Patricia 2.0 - likely the most aggressive chess engine ever made

[Alexander Schmidt]

In other words, this is how it plays. Who cares if another engine was able to refute something?

Anyway, it shouldn't just give away pieces for nothing. You don't need an engine to refute this blunders.

[BrendanJNorman]

In other words, this is how it plays. Who cares if another engine was able to refute something?

Anyway, it shouldn't just give away pieces for nothing. You don't need an engine to refute this blunders.

Yeah man, I'm sure you'd crush Patricia effortlessly.

[Whiskers]

Is this a search bug or just part of the way it plays? Patricia was White. Black was a 2000 Elo bot.
Game in 15.
Black wins.

[pgn]1. e4 c5 2. b4 cxb4 3. a3 Qa5 4. Nf3 e5 5. d4 Nf6 6. Nxe5 Nxe4 7. Bd3 bxa3+ 8. c3 Nxc3 9. Qd2 Bb4 10. O-O Nxb1 11. Qg5 Rf8 12. Qxg7 a2 13. Bg5 h6 14. Bf6 d5 15. Re1?? Bxe1 16. Rxa2?? Qxa2 17. Nxf7 Rf7
0-1[/pgn]

0-1

Yeah, I can reproduce this one locally. It's not a bug, it comes about because Patricia just wasn't able to see she gets mated in time and thought showing off with the "sacrifices" was a great idea, but stuff like that is something I want to stamp out for v3.

[Peter Berger]

Is this a search bug or just part of the way it plays? Patricia was White. Black was a 2000 Elo bot.
Game in 15.
Black wins.

My answer is: neither.
It is just a feature of the way strong engines are tested currently. They are supposed to play against a pool of engines of similar strength for loads of games all the time according to current religious belief of programmers.
As the testing is usually done well, the engines have become mighty good at this job.
Per engine design Patricia's rating would be _way_ lower, if it had to fight against weaker engines more often.
In kind of a similar way the same would apply to Stockfish, which would not lose a game, but draw way too many.
From time to time people bring this topic up - but as engine developpers don't care too much as they are driven by other objectives, we get more and more extreme examples for this phenomenon all the time.
And before someone asks for more games : how often is an 3200 engine expected to lose (or draw) against a 2000 engine?
Peter

[Alexander Schmidt]

Yeah man, I'm sure you'd crush Patricia effortlessly.

That's not what I wrote, I wrote you don't need an engine to refute 15. Re1, do you?

[Werewolf]

In other words, this is how it plays. Who cares if another engine was able to refute something?

Anyway, it shouldn't just give away pieces for nothing. You don't need an engine to refute this blunders.

Yes that was my point. I’ve beaten the chess.com bot that beat Patricia - it’s only about 2000 Fide Elo.

[Werewolf]

Is this a search bug or just part of the way it plays? Patricia was White. Black was a 2000 Elo bot.
Game in 15.
Black wins.

[pgn]1. e4 c5 2. b4 cxb4 3. a3 Qa5 4. Nf3 e5 5. d4 Nf6 6. Nxe5 Nxe4 7. Bd3 bxa3+ 8. c3 Nxc3 9. Qd2 Bb4 10. O-O Nxb1 11. Qg5 Rf8 12. Qxg7 a2 13. Bg5 h6 14. Bf6 d5 15. Re1?? Bxe1 16. Rxa2?? Qxa2 17. Nxf7 Rf7
0-1[/pgn]

0-1

Yeah, I can reproduce this one locally. It's not a bug, it comes about because Patricia just wasn't able to see she gets mated in time and thought showing off with the "sacrifices" was a great idea, but stuff like that is something I want to stamp out for v3.

Your work is amazing, thanks for engaging on all this.

[AdminX]

I see what you mean by 'Aggressive'. Patricia wiped the floor with Benjamin in a short 20 game blitz match (18-2). Here was a rare win by Benjamin.

