eval weakness (passed pawns)

Get your specific HIARCS/Junior support questions answered here as well as up-to-the-minute news!

Moderators: Watchman, Mark Uniacke, mrudolf

Post Reply
bnemais
Member
Posts: 19
Joined: Sat Oct 13, 2007 6:20 pm

eval weakness (passed pawns)

Post by bnemais »

I've watched many hiarcs games, mostly blitz. I would say hiarcs exhibits a blindness to enemy passed pawns. There doesn't seem to be any UCI option to make adjustments in this area either, unfortunately.

I would say 20-33 percent of its losses are due to this fact. I've seen it dozens of times, often enough that I can predict the loss while hiarcs still evals the position favorable to itself.

I can't really comment on if it fails to produce its own passed pawns due to such an eval problem-- what I notice are the times when it allows enemy passed pawns and seems unconcerned, then ultimately loses because of it.
Andrew
Member
Posts: 112
Joined: Thu Aug 23, 2007 11:38 am
Location: Canberra, Australia

Re: eval weakness (passed pawns)

Post by Andrew »

bnemais wrote:I've watched many hiarcs games, mostly blitz. I would say hiarcs exhibits a blindness to enemy passed pawns. There doesn't seem to be any UCI option to make adjustments in this area either, unfortunately.

I would say 20-33 percent of its losses are due to this fact. I've seen it dozens of times, often enough that I can predict the loss while hiarcs still evals the position favorable to itself.

I can't really comment on if it fails to produce its own passed pawns due to such an eval problem-- what I notice are the times when it allows enemy passed pawns and seems unconcerned, then ultimately loses because of it.
Please post some examples if you find any, so they might hopefully be of use
for H12 !

Andrew
bnemais
Member
Posts: 19
Joined: Sat Oct 13, 2007 6:20 pm

Post by bnemais »

Here are some examples.

Here, hiarcs doesn't see white's compensation for the exchange. As in many examples, giving the passed pawn isn't the end of the world, but later play turns it into a monster because hiarcs doesn't see the problem with it.

1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.g3 e5 7.Nde2 b5 8.Bg2 Bb7 9.h3 Nbd7 10.a4 b4 11.Nd5 Nxd5 12.exd5 a5 13.O-O Be7 14.Be3 O-O 15.c3 f5 16.Rc1 bxc3 17.Rxc3 Ba6 18.Qd2 Nf6 19.Rc6 Bb7 20.Qc2 Kh8 21.Rc1 Bxc6 22.dxc6 Qc8 23.c7 d5 24.Qc6 Bb4 25.Bxd5 Ra6 26.Qb5 Nxd5 27.Qxd5 Bd6 28.Qc4 Be7 29.f4 Bf6 30.fxe5 Bxe5 31.Bd4 Re6 32.Nf4 Bxf4 33.gxf4 Rfe8 34.Be5 h6 35.Qd5 Kh7 36.Qxa5 Qb7 37.Qd2 Rc8 38.a5 Qb3 39.Rc3 Qa2 40.Rc2 Rg6+ 41.Kh2 Qa1 42.Rc1 Qa4 43.Qd5 Qb4 44.b3 h5 45.Rc2 Qe1 46.Rg2 Rxg2+ 47.Qxg2 g6 48.a6 h4 49.b4 Qxb4 50.Qd5 Qf8 51.a7 Qg8 52.Qd7+ Kh6 53.Bb2 Qf8 54.Kh1 Ra8 55.Bd4 Rc8 56.Ba1 Ra8 57.Be5 Rc8 58.Bc3 Ra8 59.Kg2 Rc8 60.Kf3 Ra8 61.Qe6 Kh5 62.Qf6 Qxf6 63.Bxf6 Kh6 64.Ke3 Kh7 65.Kd4 Kg8 66.Kd5 Kf7 67.Bxh4 Rc8 68.Kc6 Ke6 69.Kb7 Kd7 70.Bd8 g5 71.fxg5 f4 72.a8=Q Rxa8 73.Kxa8 f3 74.Kb7 Kd6 75.c8=Q Ke5 76.g6 f2 77.Qc5+ Ke4 78.Qxf2 Ke5 79.g7 Ke4 80.Qe1+ Kf4 81.g8=Q 1-0

After 21. Rc1.
[fen]r2q1r1k/1b2b1pp/2Rp1n2/p2Ppp2/P7/4B1PP/1PQ1NPB1/2R3K1 b - - 0 21[/fen]

The position is equal, but hiarcs sees a big plus. It eventually loses because it continues to be unconcerned about the monster passer.

