OpenTal - almost new engine

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

Moderators: hgm, Rebel, chrisw

BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman »

PK wrote:Since the next Rodent release will be delayed, I decided to present you a program with a proud name "OpenTal 1.0"

http://www.pkoziol.cal24.pl/opental/

In short, this is a shameless Rodent clone that does not expose tuning options to the user. It plays in one style only. To make matters worse, it uses internal opening book before main book. And the entire project hinges on evaluation weights kindly provided by Brendan J. Norman.

These weights are magic. OpenTal still plays at about 2500-2600 Elo, despite sacrificing pieces right and left. Despite apparent craziness it managed to beat RedQueen 1.1.98 in a short match. It won a couple of games with a direct attack, plus one crazy endgame rook down. It also lost a few because of unsound sacrifices.

OpenTal will lose to anything over 2800 Elo, especially untweaked Rodent manhandles it really badly. But this is not an opponent for engines. This is an opponent for you!
Such an exciting release, thanks Pawel!

This is an engine that emulates style of Tal more than any other I've seen...

More than ProDeo Tal, Szint Tal, Homer Tal, Chess System Tal or anything!

With regard to playing strength, the "weak play" against other engines is deliberate.

If you check Tal's games - even his peak strength games from 1959-60 when he was world champ - with an engine (something that strangely, not many personality creators do), you'll see that many, many, many of his sacrifices were objectively unsound and that the level of his opponent's defense was very poor.

That human Tal too, if facing an engine of 2500-2600 CCRL would also be humiliated.

So with this in mind, we had about 300-400 points to "sacrifice" from default Rodent, which means, we had about 300-400 points to play with when tuning the parameters.

The current OpenTal 1.0 matches a LOT of Tal's historical sacrifices that modern engines simply laugh at.

I wanted to emulate TAL, not an engine that attacks and happens to also have the same playing style as Tal.

All credit to Pawel, this is his work (with a sprinkling of my ideas), so thanks for making OpenTal a reality!
BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman »

jdart wrote:Nice, but not really Tal-like, because the sacrifices are sound!

--Jon
This is the only engine I've seen that plays Tal's ...Nf4 here.


Image

Speculative as hell! :)

I actually used this position (as well as others) while tuning, as a guide.
BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman »

PK wrote:But this is not an opponent for engines. This is an opponent for you!
Did you see that Fern? I believe this is the engine opponent you've waited years for.

:wink:
BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman »

Tal is back!

A 2560 engine humiliated in 19 moves!

[pgn][Event "Open Tal 1.0 Test"]
[Site "BRENDANNORMD8A2"]
[Date "2017.12.18"]
[Round "1"]
[White "OpenTal 1.0"]
[Black "Frenzee 2.00"]
[Result "1-0"]
[ECO "B43"]
[WhiteElo "2200"]
[BlackElo "2564"]
[PlyCount "37"]
[EventDate "2017.??.??"]
[TimeControl "180+2"]

