Skip to content

What a known tile is worth under row scan

Under a row scan, one known tile removes between 1 and 10 bits of the full search tree depending only on where its cell sits in the scan (median 5.8 bits, on painted 16x16 boards), and several tiles together are worth less than the sum. On those boards, 10 spread tiles leave the tree above 10^15 nodes and 16 to 24 bring it under, a figure checked end to end only. Nothing here is a measurement of the real puzzle beyond two transfer checks.

by Raphaël Anjou

experimentmeasurementmeasuredReduce the searchBacktrackingUpdated 2026-11-14
Reproduceseeded: reproduces with the given seedreruns the searchBudget: about 10.1 core-hours for the 16x16 pin study, about 1.4 core-hours for the small-board map
Hardware & run
Native runCPU only
Cores
10
GPUs
0
CPU
Apple silicon laptop (10 cores) and an 8-core Apple M1
Machine
10-core Apple silicon laptop, at most 2 processes at a time, plus an 8-core Apple M1 for three stages; the small-board map ran on one core
Budget
about 10.1 core-hours for the 16x16 pin study, about 1.4 core-hours for the small-board map
Start
seeded; painter seeds 300051, 300061, 300097, 300117, 300131 for the 16x16 boards; small boards seeds 1 to 20 (8x8) and 1 to 5 (10x10)
Reproduce this result

seeded: reproduces with the given seed

just experiments orders-and-pins all

Re-runs the 16x16 pin stages (one-pin map, leave-one-out clues, spread and random pin sets, whole-tree check) and then analyze.py; seeds are fixed and rows are deterministic, but the full run is about 10 core-hours. The small-board map is a second recipe, just experiments pin-map all (about 1.6 core-hours). Neither run touches the real pieces beyond the five clues.

Keep exploring

Referenced by

Page sourceView as Markdown