Couple more ideas

[Lyudmil Tsvetkov]

Fischer used to fianchettoe kingside in more than half of his games.

Only when black. Pulling up Fischer's games in a database:

As white: 558 games, of which 66 feature a king-side fianchetto (defined as a bishop on g2 and a pawn on g3 on or before move 15)

As black: 396 games, of which 236 feature a king-side fianchetto.

Bobby sure loved his KID

Of course this is due to Bobby being a 1.e4 guy… so perhaps the issue is that one should only fianchetto in closed/semi-closed openings.

You might also want to check numbers for Kasparov, Smyslov, Petrosian, etc., but also for practically every world champion after 1945.
Certainly king fianchettoes there are more than for the average GM.

The reason why many players with white do not fianchettoe their bishops is of course the availability of e4, leading to Ruy Lopez and Sicilian.

However, I think the last word with white king fianchettoes is still not said, and in a while people will discover that in many more lines white kingside fianchettoe gives a more decisive advantage than otherwise.

You amaze me with your stats, how are you able to do that?
You provide very valuable information.

[PK]

Perhaps the problem with fianchetto is that evaluation function must deal with two completely different scenarios, and we haven't found a way to reconcile them yet. There is Grunfeld/Catalan scenario, where we want to keep the f-pawn on its initial square and use the fianchettoed bishop to exert pressure to opponent's queenside. On the other hand, there is a KID scenario, where the fianchetto formation is built and then destroyed for the greater good of the kingside attack. Then there are hybrids, like Botvinnik's triangle (c4-d3-e4) which plays much like a KID, or Leningrad Dutch, where White can prevent forming the ideal formation with a fast d4-d5.

Giving a static bonus for a fianchetto helps in the Grunfeld scenario, but backfires in the KID scenario. I think that at the beginning a rule like "apply static fianchetto bonus only when d and e pawns are not in direct contact, i.e. not rammed and not attacking each other" might make sense. If this works, we may start thinking about the hybrids.

[Lyudmil Tsvetkov]

Perhaps the problem with fianchetto is that evaluation function must deal with two completely different scenarios, and we haven't found a way to reconcile them yet. There is Grunfeld/Catalan scenario, where we want to keep the f-pawn on its initial square and use the fianchettoed bishop to exert pressure to opponent's queenside. On the other hand, there is a KID scenario, where the fianchetto formation is built and then destroyed for the greater good of the kingside attack. Then there are hybrids, like Botvinnik's triangle (c4-d3-e4) which plays much like a KID, or Leningrad Dutch, where White can prevent forming the ideal formation with a fast d4-d5.

Giving a static bonus for a fianchetto helps in the Grunfeld scenario, but backfires in the KID scenario. I think that at the beginning a rule like "apply static fianchetto bonus only when d and e pawns are not in direct contact, i.e. not rammed and not attacking each other" might make sense. If this works, we may start thinking about the hybrids.

Hi Pawel.

Did not fianchettoe work in Rodent?
Or maybe some problems appeared again.

I think it is part of the bigger picture, with all terms being tuned together, interactions with king safety, etc. For example, SF likes to fianchettoe its queenside bishop, but not its king bishop. Why?
Because its queen bishop usually attacks on the long diagonal the enemy king shelter, and the engine gets bonus for this. It also likes to develop its king bishop to d6 or c5, where it attacks the enemy king shelter, but on g7 does not do so.

So you need an additional bonus for king fianchettoe, but also tuned to reflect interactions with other terms, not in the last place also mobility and psqt.

[Lyudmil Tsvetkov]

Also, kingside fianchettoe is closely king safety related.

Of course, part of the importance of king fianchettoe is due to the bishop attacking enemy pieces on the long diagonal and controlling the center, and part to making the king shelter stronger. I think the split between the two is somewhere around 50-50.

So, kingside fianchettoe is closely king safety related and I would even treat it as a king safety term.
King safety is not all about pawn shelter and storming pawns, it is also about kingside fianchettoe, king binds, etc. A shelter involving a king fianchettoe is 30% more difficult to attack and destroy than in the usual case, so a big benefit from it. Not only you have a minor piece within the shelter, but this minor piece and the pawns around it complement each other by controlling squares of different colour.

So nothing more natural than to give a king safety bonus for king fianchettoe.

I would be surprised if this would not work in SF after repeated attempts. It will of course, as this is a sound term. It is only a matter of implementation and tuning of parameters. And the first thing I would like to tune king fianchettoe bonus with is pawn shelter bonus/penalty.

[vincenegri]

You amaze me with your stats, how are you able to do that?
You provide very valuable information.

It's very easy if you have a large database and some good software that allows sophisticated queries. SCID will do it, but Chess Assistant and Chessbase are solid commerical options. Large databases of historical games are available legitimately for free on the internet (since the unannotated games are not subject to copyright) but imho if your budget allows it I would pay for a subscription service that takes the hassle out of keeping your database up to date.

If you are interested in following current trends or just historical research such a tool is invaluable. You can effectively investigate questions like "how often do people manage to win R&B vs R in practice?"

