses I heard GMs mention many times that a move was "computer-like", say "a human move would be", so it is clear that they make the difference too.

I thought that NNs would make all engines play more like men, and I think it was true generally speaking. But even there, differences are huge. The Maia nets/engines are good, although they are not quite at the labeled level. The Harmon net should be used with Fire, not SF, and at shallower depths.

Of the others, the new Rebel feels quite human. I also like Lc0 CPU, and Winter. I did not trully tried Zahak's levels to be sure.

In general, human-like engines are designed by chess players who focused mainly on the evaluation. The nets created from these engines have similar characteristics. A slower search makes engine feel even more human.
User avatar
Rebel
Posts: 7476
Joined: Thu Aug 18, 2011 12:04 pm
Full name: Ed Schröder

Re: Positional-style chess engines

Post by Rebel »

mclane wrote: Sat Feb 05, 2022 10:43 amLC0
Yep.
90% of coding is debugging, the other 10% is writing bugs.
BrendanJNorman
Posts: 2589
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: Positional-style chess engines

Post by BrendanJNorman »

carldaman wrote: Tue Feb 08, 2022 4:01 am
dkappe wrote: Mon Feb 07, 2022 11:05 pm
carldaman wrote: Mon Feb 07, 2022 11:01 pm +1
One only has to look at the previous context and patterns...
There is such a thing as constructive criticism, but we're not seeing that here!
+1

Indeed. Brendan, you can do better than this. Maybe think a bit before sending off a nastygram without chess content.
Brendan can come on strong, but only if provoked. The poster he was replying to has a history of routinely putting down others.
Nice attempt at humor - but should we really placate the original instigator?!
Thanks for the more rational outlook mate.

Put it this way...

If you saw someone physically bullying someone on the street in real life, approached and was shown aggression yourself...

...and THEN responded with force...

Most would say "dude deserved it...brought it upon himself...he was looking for trouble" and others would be clutching pearls and pontificating about whether the force was "reasonable" or "appropriate" or whatever...

Victim-blaming in other words.

I'm the same online as I am in real life. If somebody is mocking and bringing others down or otherwise ruining someone else's experience, I'm gonna confront them. Period.

It's costs nothing for people (like Vernon) to just be nicer to people.

And anybody can see that my tone didn't change until he started bullying ppl.

And just to keep some chess content in here, here is an interesting positional game by an unreleased engine.

It features a strategic exchange sacrifice (...32 Rxd4!?) followed by a gradual turning of the screws until an inevitable tactical flourish ends the game.

Could both sides have played better? Sure. Is the exchange sac theme very interesting for *real* chess players? Definitely.

That's who we are.


[pgn][Event "Test Game"]
[Site "Shantou"]
[Date "2021.11.13"]
[Round "1.113"]
[White "Gandalf 4.32f"]
[Black "Unreleased Engine"]
[Result "0-1"]
[ECO "A46"]
[PlyCount "186"]
[EventDate "2021.??.??"]

