Gaviota, Wasp, The Baron, Pro Deo

[Rebel]

Ed Schroder: dear Ed, a new Pro Deo perhaps?

Actually I just released one, but you need to go to the CTF sub forum. Not suitable to post it here.

Ed, you are literally the Einstein of chess programming.

So many crazy (but creatively brilliant) ideas coming in each release.

Yeah, that would be me, crazy, bored with the traditional ways

Here's the link for anybody interested in what I'm talking about (I know you're curious)

http://rebel13.nl/prodeo/prodeo25.html

By the way, keeping the theme of this thread, how would you suggest making a ProDeo personality as positional as possible?

Well actually the new playing styles are exactly about that, I love the gambit style but thee is still work to do. It's a starting point.

[BrendanJNorman]

Ed Schroder: dear Ed, a new Pro Deo perhaps?

Actually I just released one, but you need to go to the CTF sub forum. Not suitable to post it here.

Ed, you are literally the Einstein of chess programming.

So many crazy (but creatively brilliant) ideas coming in each release.

Yeah, that would be me, crazy, bored with the traditional ways

Here's the link for anybody interested in what I'm talking about (I know you're curious)

http://rebel13.nl/prodeo/prodeo25.html

By the way, keeping the theme of this thread, how would you suggest making a ProDeo personality as positional as possible?

Well actually the new playing styles are exactly about that, I love the gambit style but thee is still work to do. It's a starting point.

Here's a quick blitz game (I'm in my lunch break) I played with Prodeo 2.5 just now.

[pgn][Event "Computer chess game"]
[Site "ADMINRG-GT4VPMO"]
[Date "2017.09.16"]
[Round "?"]
[White "Prodeo 2.5"]
[Black "Brendan"]
[Result "1-0"]
[BlackElo "2400"]
[Time "15:54:10"]
[TimeControl "60+1"]
[Termination "time forfeit"]
[PlyCount "79"]
[WhiteType "program"]
[BlackType "human"]