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 d6 5.Be2 O-O 6.Nf3 e5 7.Be3 Ng4 8.Bg5 f6 9.Bh4 Qd7 10.O-O Na6 11.h3 exd4 12.Nxd4 Nh6 13.Qd2 Nc5 14.Rae1 c6 15.Bd1 Nf7 16.b4 Ne6 17.Bg4 Qe8 18.Bxe6 Bxe6 19.Nd5 Qd7 20.Nxe6 Qxe6 21.Nc7 Qxc4 22.Nxa8 Rxa8 23.Bg3 Bh6 24.f4 a6 25.Bf2 Bg7 26.Bd4 c5 27.bxc5 dxc5 28.Bf2 Rd8 29.Qb2 b5 30.Qb3 Qxb3 31.axb3 Rc8 32.Ra1 Nd6 33.Rfe1 Rc6 34.Rad1 Nf7 35.g3 c4 36.bxc4 bxc4 37.Bd4 Bf8 38.Bc3 Nd6 39.Bxf6 c3 40.Rc1 Nb5 41.Be5 Ba3 42.Rc2 Bc5+ 43.Kg2 Bd4 44.Bxd4 Nxd4 45.Ra2 c2 46.Rc1 Rc3 47.h4 a5 48.h5 gxh5 49.Kh3 a4 50.Kh4 a3 51.Kxh5 Kf7 52.Kg4 Ke6 53.Kh4 Kd6 54.g4 Kc5 55.f5 Kc4 56.Rca1 Kb4 57.Rc1 Kb3 58.Raa1 Ne2 59.Rxc2 Rxc2 60.Rb1+ Rb2 61.Ra1 Nd4 62.g5 a2 63.Kg4 Rb1 64.Rxa2 Rg1+ 65.Kf4 Kxa2 66.Ke5 Nf3+ 67.Kf4 Nxg5 68.Ke5 h5 69.Kf4 h4 70.e5 h3 71.e6 h2 72.Ke5 h1=Q 73.Kd6 Qb7 74.e7 Rd1+ 75.Ke5 Qxe7+ 76.Kf4 Rg1 0-1

After 36... bxc4
[fen]6k1/5nbp/p1r2pp1/8/2p1PP2/6PP/5B2/3RR1K1 w - - 0 37[/fen]

Most engines agree hiarcs has a great position here, but after several inaccurate moves due to not seeing the passer threats, it is lost.