And of course, structured query tools allow the games from engine tests to be analysed in a way that would be simply impossible by hand. Imagine playing through 100,000 games yourself looking for certain pawn structures!!

[PK]

As for fianchetto eval in Rodent: general bonus for pieces on bishop's diagonal, works like a charm once fianchetto is in place. Right now I'm investigating whether a frequency of moves like g3/g6 can become greater without decreasing playing strength.

[Lyudmil Tsvetkov]

As for fianchetto eval in Rodent: general bonus for pieces on bishop's diagonal, works like a charm once fianchetto is in place. Right now I'm investigating whether a frequency of moves like g3/g6 can become greater without decreasing playing strength.

Yeah, yeah, yeah, every second move you can play g3/g6.

[Dann Corbit]

Some analyzed positions from this thread:

Code: Select all

rnbqk1nr/pppp1ppp/4p3/8/1b4P1/3P4/PPP1PP1P/RNBQKBNR w KQkq - acd 33; acs 600; bm Bd2; ce -22; pm Bd2; pv Bd2 Be7;
r4rk1/ppqn1pp1/2pbpn1p/8/2PP4/3Q1NN1/PP1B1PPP/3RR1K1 w - - acd 35; acs 600; bm Ne4; ce 19; pm Ne4; pv Ne4 Be7 Nc3 Rfd8 Qe2 a6 g3 Qb6 Kg2 Bf8 Na4 Qc7 c5 Be7 Rb1 a5 Bf4 Qc8 Rbc1 Nd5 Bd2 Re8 Nc3 N5f6 h4 Rd8 Bf4 b6 Ne5 Nxe5 dxe5 Nd7 cxb6 Nxb6 Ne4;
r2qkb1r/pp1nnppb/2p1p2p/3pP3/3P4/2P2N2/PP1NBPPP/R1BQ1RK1 w kq - acd 38; acs 600; bm b4; cce -16; ce 36; pm b4; pv b4 a5 Nb3 axb4 cxb4 Nc8 Bd2 Ncb6 a4 Be7 a5 Nc4 Bc3 O-O Nfd2 Nxd2 Qxd2 Bg6 f4 Be4 Qe3 Qc7 Rac1 Rfc8 Bd3 f5 g4 Rf8 gxf5 Bxf5 Bxf5 Rxf5 Qg3 Kh7 Rc2 Raf8 Bd2 Qd8; white_wins 14; black_wins 16; draws 24; 
r1q2r1k/1bp1bp1p/1pn1p1p1/pB1pP1Pn/P2P3P/1NP1BN2/1P2QP2/R3K2R w KQ - acd 39; acs 600; bm Qc2; ce 23; pm Qc2; pv Qc2 Na7 Be2 Ba6 Nh2 Bxe2 Qxe2 c5 Ng4 Nc6 Nf6 Bxf6 gxf6 Qc7 Rc1 Qd7 Rd1 Rfc8 O-O cxd4 Bxd4 Rab8 Ra1 Nxd4 Nxd4 Rc4 Rac1 Rbc8 Qd2 R4c5 Ra1 Rc4 Kg2 Rxa4 b3 Rxa1 Rxa1 Rc5;
6k1/8/p3p3/Pp1pP3/1PpP4/2P5/8/6K1 w - - acd 87; acs 600; bm Kg2; ce 0; pm Kg2; pv Kg2 Kh8 Kf3 Kh7 Kg4 Kg6 Kf4 Kh6 Kg4 Kg6;
6k1/8/8/p1p1pP2/PpPpP3/1P1P4/8/6K1 w - - acd 48; acs 3; bm Kf2; ce 32728; pm Kf2; pv Kf2 Kf7 Kg3 Ke7 Kg4 Kf6 Kh5 Kg7 Kg5 Kf7 f6 Kf8 Kf5 Kf7 Kxe5 Kg6 Ke6 Kg5 f7 Kf4 f8=Q+ Ke3 Qxc5 Kxd3 Kd5 Kc2 Qxd4 Kxb3 c5 Ka2 Kc4 b3 Kc3 b2 Qf2 Ka3 Qxb2+ Kxa4 Qb3#;
6k1/8/8/4P2P/3P2P1/2P2P2/8/6K1 w - - acd 26; acs 0; bm d5; ce 32746; dm 11; pm d5; pv d5 Kf7 d6 Ke6 h6 Kd7 h7 Ke6 h8=Q Kd7 Qg8 Kc6 Qc4+ Kb7 Qb5+ Ka7 d7 Ka8 d8=Q+ Ka7 Qbb6#;