1. c4 {+0.01/1 0} e5 2. Nc3 {+0.01/1} Nf6 3. Nf3 {+0.01/1 0} Nc6 4. g3
{+0.01/1 0} Bb4 5. Bg2 {+0.01/1 0} O-O 6. O-O {+0.01/1 0} e4 7. Ng5
{+0.01/1 0} Bxc3 8. bxc3 {+0.01/1} Re8 9. f3 {+0.01/1 0} e3 10. dxe3
{+0.01/1 0} h6 11. Nh3 {+0.01/1 0} d6 12. Nf4 {(12.Nf4 Nd7 13.e4 Nce5
14.Qd4 Nc6 15.Qd1 Nce5 16.Qd4 Nc6) 0.00/13 2} Ne5 13. Qb3 {(13.Qb3 a5 14.a4
Nfd7 15.Nd3 Nb6 16.Nxe5 Rxe5 17.f4 Re8 18.f5 Qf6 19.e4 Bd7) -0.02/13 3} b6
14. Nd5 {(14.Nd5 Ned7 15.Qa4 Bb7 16.Nxf6+ Nxf6 17.Bh3 Qe7 18.e4 Qe5 19.Bf4
Qxc3 20.Rad1) +0.29/12 2} Ba6 15. Nxf6+ {(15.Nxf6+ Qxf6 16.f4 Nxc4 17.Bxa8
Rxa8 18.Qa4 Na5 19.Qd4 Qe7 20.Rb1 Re8 21.f5 Bxe2 22.Rf2 c5 23.Qf4) +0.67/14
2} Qxf6 16. f4 {(16.f4 Nxc4 17.Bxa8 Rxa8 18.Rd1 Qe7 19.Qa4 Nxe3 20.Qxa6
Nxd1 21.Qd3 Re8 22.Qxd1 Qxe2 23.Qa4 Kh8 24.Ba3) +0.62/15 2} Nxc4 17. Bxa8
{(17.Bxa8 Rxa8 18.Qa4 Qd8 19.Qxa6 Nxe3 20.Bxe3) +0.63/12 0} Rxa8 18. Qa4
{(18.Qa4 Na5 19.Qd4 Qg6 20.f5 Qh5 21.Rf2 Bxe2 22.f6 Nc6 23.Qe4 d5 24.Qc2)
+0.69/13 1} Na5 19. Qd4 {(19.Qd4 Qe6 20.f5 Qc4 21.Rf2 Re8 22.f6 Re4 23.Qxc4
Nxc4 24.Rb1 d5 25.fxg7 Kxg7 26.Rf4 Re8) +0.81/14 2} Qe7 20. f5 {(20.f5 c5
21.Qg4 Qf6 22.e4 Re8 23.Rb1 Bb7 24.Bb2 Rxe4 25.Rf4 Rxf4 26.Qxf4 Qe7 27.e3
Bd5) +0.71/14 2} Bb7 21. Bd2 {(21.Bd2 Re8 22.Qa4 Qe5 23.Rad1 Be4 24.c4 Nc6
25.Rf2 a5 26.Rf4) +0.80/11 1} c5 22. Qf4 {(22.Qf4 Qe4 23.Qxe4 Bxe4 24.Rf4
d5 25.Rd1 Re8 26.f6 Nc4 27.Bc1 b5 28.fxg7 Kxg7 29.Kf2 a5 30.Bd2 a4 31.Bc1
Re6 32.Bd2) +0.53/15 2} f6 23. c4 {(23.c4 Nc6 24.Rfd1 Re8 25.Rac1 Qd7
26.Bc3 Ne5 27.Bb2 Qc6 28.e4 Qxe4 29.Qxe4 Bxe4 30.Bxe5 Rxe5 31.Rxd6 Bxf5)
+0.90/13 1} Nc6 24. Rfd1 {(24.Rfd1 Re8 25.Rac1 Ne5 26.Bc3 Qd7 27.Bb2 Qc6
28.e4 Qa4 29.Bxe5 dxe5 30.Qe3 Qxa2 31.Rd7 Bc6 32.Rc7) +0.92/14 2} Ne5 25.
Bc3 {(25.Bc3 Nf7 26.Bb2 Re8 27.Rd3 Qd7 28.Ra3 Ng5 29.Qf1 Qc6 30.Kf2 Ne4+
31.Kg1) +0.77/13 2} Nf7 26. Kf2 {(26.Kf2 Ng5 27.Kf1 Re8 28.Be1 Ne4 29.Qg4
Ng5 30.Bf2 Ne4 31.Qf4 Qf7 32.Rac1 Qh5 33.Be1 Qxh2) +0.56/13 2} Re8 27. Ke1
{(27.Ke1 Qxe3 28.Qxe3 Rxe3 29.Kd2 Ng5 30.Kxe3 Ne4) +0.48/13 2} Qxe3 28.
Qxe3 {(28.Qxe3 Rxe3 29.Kd2 Re4 30.Kd3 Re8 31.a4 d5 32.cxd5 Bxd5 33.Kd2 Ng5
34.Ra3 Bc4 35.e3 Nf3+ 36.Kc2 Nxh2 37.Bb2) -0.31/15 2} Rxe3 29. Kd2 {(29.Kd2
Re7 30.h4 Ba6 31.Re1 Bxc4 32.e4 Ne5 33.Bxe5 Rxe5 34.Kc3 Bf7 35.a4 d5
36.exd5 Bxd5) -0.22/14 2} Re4 30. Kd3 {(30.Kd3 Rg4 31.Rac1 Be4+ 32.Kd2 Bxf5
33.Bb2 Be6 34.Kd3 d5 35.cxd5 Bxd5 36.a3 Nd6 37.Rf1) 0.00/14 2} Kf8 31. Rf1
{(31.Rf1 d5 32.cxd5 Bxd5 33.Rf2 Ra4 34.a3 Nd6 35.Kc2 Ne4 36.Rff1 c4 37.Bb4+
Kf7) +0.15/13 3} Ke7 32. Rf4 {(32.Rf4 Ng5 33.h4 Rxf4 34.gxf4 Nh3 35.Ke3 h5
36.Rf1 Bc8 37.Kd3 Bxf5+ 38.e4 Be6 39.a3 g6) +0.66/15 1} Rxf4 33. gxf4
{(33.gxf4 Ba6 34.e4 b5 35.cxb5 Bxb5+ 36.Ke3 Bc6 37.Rd1 h5 38.h4 Bd7)
+1.23/10 0} Ba6 34. e4 {(34.e4 d5 35.exd5 Nd6 36.Re1+ Kd7 37.Ke3 Bxc4
38.Rg1 Nxf5+ 39.Ke4 Nd6+ 40.Ke3 Ne8 41.Ke4 Bxa2) -0.42/13 1} d5 35. exd5
{(35.exd5 Nd6 36.Be5 Bxc4+ 37.Kc3 Bxd5 38.Bxd6+ Kxd6 39.a3 Ke7 40.Rg1 Kf7
41.Rd1 Bc6 42.Kc4 Ke7 43.h4) -0.30/15 1} Nd6 36. Be5 {(36.Be5 Bxc4+ 37.Kc3
Bxd5 38.Bxd6+ Kxd6 39.a3 Ke7 40.Rg1 Kf7 41.Rd1 Bc6 42.Kc4 h5 43.Rd6 Ba4
44.Rd2 Ke7 45.Re2+ Kd6 46.Re6+ Kd7) -0.16/15 1} fxe5 37. fxe5 {(37.fxe5
Bxc4+ 38.Kc3 b5 39.exd6+ Kxd6 40.a3 Bxd5 41.Rg1 Ke5 42.Rxg7 a5 43.Rc7 b4+
44.Kb2 c4 45.Rc5) +0.21/13 0} Nxf5 38. Rc1 {(38.Rc1 Nd4 39.Ke3 Bc8 40.Rg1
g5 41.Rf1 Bf5 42.Rf2 Nc2+ 43.Kd2 Nd4 44.Kc3 h5 45.a3 a6 46.Rf1) +0.61/14 1}
g6 39. Rc3 {(39.Rc3 Bb7 40.Ra3 a6 41.Rb3 b5 42.Ke4 Nd4 43.Rg3 Bc8 44.d6+
Kf7 45.cxb5 axb5 46.Kd5 Ne6 47.Rf3+ Kg7) +0.75/14 1} h5 40. Ra3 {(40.Ra3
Bc8 41.Rxa7+ Kd8 42.e6 Nd6 43.Rg7 Nf5 44.Rxg6 Ke7 45.a3 Ba6 46.Rg5 Nd6
47.Rg7+ Kf6) +2.45/14 1} ... {1-0 Black forfeits on time} 1-0 [/pgn]

