Drop7 Research
← Experiments

CHECK gates for the one-entry leaf memo: bit identity, search parity, determinism, interleaved timing on real probe seeds

completedtier CHECKengineeringpreviously-evaluated-developmentseed-freeEX-20260822-leaf-memo-gates-b31e4e7f
No explanation has been written for this experiment yet. Add web/content/research/EX-20260822-leaf-memo-gates-b31e4e7f.mdx and it will appear here. The registered protocol is shown below.

The registered protocol

MemoSearch (fast-search.hpp with the leaf call replaced by fastFairLeafMemo below ++work_) passes: leaf bit identity in the search's feeding order on probe games at d4s5 and d4s7; parity with FastSearch on every move (column, work, nodes, cache hits, completed depth); identical results at 1 and 4 threads; interleaved timing ratio >= 1.5x on real roots.

Candidate
fast-engine-memo MemoSearch
approaches/lifetime-objective/fast-engine-memo/gate.cpp
Comparator
fair-d4
approaches/fair-expectimax/reference/fair-only-depth4.cpp

Primary metric

leaf bit mismatches and search parity mismatches (must be zero)

Statistical unit: not-applicable

Pass criteria, fixed in advance

  • 0 leaf bit mismatches at d4s5 and d4s7
  • 0 action/work/node/hit/depth mismatches vs FastSearch
  • 0 determinism mismatches across thread counts
  • timing ratio >= 1.5 at both strata settings (indicative)

On pass: Record pass (engineering, mechanics-only); the memo may be adopted by cohort runners after their own gates are re-run; an idle-host timing measurement is owed before any throughput claim.

On fail: Record fail; do not adopt the memo in any cohort runner.

Data and reuse

Probe games on previously opened development seeds 0xa5278000-0xa52784ff only (CHECK probes); no strength data; no new seed.

seed leases: none

What happened

valid run · outcome: passsupported-as-testedmechanics-onlyRS-20260822T074305Z-bd1697c8

One-entry leaf memo (fast-engine-memo/MemoSearch) on real probe seeds: leaf bits 0/4,260 (d4s5) and 0/2,500 (d4s7) mismatches in the search's feeding order; parity with FastSearch 0 action/work/node/cache-hit/completed-depth mismatches over 160 (d4s5, 117,732,775 leaf calls) and 50 (d4s7, 119,607,901 leaf calls) moves; determinism 0 mismatches at 1 vs 4 threads; memo hit rate 61.74% (d4s5) and 68.47% (d4s7); interleaved timing plain/memo 1.581x (d4s5, 12 real roots x 3 reps) and 1.634x (d4s7, 6 roots x 2 reps) under load ~31, indicative. Engineering result: semantics-preserving; no strength claim.

What it had to pass
  • 0 leaf bit mismatches at d4s5 and d4s7 — observed: 0/4,260 and 0/2,500
  • 0 parity mismatches vs FastSearch — observed: 0 over 160 and 50 moves
  • 0 determinism mismatches — observed: 0 over 4 games
  • timing ratio >= 1.5 at both strata (indicative) — observed: 1.581 and 1.634 under load
Recorded metrics
leafBits
d4s5
boards
4,260
mismatches
0
hitRateFeedOrder
0.5655
d4s7
boards
2,500
mismatches
0
hitRateFeedOrder
0.6392
parity
d4s5
moves
160
mismatches
0
leafCalls
117,732,775
hitRate
0.6174
d4s7
moves
50
mismatches
0
leafCalls
119,607,901
hitRate
0.6847
determinism
games
4
threads
  1. 1
  2. 4
mismatches
0
timingIndicative
d4s5
roots
12
reps
3
plainSeconds
7.1319
memoSeconds
4.5098
ratio
1.5814
d4s7
roots
6
reps
2
plainSeconds
12.5726
memoSeconds
7.6951
ratio
1.6338
Limitations
  • Timing ratios were measured on a host at load ~31 and are indicative; an idle-host, three-repeat measurement is owed before any throughput claim.
  • Probe cohorts are small (210 parity moves); the bit-identity argument is structural and the gates are confirmation, not the proof.
  • Not adopted by any cohort runner yet.