1. e4 c5 2. Nf3 e6 3. Nc3 a6 4. d4 cxd4 5. Nxd4 Qc7 6. Be2 b5 7. Bf3 Bb4 {
(7. ... Bb4 8.Bd2 Nc6 9.Nxc6 dxc6 10.0-0 Bd6 11.h3 Nf6 12.Bg5 b4 13.Bxf6) -0.
74/11 5} 8. O-O {(8.0-0 Bb7 9.a4 Nf6 10.e5 Bxf3 11.Qxf3 Bxc3 12.bxc3 Nd5 13.
axb5 0-0 14.Rd1 f6 15.Ba3 axb5 16.Bxf8 Rxa1 17.Rxa1 Kxf8 18.exf6 Qxc3 19.fxg7+
Kxg7) +2.35/18 6} Bxc3 {(8. ... Bxc3 9.bxc3 Qxc3 10.Bf4 Ne7 11.Bd6 Nbc6 12.e5
Nd5 13.Ne2 Qb2 14.Bxd5 exd5) -0.18/12 5} 9. bxc3 {(9.bxc3) +2.63/18 6} Qxc3 {
(9. ... Qxc3 10.Rb1 Ne7 11.Bb2 Qc4 12.Nb3 0-0 13.Qd6 Nbc6 14.Rbc1 Bb7 15.Qxd7
Rab8 16.Rfd1) -0.14/125} 10. Be3 {(10.Be3 Bb7 11.Nb3 Nf6 12.Nc5 Nc6 13.Nxb7
0-0 14.a4 Ne5 15.Nd6 Ne8 16.Be2 Nxd6 17.Qxd6 Nc4 18.Bxc4 bxc4) +3.33/16 6} Ne7
{(10. ... Ne7 11.a4 bxa4 12.Rxa4 0-0 13.Bd2 Qc7 14.Nb5 Qc5 15.c4 Nbc6 16.Be3
Qe5) +0.05/11 5} 11. e5 {(11.e5 Nbc6 12.Nxc6 Nxc6 13.Qd6 Qxe5 14.Qxe5 Nxe5 15.
Bxa8 Nc4 16.a4 Nxe3 17.fxe3 b4 18.a5 d5 19.Rab1 Bd7 20.Rxb4 Bb5 21.Rd1) +2.90/
17 6} Nbc6 {(11. ... Nbc6 12.Nxc6 dxc6 13.Qd6 Bd7 14.Rfd1 Nd5 15.Rac1 Qc4 16.
a3 0-0-0 17.Bg5 Nc3 18.Rd4 Ne2+ 19.Bxe2) +0.02/12 5} 12. Nxc6 {(12.Nxc6 Nxc6
13.Qd6 Qxe5 14.Qxe5 Nxe5 15.Bxa8 0-0 16.a4 d6 17.Bf4 b4 18.Rad1 a5 19.Bxe5
dxe5 20.Rd6 f6 21.Rfd1 Bd7 22.Rxd7 Rxa8 23.h4 Rc8) +2.98/21 6} dxc6 {(12. ... d
xc6 13.Qd6 Bd7 14.Rfd1 Nd5 15.Bc5 Kd8 16.Rac1 Re8 17.Bh5 g6 18.Be2 h5 19.f4)
-0.07/12 5} 13. Qd6 {(13.Qd6 Bd7 14.Rad1 Rd8 15.Bg5 Nf5 16.Qc7 0-0 17.Bxd8 Be8
18.Be7 Nxe7 19.Qxe7 Qxe5 20.Rd8 Qf6 21.Qxf6 gxf6 22.Rfd1 f5 23.h4 f6 24.h5) +4.
83/22 6} Bd7 {(13. ... Bd7 14.Rfd1 Nd5 15.Bc5 Kd8 16.Be4 a5 17.a4 b4 18.Kh1 h5
19.Bd4 Qc4) -0.07/12 5} 14. Rad1 {(14.Rad1) +5.33/21 6} Rd8 {(14. ... Rd8 15.
Bg5 f6 16.exf6 gxf6 17.Rd3 Qb2 18.Rb3 Qe5 19.Bh5+ Ng6 20.Qxe5 fxe5 21.Bxd8
Kxd8 22.Rd3) -1.58/12 5} 15. Bg5 {(15.Bg5) +5.45/23 6} f6 {(15. ... f6 16.exf6
gxf6 17.Rd3 Qb2 18.Rb3 Qe5 19.Bh5+ Ng6 20.Qxe5 fxe5 21.Bxd8 Kxd8 22.Bf3 Nf4 23.
Rd1 Rg8) -1.69/13 6} 16. Bh5+ {(16.Bh5+) +5.93/16 5} Ng6 {(16. ... Ng6 17.exf6
gxf6 18.Rd3 Qb2 19.Rb3 Qe5 20.Qxe5 fxe5 21.Bxd8 Kxd8 22.Ra3 Bc8 23.Rd1+ Ke7 24.
Rad3) -1.92/12 6} 17. f4 {(17.f4) +6.11/17 5} fxg5 {(17. ... fxg5 18.f5 exf5
19.Rxf5 Qe1+ 20.Rxe1 Bxf5 21.Bxg6+ hxg6 22.Qxc6+ Kf8 23.e6 Ke7 24.Qb7+ Ke8) -3.
73/11 8} 18. f5 {(18.f5 exf5 19.Rxf5 Qc5+ 20.Qxc5 Bxf5 21.Rf1 Rf8 22.e6 Rf6 23.
Rxf5 Rxf5 24.Qxf5 Ke7 25.Bf3 Kd6 26.Qxg5 Kc7 27.Qc5 Rd1+ 28.Bxd1 a5) +6.16/18
5} exf5 {(18. ... exf5 19.Rxf5 Qe1+ 20.Rxe1 Bxf5 21.Qxc6+ Ke7 22.Qc7+ Rd7 23.
Qc5+ Ke6 24.g4 Bxc2 25.Qc6+ Ke7 26.Qxc2) -5.65/11 6} 19. Rxf5 {(19. ... Qc5+
20.Qxc5 Bxf5 21.Qxc6+ Ke7 22.Qc7+ Ke6 23.Rxd8 Rxd8 24.Qxd8 Kxe5 25.Qc7+ Kf6 26.
Qb6+ Be6 27.Bxg6 hxg6 28.Qxa6 b4 29.a4 bxa3 30.Qxa3) -7.24/13 9 Arena
Adjudication (19) (19.Rxf5 Qc5+ 20.Qxc5 Bxf5 21.Rf1 Rf8 22.e6 Rf6 23.Rxf5 Rxf5
24.Qxf5 Ke7 25.Qxg5+ Ke8 26.Qf4 Ke7 27.Qc7+ Kxe6 28.Bg4+ Kd5 29.Qxd8+ Kc4 30.
Qd3+ Kc5 31.Qe3+ Kb4 32.Be6 Ka5 33.Qc3+ Kb6) +7.25/21 5} 1-0[/pgn]

