STS (v4.0) - "Square Vacancy"

[opraus]

Update for Xpdnt [not yet released]

It now recognizes the alternative moves, and disambiguation

90/100 on core-2 duo 2Ghz @ 30 sec /pos

missed the following completely

Code: Select all

6k1/p2pp2p/bp4n1/q1r4R/1RP1P3/2P2B2/P2Q2P1/4K3 w - - bm Rd5; c0 "Rd5=10, Rf5=6, g4=7"; id "STS(v4.0) Square Vacancy.001";
Q7/2pq3k/1rp2b1p/2R5/8/1P1NP3/P2K1Pr1/5R2 w - - bm Qf8; c0 "Qf8=10, Qa4=5, Rc2=4"; id "STS(v4.0) Square Vacancy.025";
q5k1/p2p2bp/1p1p2r1/2p1np2/6p1/1PP2PP1/P2PQ1KP/4R1NR b - - bm Qd5; c0 "Qd5=10, Re6=4, a5=1, c4=2"; id "STS(v4.0) Square Vacancy.026";
3rk3/2p1r1p1/7q/p1p1P1R1/2B4P/4P3/PPQ5/6K1 w - - bm Qe4; c0 "Qe4=10, Kg2=4, b3=3"; id "STS(v4.0) Square Vacancy.039";
r5k1/1q4b1/6n1/3pp1Nn/1N3p2/3P2Pp/2QBP2P/1R4K1 w - - bm Qc5; c0 "Qc5=10, Nxh3=4"; id "STS(v4.0) Square Vacancy.053";
2rqrbk1/1b1n1p2/p2p1npp/1p6/4P2B/1B3NNP/PP1Q1PP1/3RR1K1 w - - bm Qf4; c0 "Qf4=10, Re2=3, a3=1, a4=2"; id "STS(v4.0) Square Vacancy.057";
5b1k/r7/Pq5p/2p5/3p3N/R2Q2P1/5PK1/8 w - - bm Qf5; c0 "Qf5=10, Ng6+=2, Qf3=8, Qg6=2"; id "STS(v4.0) Square Vacancy.061";
7r/q3kp2/2rp1p2/3RpPb1/1p2P1P1/1P1N4/P4R1P/1K2Q3 b - - bm Qa3; c0 "Qa3=10, Kd7=2"; id "STS(v4.0) Square Vacancy.085";
4r1k1/1rp2ppp/p3b3/1n1p4/1q1P1B2/1P3P1P/P1Q3P1/3RRNK1 w - - bm Qc6; c0 "Qc6=10, Ne3=4, Qd2=5, Qd3=5, Qf2=3"; id "STS(v4.0) Square Vacancy.088";
7k/ppr1qpb1/1n2pN1p/4P1p1/8/1B3QP1/PP3PKP/4R3 w - - bm Qh5; c0 "Qh5=10, Re2=3, Re3=1, h3=2"; id "STS(v4.0) Square Vacancy.097";

[Spock]

Onno x64

60 secs per move on QX6700 @ 2.93GHz

85 of 100 matching moves

Thank you. This is quite good score for a single CPU engine, It's on par with other commercial engines.

Yes - just got an identical 85 of 100 with Hiarcs 12.1 SP

[swami]

Update for Xpdnt [not yet released]

