diff options
| author | Scott Gasch <[email protected]> | 2026-08-29 00:30:53 -0700 |
|---|---|---|
| committer | Scott Gasch <[email protected]> | 2026-08-29 00:30:53 -0700 |
| commit | 917bf1260ae217bdcbdbc11c2c37cdb5d3a8eae9 (patch) | |
| tree | 8013781f51e53eaf1fe3d4f45367e920d4c39786 /tests/ecm_ringers.ep_ | |
| parent | 366eebaf6b72fc6b56ea06a71f0276069a501dc7 (diff) | |
Switch to Crafty-style killer ordering; fix mvNullmoveRefutations type-mixing bug and add a quiet-refutation killer backfill.
Killer tiers now try both of this ply's own killers before either
ply-2-back one, matching Crafty's ordering. Two earlier attempts at
this same swap were reverted for regressing; this pass lands on top of
NumLeftoverMovesToSelect (more SelectBestWithHistory budget to reach
these lower-tier slots) and a real bug fix below, and beats interleaved
order head-to-head on solves, node count, and first-move beta cutoff
across the three curated suites.
The bug: mvNullmoveRefutations's empty-killer-slot backfill could only
ever contain a capturing move (TryNullmovePruning only wrote it inside
the capture-refutation branch), but IS_SAME_MOVE's mask includes the
pCaptured bits, so that backfilled value could never match a real
quiet candidate -- the backfill was silently dead code. Fixed by
recording genuinely quiet null-move refutations into a new, separate
mvNullmoveQuietRefutations array (kept separate so it can't clobber the
capture history mvNullmoveRefutations still needs for the
Botvinnik-Markoff same-piece-two-squares extension check) and
backfilling the regular killer table from that instead. The
check-evasion killer table intentionally does *not* get this backfill:
a null-move refutation can never legitimately be an escaping-check
move (null moves can't deliver check), so backfilling there risks
IS_SAME_MOVE cross-context false positives instead of the old
guaranteed-inert no-op.
Measured at sd10 across ecm_ringers/ecm_confident_quick/ecm_hard_quick
against head_reference (commit d11e973): 115/191 solves (vs. 116
baseline), 924.36M total nodes (vs. 933.23M), first-move beta cutoff
within 0.1-0.9 points of baseline on all three suites -- and clearly
better than the same fix under interleaved order (113/191 solves,
963.10M nodes), which loses to head_reference on every metric.
Co-Authored-By: Claude Sonnet 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01EortUUkDVpsfrbqshBJYJg
Diffstat (limited to 'tests/ecm_ringers.ep_')
| -rw-r--r-- | tests/ecm_ringers.ep_ | 44 |
1 files changed, 44 insertions, 0 deletions
diff --git a/tests/ecm_ringers.ep_ b/tests/ecm_ringers.ep_ new file mode 100644 index 0000000..e8fd0e7 --- /dev/null +++ b/tests/ecm_ringers.ep_ @@ -0,0 +1,44 @@ +setboard k3r3/pR5p/PppR1p1p/4nP2/1PP5/8/4B1rP/2K5 w - - +solution Re7 +id "ECM.213" +go +setboard 6k1/1p3p2/4r1p1/1PPrbq1p/2QN4/P2R1P1P/R4KP1/8 b - - +solution Bg3+ +id "ECM.270" +go +setboard 8/5p2/4p1k1/4P1p1/2rBRbK1/5P2/r4P2/7R b - - +solution Rxd4 +id "ECM.335" +go +setboard 4rqk1/p5b1/3p3p/3Pn1p1/1pP1NrP1/4NP2/P3Q1P1/1K1RR3 b - - +solution Nxf3 +id "ECM.501" +go +setboard R2b2k1/2rq1n1p/3p2p1/1p1Ppp1n/1P6/1B1PQN1P/1B3PP1/6K1 w - - +solution Nxe5 +id "ECM.406" +go +setboard r1qb1r1k/2p3pp/p1n1bp2/1p1Np2Q/P3P3/1BP3R1/1P3PPP/R1B3K1 w - - +solution Rxg7 +id "ECM.717" +go +setboard 4rrk1/2pn2pb/p1p1qp2/1pb1pN2/P3P1PN/1P1P4/1BP2PK1/R2Q3R w - - +solution Nxg7 +id "ECM.721" +go +setboard r3k2r/p1qb1p2/1p2p2p/3pPpN1/P1nP3Q/8/2P2PPP/R1B1R1K1 w kq - +solution Nxf7 +id "ECM.746" +go +setboard 2r1r1k1/5ppp/pq3b2/2pB1P2/2p2B2/5Q1P/Pn3PP1/2R1R1K1 w - - +solution Bxf7+ +id "ECM.750" +go +setboard 6rk/3b1n1p/1p1q3b/1PpNp3/2P1Pp2/2Q2NrP/5RP1/2R2B1K b - - +solution Bxh3 +id "ECM.783" +go +setboard r6r/ppnqpk2/3pbpp1/5N2/1PP1P3/5RR1/P2QB1PP/6K1 w - - +solution Nh6+ +id "ECM.801" +go |