6k1/8/8/3PP3/2PP4/8/8/6K1 w - - acd 28; acs 0; bm Kf2; ce 32744; dm 12; pm Kf2; pv Kf2 Kf7 c5 Ke7 c6 Ke8 d6 Kf7 c7 Ke6 c8=Q+ Kd5 Qc5+ Ke6 Qc4+ Kd7 Kf3 Ke8 Qc7 Kf8 Qh7 Ke8 Qe7#;
6k1/8/8/3PP3/2P2P2/8/8/6K1 w - - acd 27; acs 0; bm c5; ce 32744; dm 12; pm c5; pv c5 Kf7 c6 Ke7 Kf2 Kd8 d6 Kc8 e6 Kb8 e7 Ka7 e8=Q Kb6 c7 Kc5 Qe5+ Kb4 c8=Q Ka3 Qcc3+ Ka4 Qea5#;
6k1/8/8/3P4/4P3/3P4/8/6K1 w - - acd 47; acs 7; bm Kf2; ce 32736; pm Kf2; pv Kf2 Kf7 Ke3 Ke7 Kd4 Kd6 e5+ Kc7 e6 Kd6 Kc4 Ke7 Kc5 Kf6 Kd6 Kg5 e7 Kf4 e8=Q Kf3 Qe1 Kf4 Ke6 Kf3 Kf5 Kg2 Kg4 Kh2 Kf3 Kh3 Qh1#;
6k1/8/8/3P4/2PBP3/3P4/8/6K1 w - - acd 26; acs 0; bm e5; ce 32748; dm 10; pm e5; pv e5 Kf8 c5 Ke7 c6 Ke8 c7 Kd7 d6 Kc8 e6 Kb7 e7 Kc6 e8=Q+ Kxd6 c8=Q Kd5 Qed7#;
6k1/8/8/3P4/2P1P3/8/8/6K1 w - - acd 4; acs 1; bm c5; ce 32744; dm 12; pm c5; pv c5;
6k1/8/8/3P4/2P1P3/3P4/8/6K1 w - - acd 26; acs 0; bm d6; ce 32632; dm 68; pm d6; pv d6 Kf7 d7 Ke7 d8=Q+ Kxd8;
6k1/8/6P1/5P2/4P3/3P4/8/6K1 w - - acd 27; acs 0; bm f6; ce 32748; dm 10; pm f6; pv f6 Kf8 e5 Ke8 e6 Kd8 g7 Kc7 e7 Kb6 e8=Q Kc5 Qe5+ Kc6 g8=Q Kb6 Qb3+ Ka6 Qa1#;
6k1/8/5P2/4P3/3P4/2P5/8/6K1 w - - acd 98; acs 1326; bm e6; ce 32748; dm 10; pm e6; pv e6 Kf8 d5 Kg8 d6 Kh7 d7 Kg6 f7 Kf5 f8Q+ Kg5 d8Q+ Kg4 Qg7+ Kf5 Qgg5+ Kxe6 Qdd5+;
6k1/8/4P3/3P4/2P5/3P4/8/6K1 w - - acd 29; acs 0; bm d6; ce 32744; dm 12; pm d6; pv d6 Kg7 e7 Kf7 c5 Ke8 d4 Kf7 d5 Ke8 c6 Kf7 c7 Kf6 e8=Q Kf5 Qe6+ Kf4 c8=Q Kf3 Qc3+ Kf4 Qce3#;
6k1/8/3p4/p1p1pP2/PpP1P3/1P1P4/8/6K1 w - - acd 49; acs 22; bm Kg2; ce 32734; pm Kg2; pv Kg2 Kf8 Kg3 Kg7 Kh4 Kh6 Kg4 Kh7 Kg5 Kg7 f6+ Kg8 Kg6 Kf8 f7 d5 cxd5 c4 dxc4 Ke7 c5 Kf8 c6 Ke7 c7 Kd7 d6 Kc6 c8=Q+ Kb6 f8=Q Ka7 Qf2#;
6k1/8/3P4/2P5/1P2P3/5P2/6P1/6K1 w - - acd 26; acs 0; bm d7; ce 32754; dm 7; pm d7; pv d7 Kg7 d8=Q Kf7 c6 Ke6 c7 Kf7 c8=Q Kg7 Qg5+ Kf7 Qcg8#;
6k1/8/2p1p3/1p1pP3/pP1P4/P1P5/8/6K1 w - - acd 73; acs 600; bm Kf2; ce 0; pm Kf2; pv Kf2 Kg7 Kg3 Kh7 Kf3 Kh6 Kf4 Kh5 Kf3 Kg5 Ke3 Kg4 Ke2 Kf4 Kf2 Ke4 Ke2 Kf4;
6k1/1p6/2p4p/2P3p1/2N3P1/6N1/8/6K1 w - - acd 29; acs 8; bm Nd6; ce 32612; dm 78; pm Nd6; pv Nd6 Kf8 Ngf5 Kg8 Nxh6+ Kh7 Nxb7 Kg7 Nd8 Kf6 Nxc6 Ke6 Nd4+ Kd5 Nf7 Ke4 c6 Ke3 c7 Kd2 c8=Q Kd3 Qc2+ Ke3 Qe2+ Kxd4 Nxg5;
2rqnrk1/1b1nbp1p/p3p1p1/1p1pP1P1/1P1N1P2/P2B4/2P1N2P/R1BQ1RK1 w - - acd 36; acs 600; bm Rf3; ce 31; pm Rf3; pv Rf3 Bxg5;