1. d4 Nf6 2. Nf3 e6 3. e3 b6 4. Bd3 Bb7 5. O-O d5 6. c4 dxc4 7. Bxc4 a6 8. Nc3
b5 9. Be2 Nbd7 10. a3 c5 11. dxc5 Bxc5 12. b4 Be7 13. Bb2 O-O 14. Qb3 Qb8 15.
h3 Nb6 16. Rfd1 Rc8 17. Rac1 h6 18. Nd2 Bd6 19. Bd3 Rd8 20. Qc2 Rc8 21. Qb1
Bh2+ 22. Kh1 Bc7 23. Nce4 Nxe4 24. Nxe4 Na4 25. Bd4 Rd8 26. Nc5 Nxc5 27. bxc5
Bc6 28. Be4 Qb7 29. Bxc6 Qxc6 30. f4 Rd7 31. Rf1 Rad8 32. Kg1 Rxd4 33. exd4
Rxd4 34. f5 e5 35. Qb3 e4 36. Rfd1 Qd7 37. Qc2 e3 38. c6 Qd6 39. Rxd4 Qxd4 40.
Qe2 Qe5 41. g4 Bb6 42. Kh1 Qd6 43. Qd1 Qg3 44. Qg1 Qxh3+ 45. Qh2 Qf3+ 46. Qg2
Qf4 47. Qe2 Bc7 48. Kg1 Qg3+ 49. Kf1 Bb6 50. Rc2 Qf4+ 51. Kg1 h5 52. gxh5 Qg5+
53. Kh1 Qxf5 54. Kg2 Bc7 55. Rc1 Qg5+ 56. Kf1 Bb6 57. Rc2 Qd5 58. Kg1 Bc5 59.
c7 Qc6 60. Rc1 Qxc7 61. Qf3 Qe5 62. Kf1 Bxa3 63. Rc6 a5 64. Ra6 a4 65. h6 gxh6
66. Qg2+ Kh7 67. Qc2+ f5 68. Ra7+ Kg6 69. Qg2+ Kh5 70. Qh3+ Kg5 71. Qg2+ Kf4
72. Qh2+ Ke4 73. Qc2+ Kd4 74. Qd1+ Kc5 75. Ke2 f4 76. Kf1 Qe4 77. Rc7+ Kb6 78.
Rg7 Qh1+ 79. Rg1 Qh3+ 80. Ke2 f3+ 81. Kxe3 f2+ 82. Kf4 Qe3+ 83. Kxe3 fxg1=Q+
84. Kd2 Bb4+ 85. Kc1 Qxd1+ 86. Kxd1 h5 87. Kc1 h4 88. Kb1 h3 89. Ka1 h2 90. Ka2
h1=Q 91. Kb2 a3+ 92. Ka2 Qg2+ 93. Ka1 Qb2# 0-1

[/pgn]
Dayffd
Posts: 424
Joined: Wed Sep 30, 2009 5:30 am

Re: Positional-style chess engines

Post by Dayffd »

A lot of engines have been indicated here; would Chiron 5.01 fit in one of these lists?
David S.
Cornfed
Posts: 511
Joined: Sun Apr 26, 2020 11:40 pm
Full name: Brian D. Smith

Re: Positional-style chess engines

Post by Cornfed »

Vernon Crawford wrote: Mon Feb 07, 2022 5:14 am
picarito wrote: Fri Feb 04, 2022 4:08 pm Hello everyone, I would like to know your opinion about which are the positional style chess engines, it doesn't matter (a list if possible). Thanks in advance
The whole 'positional' thing is a purely human construct.
Maybe you mean 'human-like'...which certainly means 800-2000 Elo weaker (in line with human strength) than todays engines.
It's amusing to see engine authors (supported by their ignorant fan-boys) claiming their engine plays more human-like?

Please, lets step away from the 'style bashing' (of human words...not 'style' of engines).
Vernon makes a good point - are we referring to 'positional' or 'human like'? We should probably define our terms...I'm not sure anyone has done that so I can see why he headed this way.

If 'human like', then Vernon is probably right. Humans 'blunder', make poor choices, play poor positionally sometimes, are prone to missing even fairly simple tactical ideas....it varies and the average human is...well, probably around 1500.

If 'positional'...then to my mind you have to be looking at VERY low ply (to take away much of the tactical element)...say the famous 2 1/2 moves (5 ply) and a very good evaluation function.
supersharp77
Posts: 1266
Joined: Sat Jul 05, 2014 7:54 am
Location: Southwest USA

Re: Positional-style chess engines

Post by supersharp77 »

Rebel wrote: Mon Feb 07, 2022 4:27 pm
mclane wrote: Sat Feb 05, 2022 10:43 amLC0
Yep.
Well LC0....(MAYBE) I've personally watched Thousands of LC0 (Leela Zero...LeelaFish) games up to the current day and It's "Style" I find a bit confusing at Times...Is It Positional..Or a Positional Attacker? Most of What it tries to do seems to end with a Kingside Attack Or In some sort Of Zugzwang Sort Of Position...Don't see The Queenside Positional Attacks Much...I see excellent Central control and Great attacking ideas with LC0...No Doubts!! :) :wink: