version 0.6.0 Alpha quickly finds tactics

[tissatussa]

just by chance i happened to test your new version v0.6.0 and the PR-192 alpha version , giving both this position :

r1q2k1r/p3R1pp/2n5/1p3pB1/2pp4/2P5/3Q1PPP/4R1K1 w - - 0 21

here Bh6 is clearly best move.
newest v0.6.0 finds it at depth 14 after 20 seconds,
Alpha even at depth 11 after 4 seconds !?

i test in CuteChess with infinite time.
a few strong engines do NOT find the "solution" to this position, or only after a long time.

i don't know if this can help you in any way.
it just happened here and i want to share it.
is LMR involved ? i have no clue ..

[ i'm on Xubuntu 22.04 ]

[FitzOReilly]

Interesting position. Thanks for letting me know!

I tested it with a few older versions: Both 0.4.0 and 0.5.0 find it at depth 9. It's probably due to pruning. Version 0.6.0 adds late move reductions and futility + reverse futility pruning at higher depths.

[tissatussa]

here are 3 other puzzle positions which i used to test engines - they all have a "solution" which is clearly best.
see the results for Fatalii v0.6.0 and its Alpha below.
i attached my ZIPped screenshots, which give evaluation details when depth is climbing : 3-puzzles.zip

---

r1bqk2r/pp1n2bp/2p3p1/4pp2/2P5/2N1PNP1/PPQ2PP1/2KR1B1R w kq - 0 1

solution : g4 (=clearly best move)
0.6.0 finds it after 9 minutes at depth 17
Alpha finds it after 3 seconds at depth 9 (and 10,11) BUT then it finds OTHER moves to be best..

---

7Q/p2r4/kp6/4Bp2/P3b3/KP1q4/8/2R5 w - - 0 1

solution: Rd1 (=clearly best move)
0.6.0 finds it after 2 seconds at depth 13
Alpha finds it after 27 seconds at depth 15

---

4k3/2p1np2/p4qrp/3Bp3/4P1b1/2P1Q3/Pr1N1PP1/1R2R1K1 b - - 0 1

solution: Nxd5 (=clearly best move)
0.6.0 finds it after 94 seconds at depth 17
btw. BUT after 7 minutes at depth 18 it sees this move is CLEARLY better for Black.
Alpha finds it within 3 minutes at depth 14
btw. BUT after 13 minutes at depth 17 it sees this move is CLEARLY better for Black.