Mind you, I had a couple of beers just now with a business lunch, so my play is a bit sloppy, BUT....WHAT THE HELL is 36.Be5!?!?!?

Gonna test Prodeo 2.5 and Gaviota when I get home today.

BTW...that was Prodeo on its default setting!

[Rebel]

P.S Which version of Gaviota should I test?

I use the Gaviota 1.0 version. It's a fine positionnal player, and if you use it as a WB engine, there are a lot of options and features: books, learning, book learning... It's very balanced, not too fast, and I like its plans. It's a tough positionnal player, a bit weaker than Wasp, probably, although better in simple positions (Wasp is faster). It's often able to find original and active plans that make sense while being refreshing.

I love Gaviota for its positional play. It can play very unusual moves that raises an eyebrow but then a few moves later everything becomes clear, meaning there was a plan after all. Quite unique.

Ed tried to use Toga's search with different evaluations, but, unfortunately, he's semi-retired, because he could have discovered new ways of doing things.

The idea with TOGA was to create an universal platform for a new type of competition. The evaluation is now on par with Toga and I am waiting for the first entry.

Also I made some changes in Toga's evaluation, seems to me there is still room for improvement.

[Rebel]

Mind you, I had a couple of beers just now with a business lunch, so my play is a bit sloppy, BUT....WHAT THE HELL is 36.Be5!?!?!?

[d]8/p3k1p1/bp1n1p1p/2pP1P2/2P2P2/2BK4/P6P/R7 w - -

Make you forfeit on time

[Lyudmil Tsvetkov]

I have a few questions to some of our honourable members:

Miguel Ballicora: dear Miguel, do you still work on Gaviota? It's a fine engine I use a lot to analyze and I would be very disappointed if its development was finished. Will there be a new version anytime soon?

Frank Quisinsky: Do you know if John Stanback will release the version of Wasp you tested recently?

Ed Schröder: dear Ed, a new Pro Deo perhaps?

I am also interested in a more recent version of The Baron. Basically, I like to analyze and sometimes play against positionnal engines, and those some of my favourite (although I have recently discovered Zarkov 6.55, which also has a very... understandable style of play). I also think that the paradigm in computer chess these engine represent is not a dead end, quite the opposite, and that these engines have awful lot to offer.

I would be very grateful for answers, and some good news.

Miguel has moved to the political field.

He will soon be releasing a political engine.

[matejst]

In your experience which (Giovta or Rodent III, Rubinstein personality) plays better positional chess?

I just started to test Rodent III. I hope I will soon have time to test the complete package: personnality, opening repertoire, on Rubinstein's own games, and to tweak the engine a bit, since I don't always agree with Pawel assessment of a player style.

Of the others, I think Wasp clearly plays the best positionnal chess in the opening and middlegame. I compared it with Komodo 5.1 (an engine I also like very much), and in simple positions where there's initiative for material -- where knowledge can be crucial -- they mostly agree. Wasp needs more time, but finds the best answers at lower depths.

What I like about Wasp is it's ability to find good strategic "plans". While engine do not make plans, it looks like Wasp does. Gaviota, there, is also good, SmarThink too, although SmarThink sometimes suddenly gets off the rail.

I would say that, from the engine I seriously tested and played with, Wasp is no 1, Komodo 5 no 2, Gaviota 3. Among older engines, Zarkov 6.55 and Gandalf, although Gandalf is very aggressive, kind a Fischer playing black. There was also a version of Diep I liked very much, but it's really too slow, and there were problems with its use.