1.d4 Nf6 2.Nf3 g6 3.c4 Bg7 4.Nc3 d5 5.Qb3 dxc4 6.Qxc4 O-O 7.e4 a6 8.Be2 b5 9.Qb3 c5 10.dxc5 Bb7 11.O-O Nxe4 12.Nxe4 Bxe4 13.Bg5 Nc6 14.Qe3 Qd5 15.Rad1 Qe6 16.Bh6 Bc2 17.Rd2 Qxe3 18.Bxe3 Bf5 19.Rfd1 Rac8 20.h3 Nb4 21.g4 Be6 22.a3 Nc6 23.b4 Bc3 24.Rc2 Bg7 25.Rd3 a5 26.bxa5 Nxa5 27.Bd2 Nb3 28.Bb4 Bc4 29.Re3 Na1 30.Rxc4 bxc4 31.Bxc4 e6 32.Re2 Rc7 33.Ng5 Bh8 34.Kg2 Bf6 35.Ne4 Bg7 36.Kg3 Rb8 37.Rd2 Rc6 38.f4 Rcc8 39.Nd6 Rc7 40.Kf3 Ra8 41.f5 exf5 42.gxf5 Rf8 43.Rg2 Be5 44.fxg6 hxg6 45.Rg5 Bg7 46.Rxg6 Kh7 47.Rg2 Be5 48.Bd5 f5 49.Re2 Re7 50.c6 Bxd6 51.Bxd6 Rxe2 52.Kxe2 Re8+ 53.Kd2 Nb3+ 54.Kc3 Nc1 55.c7 Ne2+ 56.Kc4 Kg6 57.Bb7 Kg5 58.c8=Q Rxc8+ 59.Bxc8 f4 60.a4 f3 61.Bg4 Ng1 62.Bc5 Ne2 63.Bxf3 Ng3 64.a5 Nf1 65.Bg2 Nd2+ 66.Kc3 Nb1+ 67.Kc2 Na3+ 68.Bxa3 Kf5 69.a6 Kg5 70.a7 Kf5 71.a8=Q Ke5 72.Bb2+ Kd6 73.Qd8+ Kc5 74.Qd5+ Kb6 1-0

After 9. Qb3
[fen]rnbq1rk1/2p1ppbp/p4np1/1p6/3PP3/1QN2N2/PP2BPPP/R1B1K2R b KQ - 0 9[/fen]

Here, hiarcs comes up with the silly 9... c5. That move isn't clearly losing, but it really shouldn't be considered due to the inherent danger. This time, the real danger doesn' come until move 33, but again, hiarcs doesn't see the danger quickly enough to save itself.

I can provide many more examples given enough time.
Andrew
Member
Posts: 112
Joined: Thu Aug 23, 2007 11:38 am
Location: Canberra, Australia

Post by Andrew »

Thanks for posting these, hopefully they'll help improve H12 !

Andrew
Jack Clark
Member
Posts: 21
Joined: Thu Aug 02, 2007 1:48 am
Location: Idyllwild, California

HIARCS' apparent underevaluation of passed pawn dangers

Post by Jack Clark »

Bnemais, I ran your last example wherein HIARCS allowed a passed pawn by 9...c5 and I note that HIARCS' was still in its opening book at the time. It didn't come out of book until move 19. But even when it came out of book, it didn't seem concerned.

(I'm running SIGMA CHESS 6.1.5 HIARCS 11.2 MP on a dual-core MacBook with 2GB of RAM, if it matters.)
Jack Clark
Idyllwild, California
bnemais
Member
Posts: 19
Joined: Sat Oct 13, 2007 6:20 pm

Re: HIARCS' apparent underevaluation of passed pawn dangers

Post by bnemais »

Has anyone from the Hiarcs team actually looked at this issue? I can post more examples if needed. I told myself I was simply going to wait, but continuing to see Hiarcs lose because of this is most irritating, esp. now that I can predict it way in advance.
Jack Clark wrote:Bnemais, I ran your last example wherein HIARCS allowed a passed pawn by 9...c5 and I note that HIARCS' was still in its opening book at the time. It didn't come out of book until move 19. But even when it came out of book, it didn't seem concerned.
I've already privately discussed this with Jack, but I did want to clarify for others thinking this may be an invalid example. It is indeed valid. My analysis had books off, so I missed that Hiarcs was in book. But regardless, it plays c5. My point wasn't so much that c5 is bad, but the same as Jack's last point: that Hiarcs continues to ignore the passer threat until the opponent has too big an advantage.
Chesslet
Posts: 1
Joined: Wed Aug 01, 2007 9:30 am

Passed pawn weakness

Post by Chesslet »

Hello,

I am using Hiarcs too since quite a while and I can confirm that Hiarcs detects passed pawns very late. At least later than some other engines.
My impression is that Hiarcs looses about 10-15% of the games due to that weakness, especially against Rybka.

Kind regards,
Chesslet
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

We are working on it!
Post Reply