Ouch.
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: OpenTal - almost new engine

Post by carldaman »

PK wrote:These sacrifices work against humans or weaker engines. Default Rodent has beaten poor OpenTal like 19 to 1. This is even more than their rating difference, because search code is basically the same. These games were also entertaining in a way: most of them featured a declined sacrifice or a counter-sacrifice.
It does work as advertised, emulating Tal in a very convincing and impressive way, but I can only install it in Arena and not Fritz/CB. When trying to install it in CB the engine name field remains blank and the engine cannot be saved/installed.
carldaman
Posts: 2283
Joined: Sat Jun 02, 2012 2:13 am

Re: OpenTal - almost new engine

Post by carldaman »

carldaman wrote:
PK wrote:These sacrifices work against humans or weaker engines. Default Rodent has beaten poor OpenTal like 19 to 1. This is even more than their rating difference, because search code is basically the same. These games were also entertaining in a way: most of them featured a declined sacrifice or a counter-sacrifice.
It does work as advertised, emulating Tal in a very convincing and impressive way, but I can only install it in Arena and not Fritz/CB. When trying to install it in CB the engine name field remains blank and the engine cannot be saved/installed.
OK, I've figured out a workaround for installing it in CB.
First, I installed an instance of Rodent normally. Then I copied the OpenTal exe and other files to the same folder as the Rodent I had installed.
Finally, in the Appdata/Roaming/ChessBase/Engines.UCI folder I renamed the filename to point to OpenTal instead of Rodent and also changed the engine name (inside the *.uci file) accordingly.

Now it's working flawlessly in CB. :D

[Edit] The uci file looks like this:

Code: Select all

[ENGINE]
Name=OpenTal 1.0
Author=Pawel Koziol (based on Sungorus 1.4 by Pablo Vazquez)
Filename=C:\Engines\Rodent_III_x64_POPCNT_0232\dist\opental_x64popcnt.exe
CL
BrendanJNorman
Posts: 2526
Joined: Mon Feb 08, 2016 12:43 am
Full name: Brendan J Norman

Re: OpenTal - almost new engine

Post by BrendanJNorman »

carldaman wrote:
PK wrote:These sacrifices work against humans or weaker engines. Default Rodent has beaten poor OpenTal like 19 to 1. This is even more than their rating difference, because search code is basically the same. These games were also entertaining in a way: most of them featured a declined sacrifice or a counter-sacrifice.
It does work as advertised, emulating Tal in a very convincing and impressive way, but I can only install it in Arena and not Fritz/CB. When trying to install it in CB the engine name field remains blank and the engine cannot be saved/installed.
Confirmed the CB/Fritz bug on this end too.

Hopefully, a fix will come soon.

In the meantime, here's a logo.

Image
User avatar
AdminX
Posts: 6340
Joined: Mon Mar 13, 2006 2:34 pm
Location: Acworth, GA

Re: OpenTal - almost new engine

Post by AdminX »

BrendanJNorman wrote:
carldaman wrote:
PK wrote:These sacrifices work against humans or weaker engines. Default Rodent has beaten poor OpenTal like 19 to 1. This is even more than their rating difference, because search code is basically the same. These games were also entertaining in a way: most of them featured a declined sacrifice or a counter-sacrifice.
It does work as advertised, emulating Tal in a very convincing and impressive way, but I can only install it in Arena and not Fritz/CB. When trying to install it in CB the engine name field remains blank and the engine cannot be saved/installed.
Confirmed the CB/Fritz bug on this end too.

Hopefully, a fix will come soon.

In the meantime, here's a logo.

Image
I had a similar problem but checking unblock solved it for me.

Image
"Good decisions come from experience, and experience comes from bad decisions."
__________________________________________________________________
Ted Summers
User avatar
Dariusz
Posts: 364
Joined: Sat Jun 13, 2015 10:08 am
Location: Poland
Full name: Dariusz Domagała

Re: OpenTal - almost new engine

Post by Dariusz »

Pawel, thank you very very much for this amazing engine :-)

Carl, indeed, after I did your steps, OpenTal is working under fritz/chessbase :-)
Thank you !
Regards, Darius
https://chessengeria.eu
noctiferus
Posts: 364
Joined: Sun Oct 04, 2009 1:27 pm
Location: Italy

Re: OpenTal - almost new engine

Post by noctiferus »

I have the same troubles when trying to install into Hiarcs chess Explorer:
I get the message: is not a valid UCI engine.

Does it happen to anybody else?