It now recognizes the alternative moves, and disambiguation

90/100 on core-2 duo 2Ghz @ 30 sec /pos

missed the following completely

Code: Select all

6k1/p2pp2p/bp4n1/q1r4R/1RP1P3/2P2B2/P2Q2P1/4K3 w - - bm Rd5; c0 "Rd5=10, Rf5=6, g4=7"; id "STS(v4.0) Square Vacancy.001";
Q7/2pq3k/1rp2b1p/2R5/8/1P1NP3/P2K1Pr1/5R2 w - - bm Qf8; c0 "Qf8=10, Qa4=5, Rc2=4"; id "STS(v4.0) Square Vacancy.025";
q5k1/p2p2bp/1p1p2r1/2p1np2/6p1/1PP2PP1/P2PQ1KP/4R1NR b - - bm Qd5; c0 "Qd5=10, Re6=4, a5=1, c4=2"; id "STS(v4.0) Square Vacancy.026";
3rk3/2p1r1p1/7q/p1p1P1R1/2B4P/4P3/PPQ5/6K1 w - - bm Qe4; c0 "Qe4=10, Kg2=4, b3=3"; id "STS(v4.0) Square Vacancy.039";
r5k1/1q4b1/6n1/3pp1Nn/1N3p2/3P2Pp/2QBP2P/1R4K1 w - - bm Qc5; c0 "Qc5=10, Nxh3=4"; id "STS(v4.0) Square Vacancy.053";
2rqrbk1/1b1n1p2/p2p1npp/1p6/4P2B/1B3NNP/PP1Q1PP1/3RR1K1 w - - bm Qf4; c0 "Qf4=10, Re2=3, a3=1, a4=2"; id "STS(v4.0) Square Vacancy.057";
5b1k/r7/Pq5p/2p5/3p3N/R2Q2P1/5PK1/8 w - - bm Qf5; c0 "Qf5=10, Ng6+=2, Qf3=8, Qg6=2"; id "STS(v4.0) Square Vacancy.061";
7r/q3kp2/2rp1p2/3RpPb1/1p2P1P1/1P1N4/P4R1P/1K2Q3 b - - bm Qa3; c0 "Qa3=10, Kd7=2"; id "STS(v4.0) Square Vacancy.085";
4r1k1/1rp2ppp/p3b3/1n1p4/1q1P1B2/1P3P1P/P1Q3P1/3RRNK1 w - - bm Qc6; c0 "Qc6=10, Ne3=4, Qd2=5, Qd3=5, Qf2=3"; id "STS(v4.0) Square Vacancy.088";
7k/ppr1qpb1/1n2pN1p/4P1p1/8/1B3QP1/PP3PKP/4R3 w - - bm Qh5; c0 "Qh5=10, Re2=3, Re3=1, h3=2"; id "STS(v4.0) Square Vacancy.097";

Neat improvement. From 74 to 90/100 in single version update. This score really surpasses many strong freewares. Did you tune the evaluation parameters with the goal to score more in test suite or did you just impart knowledge into it? Wondering how much elo this version would be relative to the previous one though!

[swami]

Can you post all the results you have ?

Will post the results as soon as the test gets over.

[swami]

4 seconds each move:

The King Trailblazer - 71
Baron - 70

King of Kings - 68
Alfil - 68
ETChess - 68
Aristarch - 68
Spike - 67
Scorpio - 67
Cerebro - 67
Alaric - 65
Tao - 65
Chiron - 64
Bugchess - 63
Ufim - 63
Pseudo - 63
Horizon - 62
Movei - 62
Bison - 61
SOS - 61
Little Goliath Revival - 61
Little Thought - 60
Yace Paderborn - 60

Anmon - 59
Hamsters - 59
Arasan - 58
Gaia - 58
GLC - 56
Knight Dreamer - 56
Abrok - 55
Homer - 55
Natwarlal - 54
Pepito - 51

Pupsi - 49
TCB - 49
NOW -48
Nejmet - 45

[opraus]

Several explanations:

1. My built-in 'runsuite()' function now recognizes the 'other' moves as correct answers. This probably accounts for most of the apparent improvement.

2. My built-in 'runsuite()' function now also recognizes disambiguation. [ie, it would not equate Rd1 with Red1] Which earned another several points.

3. At longer time control, it will change it's mind, because the scores are close for many variations. Some correct at 30 sec, will fail @ 300 sec

4. Fixed at least 2 bugs, which might have gotten [or lost] an answer or two.

Current scores for your other tests @ 30 sec/ pos for comparison:

open lines - 81%
centralization - 82%
undermine - 87%