Patricia 2.0 - likely the most aggressive chess engine ever made

[BrendanJNorman]

I am also noticing that Patricia 2.0 is laboring to win against engines much lower rated (2400-2650 CCRL) in games played without books. It plays adventurous but dodgy gambits such as the Wing Gambit with discrepancies in its eval. For example, at a time control of 10m +4s, it will play 1. e4 c5 2. b4 with an eval of +1.26, but after 2...cxb4 3. Nf3 it drops to -0.63. That's a huge and abnormal swing this early in the game.

I notced this too, but to fix the issue, would prefer it learn to BELIEVE in its ideas, rather than stop having them.

If you get me.

Nothing wrong with b4 vs Sicilian, as long as your opponent isnt Garry Kasparov (or an engine).

[Damas Clásicas]

https://github.com/Adam-Kulju/Patricia/ ... /tag/2.0.1
I hammered out a bunch of small bugs, including one which I think was behind the crashes (not sure, because I could never get Patricia to crash on my machine.) It passes valgrind and address sanitizer to high depths; if any problems persist, please make sure to notify me.

Thanks, you fixed it!

I tested Patricia 2.0.1 overnight (and also with those positions I sent yesterday here) and I got not more crashes!
Everything seems to be normal now.

[Whiskers]

I am also noticing that Patricia 2.0 is laboring to win against engines much lower rated (2400-2650 CCRL) in games played without books. It plays adventurous but dodgy gambits such as the Wing Gambit with discrepancies in its eval. For example, at a time control of 10m +4s, it will play 1. e4 c5 2. b4 with an eval of +1.26, but after 2...cxb4 3. Nf3 it drops to -0.63. That's a huge and abnormal swing this early in the game.

I notced this too, but to fix the issue, would prefer it learn to BELIEVE in its ideas, rather than stop having them.

If you get me.

Nothing wrong with b4 vs Sicilian, as long as your opponent isnt Garry Kasparov (or an engine).

I would too! The problem is that many sacrifices just aren’t good period, and gaslighting Patricia to believe in them will probably make the engine play much worse in regular chess too.

[towforce]

...The problem is that many sacrifices just aren’t good period, and gaslighting Patricia to believe in them will probably make the engine play much worse in regular chess too.

There may be a trade off between strength and fun/excitement. If so, I hope fun/excitement will be chosen!

[Whiskers]

...The problem is that many sacrifices just aren’t good period, and gaslighting Patricia to believe in them will probably make the engine play much worse in regular chess too.

There may be a trade off between strength and fun/excitement. If so, I hope fun/excitement will be chosen!

That’s the whole point of Patricia

[carldaman]

I am also noticing that Patricia 2.0 is laboring to win against engines much lower rated (2400-2650 CCRL) in games played without books. It plays adventurous but dodgy gambits such as the Wing Gambit with discrepancies in its eval. For example, at a time control of 10m +4s, it will play 1. e4 c5 2. b4 with an eval of +1.26, but after 2...cxb4 3. Nf3 it drops to -0.63. That's a huge and abnormal swing this early in the game.

I notced this too, but to fix the issue, would prefer it learn to BELIEVE in its ideas, rather than stop having them.

If you get me.

Nothing wrong with b4 vs Sicilian, as long as your opponent isnt Garry Kasparov (or an engine).

I tend to agree, too.
It's definitely 'trending' in the right direction!

I was just puzzled by the early and sudden drop in the eval after really "liking" the gambit at first, but the explanation makes a lot of sense.

[BrendanJNorman]

...The problem is that many sacrifices just aren’t good period, and gaslighting Patricia to believe in them will probably make the engine play much worse in regular chess too.

There may be a trade off between strength and fun/excitement. If so, I hope fun/excitement will be chosen!

That’s the whole point of Patricia

Those of us who have been into this hobby for decades witnessed so many engines that play beautiful chess (Thinker, Fizbo, ProDeo, until Ed pulled it back, Rodent, until Pawel reversed Texel tuning IIRC, Hakkapelitta, etc etc etc) eventually sacrifice style for Elo chasing.

Whether deliberately or not.

I think I speak for Carl,Thorsten, myself and a whole bunch of us, when I say "Keep doing what you're doing! Polish style/features (levels, multiPV etc) - Elo be damned!

I look forward to witnessing your great work.