The problem with Wasp, Gaviota, Zarkov, Gandalf, old Pro Deos, is that they need to get to a solid depth (14-20 plies, depends of the engine), and it takes time, so their ratings are not a measure of the quality of their game. A heavy static evalution doesn't give ELO.

There is also the question of features, and I am grateful for the effort Ed and Pawel make to offer us complete chess packages.

[matejst]

Also I made some changes in Toga's evaluation, seems to me there is still room for improvement.

I agree there. I honestly think that finding a good balance now, "the middle way" to quote Mark Uniacke, is the way to the future. Computers will be faster, and speed tricks less relevant.

When I read programmers posts, it's too often question of "reductions", "prunning", "depth". Perhaps coding "majority", "coordination", certain type of positions to avoid or to go for is more difficult. Or less rewarding in the short term.

[Dann Corbit]

Gaviota is one of my favorite engines too.
I enjoy playing against it, even though it kicks my butt.

[Dann Corbit]

I am very fond both of Gaviota and of Miguel.
Great engine, and a great person.

[BrendanJNorman]

Dear Brendan,

I already use Rodent III with the Rubinstein personnality. I hoped for a Karpov personnality, but one needs time to understand the engine one already uses. I'll stick with Rubinstein for now, then I'll try Fischer with white and Petrosian with black.

About Pro Deo: its search is faster now, but its game makes less sense. And I have already seen Ed's tweaking tool. Recently, I also found a huge collection of excellent personnalities for Pro Deo on another site.

About Wasp: I didn't use the first one. I find the latest version very sensible, and I am mostly interested in features: learning ability, multi PV. John promised all of this, and the 20 ELO improvements is totally irrelevant to me.

I often analyze openings with engines -- I do my own calculations and search my own plan, than compare with the game (or book) and the engine output. And here, in position where there are no clear paths, one can really discover the engine... let's say... "personnality". I was delighted with Gandalf 7, and the last Zarkov, at 12 plies, is somewhere on my screen, although it usually find much better moves with more time, at 14 or 15 plies. But I also need time to make my mind, so it's okay.

I am against TBs, since I have the impression that authors code less and less knowledge in engines -- seems to be the easier way. I expected a lot of Norman (the other one), but I play chess under wine, in a 32-bit system, and can't get Fire to work. I had it under windows, on another machine, but it just looks too "machine-like".

Saw that Miguel had a 1.01 version of Gaviota at TCEC, and hoped for a new version, perhaps with more knowledge in endings (although Miguel B. developed its own TBs), or a way to recognize better closed pawn structures.

Then, I think it's the right paradigm. When I see how Aronian beats Carlsen, I have no doubts that the change of time control, the transformation of chess in a full fledge sport for young players deteriorated his quality. It's kinds "the Kasparov approach", with engines instead of a team of seconds. You win in the opening or you wait for a blunder. Players like Karpov, Salov, or even Fischer, with a great positionnal technique are out of fashion.

And now, when the development of Stockfish has stalled, I guess it's time not for simplification, but for a more developped evaluation, use of selected data (Ed think that there is more in big data, but there must be a way to extract knowledge from big data, the way Larry tries).

I am very disappointed that chess engines aren't opponents any more, hadn't become teachers, but are only used for hyperbullet matches and tournaments.

About Rodent: I have a Karpov personality for Rodent Version II 0.9.33 which plays really nice chess with a dedicated opening book following Karpov's 1992-1997 opening repertoire. Can be impressive to watch.

About Prodeo: Version 2.5 is amazing. I've only tested "defensive" style so far, but it didn't seem very defensive. Reminds me of Thinker "passive" haha.

About Gaviota: I tested it yesterday and really like this engine's style. I agree, the positional manner of handling the games is really impressive.

I really agree that it's a shame that only a few programmers are trying to create OPPONENTS for humans now (which was the original intention of computer chess) and are more focussed on never-ending Elo gain.

I think the ChessMaster series was on the right track, and with UCI compatibility+ability to run tournaments with external engines - would have been unmatched.

As far as positional engines go:

I'm currently running a tournament featuring Gaviota 1.0, Rodent II Karpov, Rodent III Strangler, Prodeo 2.5 Defensive, Naum 2.0, RomiChess p3k, Gaia 2.2a, ChessTiger 2007, Komodo 5.1, Wasp, Amyan, Baron, Ares 1.005, Shredder, Frenzee, Zarkov, Hiarcs 12.1, Rhetoric 1.2 and some other engines I consider "positional/strategic"

Some interesting samples coming along.

By the way, I don't really have a preference for attacking engines, but HUMAN-like engines - the problem is, I havent found many really human-like positional ones.

Gaviota is an impressive find, I must admit.