[pgn]
[Event "5 Minutes/Game"]
[Site "Engine Match"]
[Date "2024.04.12"]
[Round "3"]
[White "Benjamin 2.0"]
[Black "Patricia 2.0"]
[Result "1-0"]
[ECO "A00"]
[PlyCount "79"]

1.Nc3 {book 0s} 1...Nf6 {book 0s} 2.d4 {book 0s} 2...d5 {book 0s} 3.Nf3 $5 {book 0s} ( {Theory prefers} 3.Bf4 $1 ) 3...c6 {book 0s} 4.Bf4 {+0.40/14 10s} 4...c5 $2 {-0.42/17 11s (Bf5) Stockfish 16.1: +1.38/31} ( 4...Nbd7 {Stockfish 16.1: +0.02/32} ) 5.dxc5 {+0.84/13 8s} 5...e6 {+1.15/18 11s} 6.b4 {+0.98/14 0s} 6...Be7 $6 {+1.74/17 14s (Bd7) Stockfish 16.1: +1.69/31} ( 6...a5 {Stockfish 16.1: +1.36/31} 7.a3 Be7 8.e3 O-O 9.Be2 b6 10.cxb6 Qxb6 11.Na4 Qd8 12.b5 Bb7 13.Nd4 Qc8 14.O-O Nbd7 15.Bg3 Ne4 16.c4 dxc4 17.Qc2 Nxg3 18.hxg3 Rd8 19.Rfc1 Qc7 20.Qxc4 Qxc4 21.Rxc4 Rac8 22.Rxc8 Rxc8 23.Bf3 Bxf3 24.gxf3 Rc4 25.Nc6 Rxa4 26.Nxe7+ Kf8 ) 7.Nb5 {+1.27/15 8s} 7...O-O {+0.70/18 13s (Na6)} 8.Nc7 {+1.71/15 10s} 8...e5 $2 {-1.05/18 12s (Ne4) Stockfish 16.1: +2.60/25} ( 8...Ne4 {Stockfish 16.1: +1.53/29} 9.Nxa8 Bf6 10.Bd2 Nc6 11.e3 Nxd2 12.Kxd2 Nxb4 13.c3 Nc6 14.Rb1 Qa5 15.Qc2 Qxc5 16.Nc7 e5 17.Be2 Bd8 18.Nxd5 Qxd5+ 19.Ke1 Ba5 20.h4 h6 21.Kf1 Qc5 22.Nd2 Qe7 23.Rd1 f5 ) 9.Bxe5 {+2.71/15 4s} 9...Nc6 {+0.16/16 10s} 10.Nxa8 {+3.01/16 0s} 10...Ne4 $1 {+2.23/16 13s (Be6)} 11.Bc7 {+3.22/12 4s} 11...Bh4 $1 {+2.79/17 12s} 12.g3 {+2.79/17 0s} 12...Qf6 {+2.62/17 40s} 13.Bg2 {+2.60/17 0s} 13...Nc3 $1 {+2.52/18 5s (Qc3+)} 14.Qc1 {+3.19/13 9s} 14...Re8 {+2.53/15 7s} 15.e3 {+2.32/13 7s} 15...Bh3 {-1.97/17 7s} 16.Nxh4 {+2.10/15 0s} 16...Bxg2 {+1.48/16 6s} 17.Nxg2 {+2.27/15 8s} 17...Nd4 $6 {-0.01/17 5s (Qf3) Stockfish 16.1: +2.75/27} ( 17...Qf3 {Stockfish 16.1: +2.25/33} 18.Kf1 d4 19.Bf4 Qe2+ 20.Kg1 Qc4 21.h3 Ne2+ 22.Kh2 Nxc1 23.Nc7 Rd8 24.Rhxc1 Qxb4 25.Rab1 Qxc5 26.Nb5 dxe3 27.Bxe3 Qe7 28.c4 Ne5 29.Ne1 h5 30.c5 h4 31.g4 Qf6 32.Nd6 b6 33.Nc4 bxc5 34.Nxe5 Qxe5+ 35.Kg1 Qe4 36.Bxc5 Qg6 37.Be3 f5 ) 18.Nf4 {+4.80/14 8s} 18...Nf3+ $6 {+0.80/16 4s (Qf5) Stockfish 16.1: +3.20/17} ( 18...Qa6 {Stockfish 16.1: +2.65/31} 19.Nd3 Nf3+ 20.Kf1 g5 21.h4 g4 22.b5 Qa4 23.Rg1 d4 24.Rg2 h5 25.Qb2 Rxa8 26.exd4 Nxb5 27.Qb4 Qxb4 28.Nxb4 Nxc7 29.Rd1 Nb5 30.Nd5 Nbxd4 31.c3 Ne6 32.Nf6+ Kh8 ) 19.Kf1 {+5.44/7 0s} 19...d4 $6 {+1.26/16 6s (g5) Stockfish 16.1: +4.61/24} ( 19...Qc6 {Stockfish 16.1: +3.12/30} 20.h4 d4 21.Rh3 dxe3 22.fxe3 h5 23.Kf2 Nd2 24.Qh1 Qd7 25.Bd6 Rxa8 26.Qg2 Qg4 27.Be5 Nce4+ 28.Kg1 Nf3+ 29.Kh1 Rd8 30.Bd4 Ned2 31.Rd1 Ne4 32.Rf1 Nfd2 33.Ra1 Nf3 34.b5 Ned2 35.c6 bxc6 36.bxc6 Qc8 37.Rd1 Qxc6 38.Rxd2 ) 20.Bd6 {+5.57/13 5s} 20...Qf5 $1 {+0.78/14 4s (dxe3)} 21.g4 $6 {+4.61/11 5s Stockfish 16.1: +2.23/34} ( 21.h4 {Stockfish 16.1: +4.35/26} 21...Ne4 22.Nc7 Rc8 23.Ncd5 Re8 24.Kg2 Ned2 25.Qd1 Qe4 26.g4 h5 27.Kg3 hxg4 28.Qe2 Qxc2 29.Rac1 Ne4+ 30.Kg2 ) 21...Qe4 {-2.66/16 10s} 22.h4 {+2.08/12 0s} 22...Nd2+ {-5.70/14 3s} 23.Kg1 {+3.01/12 0s} 23...dxe3 $2 {-3.18/14 3s Stockfish 16.1: +3.26/24} ( 23...Nf3+ {Stockfish 16.1: +2.05/35} 24.Kg2 Nxh4+ 25.Kh2 g5 26.Nc7 Re5 27.Bxe5 Nf3+ 28.Kh3 Nxe5 29.Rg1 gxf4 30.exf4 Ne2 31.fxe5 Nxc1 32.Raxc1 Qc6 33.Kg3 Qxc7 34.Rge1 a5 35.a3 Qd7 36.Rcd1 axb4 37.axb4 Qd5 38.Rd3 Qc4 39.Re4 Qxb4 40.Rdxd4 Qc3+ 41.Kg2 Qxc2 42.Rc4 Qb1 43.e6 fxe6 44.c6 bxc6 45.Rxe6 Qb5 46.Rexc6 Qd5+ 47.Kg3 Qe5+ 48.Rf4 ) 24.Kh2 {+2.33/11 1s} 24...g5 {-5.40/14 5s (Nf3+)} 25.Nc7 $2 {+4.46/12 7s Stockfish 16.1: 0.00/40} ( 25.hxg5 {Stockfish 16.1: +3.51/26} 25...h6 26.gxh6 Re6 27.Nxe6 Qxe6 28.fxe3 Nf3+ 29.Kg3 Ng5 30.Rh2 Nce4+ 31.Kg2 Qxg4+ 32.Kh1 Qf3+ 33.Rg2 Nf2+ 34.Kg1 Nh3+ 35.Kh2 Qh5 36.h7+ Kh8 37.Qb2+ Kxh7 ) 25...Re5 $4 {-6.33/13 2s Stockfish 16.1: +4.88/23} ( 25...gxf4 {Stockfish 16.1: 0.00/40} 26.Nxe8 Qxe8 27.Bxf4 Ne2 28.fxe3 Nxc1 29.Raxc1 Qe7 30.g5 h5 31.Rcd1 Qd7 32.a3 a6 33.a4 Kg7 34.Be5+ Kh7 35.Bg3 Qd5 36.Bf4 Nf3+ 37.Kg3 Nd2 38.Kh2 ) 26.Nh5 {+3.51/12 5s} 26...Rf5 {-1.42/15 3s} 27.Kh3 $2 {+3.09/12 0s Stockfish 16.1: -1.53/31} ( 27.f4 {Stockfish 16.1: +4.43/26} 27...gxf4 28.Rf1 Nxf1+ 29.Qxf1 Qxc2+ 30.Kh3 Ne4 31.Nxf4 Nf2+ 32.Kg2 Nxg4+ 33.Qe2 Qxe2+ 34.Nxe2 Rf2+ 35.Kg3 f5 36.Nd4 Nf6 37.Nce6 Kf7 38.Ng5+ Kg6 39.Ngf3 f4+ 40.Kh3 ) 27...Qf3+ $2 {-3.94/13 3s Stockfish 16.1: 0.00/44} ( 27...Ne2 {Stockfish 16.1: -1.53/31} 28.Qd1 Qf3+ 29.Bg3 Rf4 30.Kh2 Qxg4 31.Qxd2 exd2 32.Bxf4 Qxh4+ 33.Kg2 Nxf4+ 34.Kf3 Nh3 35.Nf6+ Kf8 36.Ncd5 g4+ 37.Ke2 Qxf2+ 38.Kd3 Nf4+ 39.Ke4 Qf3+ 40.Ke5 Ng6+ 41.Kd4 g3 42.Rhd1 g2 43.Rxd2 Qf1 44.Rxf1 gxf1=Q 45.Nxh7+ Kg7 46.Nhf6 Nf4 47.Ne8+ Kf8 48.Nd6 Qe1 49.Nc4 Ne2+ 50.Kd3 Nc1+ 51.Kc3 ) 28.Bg3 {+2.78/11 1s} 28...Rf4 $1 {-5.93/14 3s (Nce4)} 29.Kh2 {+3.22/11 6s} 29...gxh4 $4 {-3.95/14 2s (Rxg4) Stockfish 16.1: +6.63/27} ( 29...Qxg4 {Stockfish 16.1: 0.00/49} 30.Nxf4 gxf4 31.Ne8 fxg3+ 32.fxg3 Nce4 33.Nf6+ Nxf6 34.Qd1 Qe4 35.Kh3 Qf5+ 36.g4 Qe4 37.Rf1 Nxf1 38.Qxf1 Qxg4+ 39.Kh2 Qxh4+ 40.Kg1 Kf8 41.Qe2 Ng4 42.Rf1 Qg5 43.Qf3 Nf2+ 44.Kh2 Qh6+ 45.Kg2 Qg6+ 46.Qg3 Qc6+ 47.Kh2 Qh6+ 48.Kg2 Qc6+ ) 30.Bxf4 {+5.92/12 3s} 30...Qxf2+ {-1.72/16 4s} 31.Kh3 {+5.43/12 0s} 31...Qf3+ {-0.79/15 2s (Ne2)} 32.Kxh4 {+10.81/10 3s} 32...Nce4 {+2.25/14 3s (Qf2+)} 33.Rh2 {+13.36/13 7s} 33...Nf2 {+5.37/14 4s (Nc4)} 34.Nf6+ {+21.59/11 12s} 34...Kg7 {+6.71/15 3s (Kf8)} 35.Qb2 {+M-1373/10 2s} 35...Kg6 {+21.85/14 1s} 36.Ncd5 {+M-1373/7 0s} 36...Nc4 {+13.06/12 2s (Qe4)} 37.Ne7+ {+M-1379/4 0s} 37...Kg7 {+32.35/14 3s} 38.Bh6+ {+M-1380/4 0s} 38...Kh8 {-M-537/32 1s} 39.Nh5+ {+M-1381/2 0s} 39...Ne5 {-M-536/35 1s (Nxb2)} 40.Bg7# {+M-1382/1 0s White loses 0.16 pawn per move (2 mistakes, 1 inaccuracy), Black loses 0.43 pawn per move (2 blunders, 4 mistakes, 4 inaccuracies).} 1-0
[/pgn]

[Leo]

I downloaded this engine into Areana. For me it doesnt think. It plays the moves in 1 second. I dont understand.

[Alexander Schmidt]

You need to set a timecontrol, not just a movetime.