[BrendanJNorman]

I am also noticing that Patricia 2.0 is laboring to win against engines much lower rated (2400-2650 CCRL) in games played without books. It plays adventurous but dodgy gambits such as the Wing Gambit with discrepancies in its eval. For example, at a time control of 10m +4s, it will play 1. e4 c5 2. b4 with an eval of +1.26, but after 2...cxb4 3. Nf3 it drops to -0.63. That's a huge and abnormal swing this early in the game.

I notced this too, but to fix the issue, would prefer it learn to BELIEVE in its ideas, rather than stop having them.

If you get me.

Nothing wrong with b4 vs Sicilian, as long as your opponent isnt Garry Kasparov (or an engine).

I tend to agree, too.
It's definitely 'trending' in the right direction!

I was just puzzled by the early and sudden drop in the eval after really "liking" the gambit at first, but the explanation makes a lot of sense.

Yeah, me too.

Was analyzing some Caro Kann and its like "White's crushing already!"

I was looking at the position and played a move "But what If I do this?"

Suddenly Patricia is like "Maybe you have a point.... > -0.54"

Love her anyway though. Crazy girl.

[BrendanJNorman]

Patricia was in a particularly grumpy mood lately.

No respect for former legends.

[pgn][Event "Gd"]
[Site "DESKTOP-Q1QR5FU"]
[Date "2024.04.05"]
[Round "3"]
[White "Patricia 2"]
[Black "Komodo 1"]
[Result "1-0"]
[WhiteElo "2200"]
[BlackElo "2200"]
[EventDate "????.??.??"]
[ECO "C06"]
[PlyCount "56"]
[BlackType "program"]
[Opening "French"]
[Termination "adjudication"]
[Time "16:37:59"]
[TimeControl "120+1"]
[Variation "Tarrasch, Closed, Leningrad, 3.Nd2 Nf6 4.e5 Nfd7 5.Bd3 c5 6.c3 Nc6 7.Ne2 cxd4 8.cxd4 Nb6"]
[WhiteType "program"]

1.e4 {+0.87/17 4} 1...e6 {-0.34/15 4} 2.d4 {+1.06/18 6} 2...d5 {-0.32/15 5} 3.Nd2 {+0.84/18 4} 3...Nf6 {-0.28/15 5} 4.e5 {+1.21/19 3} 4...Nfd7 {-0.19/17 3} 5.c3 {+1.26/19 4} 5...c5 {-0.28/17 3} 6.Bd3 {+1.09/17 3} 6...Nc6 {-0.30/16 4} 7.Ne2 {+1.89/18 6} 7...cxd4 {-0.20/16 2} 8.cxd4 {+1.45/19 3} 8...Nb6 {-0.43/17 11} 9.a3 {+2.25/17 3} 9...Be7 {-0.50/16 2} 10.h4 {+3.54/17 5} 10...f6 {0.00/16 8} 11.Nf4 {+2.02/16 3} 11...Nxd4 {+0.72/15 2} 12.Qh5+ {+2.04/17 3} 12...Kd7 {+0.62/15 3} 13.exf6 {+1.86/17 3} 13...Bxf6 {+0.67/15 3} 14.Nf3 {+2.76/16 3} 14...Nb3 {+0.72/14 2} 15.O-O {+4.18/15 4} 15...Nxa1 {+1.11/10 0} 16.Nxe6 {+7.75/16 2} 16...Kxe6 {+1.39/11 0} 17.Re1+ {+7.37/17 3} 17...Kd7 {0.00/14 3} 18.Bf4 {+8.77/19 5} 18...Qe7 {+0.09/14 9} 19.Rxe7+ {+7.51/18 4} 19...Kxe7 {-0.19/12 0} 20.Ng5 {+6.86/18 6} 20...Be6 {-0.50/15 5} 21.Qe2 {+8.08/17 2} 21...Bxg5 {-0.34/12 0} 22.Bxg5+ {+7.70/18 6} 22...Kf7 {-0.63/15 2} 23.h5 {+7.30/16 2} 23...Rae8 {-0.95/15 4} 24.Qf3+ {+11.66/16 3} 24...Kg8 {-1.48/13 0} 25.h6 {+11.42/17 2} 25...Nd7 {-1.93/17 2} 26.Qg3 {+10.84/17 2} 26...Bf7 {-4.24/13 1} 27.Bb5 {+11.97/16 2} 27...Re1+ {-4.93/14 2} 28.Kh2 {+12.68/14 1} 28...Nf8 {-7.53/14 2 Arena Adjudication} 1-0

