======================================================================== FRIDAY RELEASES, REPLICATED on a held-out population rule.py sha256 63906c4b11fec868b492e3378a67af5b9a7efb6e6f2a8fcbe80d49266be632e1 seal rule.friday-releases-replication id 6014, 1789586963188 ms, 2026-09-16 21:29 Rome; the first row was on disk at 21:30 ======================================================================== overlap with the first study's sample: 0 (the sealed rule asserts zero) FILTERS rows read 128,868 no file (deleted release) 211 version unparseable 7 pre-release 31,587 outside 2016..2026-09 4,007 kept 93,056 distinct projects kept 1,464 eligible releases (168 h horizon): 92,543 follow-up within 168 h, by day (reported): Mon any 10,314/15,372 = 67.10% patch 6,131/12,183 = 50.32% Tue any 11,802/17,339 = 68.07% patch 6,606/13,125 = 50.33% Wed any 11,348/16,420 = 69.11% patch 6,495/12,572 = 51.66% Thu any 11,911/17,148 = 69.46% patch 6,661/13,089 = 50.89% Fri any 10,737/15,330 = 70.04% patch 6,013/11,611 = 51.79% Sat any 3,693/ 5,726 = 64.50% patch 2,123/ 4,530 = 46.87% Sun any 3,315/ 5,208 = 63.65% patch 2,080/ 4,368 = 47.62% ------------------------------------------------------------------------ SCORED CLAIMS (survived lo > 1.0 and RR >= 1.03; killed CI inside (0.97, 1.03)) ------------------------------------------------------------------------ any follow-up within 168 h Friday 10,737/15,330 = 0.7004 Mon..Thu 45,375/66,279 = 0.6846 RR = 1.0231 95% CI [1.0113, 1.0350] half-width 1.157% (sealed ceiling 3.0%) VERDICT: INCONCLUSIVE patch-level follow-up within 168 h Friday 6,013/11,611 = 0.5179 Mon..Thu 25,893/50,969 = 0.5080 RR = 1.0194 95% CI [0.9997, 1.0395] half-width 1.952% (sealed ceiling 3.0%) VERDICT: INCONCLUSIVE NEGATIVE CONTROL — Tuesday against Mon/Wed/Thu (sealed, scored): any RR = 0.9922 [0.9805, 1.0040] clean patch RR = 0.9876 [0.9684, 1.0072] clean SECOND CONTROL — the weekend against Mon..Thu (direction only): Sat RR = 0.9421 [0.9235, 0.9610] Sun RR = 0.9298 [0.9103, 0.9496] ------------------------------------------------------------------------ REPORTED, registered, not scored ------------------------------------------------------------------------ weekday histogram: Mon 15,460 16.61% Tue 17,429 18.73% Wed 16,491 17.72% Thu 17,239 18.53% Fri 15,414 16.56% Sat 5,754 6.18% Sun 5,269 5.66% claim A out of sample: Friday's share of Mon..Fri 0.1879 [0.1852, 0.1906], n = 82,033 the FIRST study's bars say: INCONCLUSIVE (there it was 0.1899, inconclusive) the same claim at 336 h (also day-balanced): Friday RR 1.0168 [1.0072, 1.0265] control 0.9994 [0.9898, 1.0091] the working-days table that voided the first study's 72 h proxy, recomputed here (it should still hold — it is a fact about calendars): 1 working days in a 72 h window 0.5178 n = 32,478 2 working days in a 72 h window 0.5843 n = 22,146 3 working days in a 72 h window 0.6033 n = 37,919 yank rate by weekday: Mon 134/ 15,460 0.867% Tue 138/ 17,429 0.792% Wed 123/ 16,491 0.746% Thu 112/ 17,239 0.650% Fri 142/ 15,414 0.921% Sat 77/ 5,754 1.338% Sun 65/ 5,269 1.234% ======================================================================== ANY FOLLOW-UP: INCONCLUSIVE PATCH FOLLOW-UP: INCONCLUSIVE ========================================================================