[/pgn]

[pgn][Event "Gd"]
[Site "DESKTOP-Q1QR5FU"]
[Date "2024.04.05"]
[Round "12"]
[White "Rybka v1.2n.x64"]
[Black "Patricia 2"]
[Result "0-1"]
[WhiteElo "2200"]
[BlackElo "2200"]
[EventDate "????.??.??"]
[ECO "B21"]
[PlyCount "69"]
[BlackType "program"]
[Opening "Sicilian"]
[Termination "adjudication"]
[Time "14:59:32"]
[TimeControl "120+1"]
[Variation "Smith-Morra, Andreaschek Gambit"]
[WhiteType "program"]

1.d4 c5 {+0.36/18 5} 2.e4 cxd4 {-0.13/17 6} 3.Nf3 e5 {+0.31/18 4} 4.c3 {+0.03/13 3} 4...Nf6 {+0.46/17 5} 5.Nxe5 {+0.09/13 3} 5...Nc6 {-0.28/18 6} 6.cxd4 {+0.10/13 3} 6...Nxe4 {-0.54/16 4} 7.Qe2 {+0.89/14 6} 7...d5 {-1.60/18 11} 8.Nxc6 {+0.95/14 3} 8...bxc6 {-3.15/16 4} 9.f3 {+1.13/14 3} 9...Bg4 {+0.28/16 5} 10.Qe3 {+1.26/13 4} 10...f5 {-1.51/16 15} 11.fxg4 {+1.13/14 13} 11...f4 {+0.48/17 3} 12.Qxf4 {+1.35/15 3} 12...Bd6 {+0.94/18 2} 13.Qf3 {+0.84/15 7} 13...Rf8 {+1.53/17 2} 14.Qe2 {+0.83/14 3} 14...Kd7 {-0.01/16 2} 15.Be3 {+0.62/14 3} 15...Qh4+ {+0.01/18 2} 16.Kd1 {+0.78/14 1} 16...Nf2+ {+1.63/17 3} 17.Bxf2 {+0.87/14 6} 17...Rxf2 {+0.36/19 5} 18.Qd3 {+0.76/14 4} 18...Rb8 {+0.32/18 2} 19.Be2 {+0.83/14 5} 19...Rxb2 {+1.90/15 2} 20.Nd2 {+0.38/15 6} 20...Bb4 {+1.88/18 2} 21.Nf3 {+0.38/14 3} 21...Qf6 {+2.12/17 3} 22.a3 {+0.57/13 2} 22...Ba5 {+2.27/16 2} 23.Bf1 {+0.38/13 2} 23...Kd8 {+2.89/17 2} 24.Qe3 {+0.03/14 4} 24...c5 {+4.17/15 2} 25.h4 {-1.22/13 11} 25...Qc6 {+5.34/16 2} 26.a4 {+0.03/12 2} 26...c4 {+5.97/16 1} 27.Qg5+ {+0.03/13 2} 27...Kc8 {+6.32/14 1} 28.Qf5+ {+0.12/12 2} 28...Kb7 {+7.19/16 2} 29.Rg1 {-3.38/13 7} 29...Bc3 {+8.22/13 2} 30.h5 {-1.82/12 4} 30...Ka6 {+8.55/15 2} 31.Qxh7 {-2.03/13 2} 31...Ka5 {+7.21/16 2} 32.Qg6 {-3.81/13 3} 32...Qb7 {+8.46/18 1} 33.Bxc4 {-4.20/13 2} 33...Rbd2+ {+8.82/19 1} 34.Nxd2 {-6.96/16 2} 34...Rxd2+ {+8.13/20 2} 35.Ke1 {-7.44/16 2 Arena Adjudication} 0-1

[/pgn]

This is how chess is to be played!

[carldaman]

+1 LMAO

So many devastating moves in those 2 games!
The seemingly self-forking 9...Bg4 in the second game, followed immediately by f7-f5-f4, was an absolutely exhilarating sequence!

In one of my own tests, the Rook on a1 was attacked by a Kn on b3, and Patricia responded with Rfc1, offering a choice of Rooks! LOL