floornull - the A/A FLOOR under a random drop of the same size as the steal filter's, 64 KiB EPYC guest, steal < 5.00% legs 160, 20 cells of 8; the null at each rung is EXHAUSTIVE - every C(8, d) way of dropping d legs cell m d | floor0 v0 | effS floorS vS | nmed np5 np95 pctl n | P(vS) P(fl) | ratio floor | reading t4 96 5 | 43.5% unres | 34.7% 8.6% fold *| 0.0% 0.0% 43.5%! 91.7% 48 | 100% 100% | unres fold | FLOOR alone; 100% of drops flip t4 128 1 | 3.3% fold | 26.5% 2.5% fold | 3.3% 2.5% 3.3% 25.0% 8 | 100% 0% | fold fold | no flip t4 160 3 | 14.4% fold | 23.0% 1.5% fold | 7.0% 1.5% 14.4% 14.3% 56 | 96% 4% | fold fold | no flip t4 192 1 | 7.6% unres | 8.8% 6.7% fold *| 7.6% 6.7% 7.6% 25.0% 8 | 25% 25% | fold unres | RATIO alone; 25% of drops flip t4 224 1 | 30.5% unres | 1.1% 7.2% unres | 30.5% 7.2% 30.5% 25.0% 8 | 100% 0% | unres unres | no flip t4 256 2 | 8.2% unres | 3.7% 8.2% unres | 8.2% 1.0% 8.2%!100.0% 28 | 50% 50% | unres unres | no flip t4 288 1 | 24.5% unres | 17.3% 2.7% ntt *| 24.5% 2.7% 24.5%! 25.0% 8 | 25% 25% | unres ntt | FLOOR alone; 25% of drops flip t4 320 3 | 24.6% unres | 15.4% 3.9% ntt *| 6.3% 0.9% 24.6%! 35.7% 56 | 89% 89% | unres ntt | FLOOR alone; 89% of drops flip t4 384 1 | 29.4% unres | 27.0% 9.4% ntt *| 29.4% 9.4% 29.4%! 25.0% 8 | 75% 75% | unres ntt | FLOOR alone; 75% of drops flip t4 512 1 | 27.3% ntt | 64.6% 6.2% ntt | 27.3% 6.2% 27.3% 25.0% 8 | 100% 0% | ntt ntt | no flip t8 96 3 | 10.5% fold | 30.1% 6.2% fold | 6.2% 1.0% 10.5% 64.3% 56 | 100% 0% | fold fold | no flip t8 128 3 | 54.3% unres | 19.3% 4.9% fold *| 26.2% 1.5% 54.3%! 35.7% 56 | 45% 45% | unres fold | FLOOR alone; 45% of drops flip t8 160 7 | 15.9% unres | - - GONE | - - - - - | - - | - - | the filter left no fold leg t8 192 1 | 5.5% unres | 1.1% 4.6% unres | 5.5% 4.6% 5.5% 25.0% 8 | 100% 0% | unres unres | no flip t8 224 2 | 31.9% unres | 4.3% 2.3% ntt *| 31.9% 1.6% 31.9%! 46.4% 28 | 32% 39% | unres unres | needs both; 39% of drops flip t8 256 0 | 6.6% ntt | 9.3% 6.6% ntt | - - - - 1 | - - | - - | no leg dropped t8 288 0 | 8.8% ntt | 16.6% 8.8% ntt | - - - - 1 | - - | - - | no leg dropped t8 320 1 | 6.6% ntt | 22.6% 6.6% ntt | 6.6% 5.5% 6.6% 100.0% 8 | 100% 0% | ntt ntt | no flip t8 384 3 | 11.4% ntt | 43.0% 4.3% ntt | 10.7% 4.3% 11.4% 14.3% 56 | 100% 0% | ntt ntt | no flip t8 512 3 | 5.5% ntt | 69.8% 5.5% ntt | 4.1% 0.9% 5.5% 100.0% 56 | 100% 0% | ntt ntt | no flip cell/m/d the (threads, m) rung and how many of its 8 legs steal < 5.00% removes floor0/v0 the A/A floor and verdict over all 8 legs (the PUBLISHED reading) effS the filtered rung's EFFECT, max(F/T, T/F) - 1, on the floor's own scale; `!` after np95 marks a null of floors that STRADDLES it, which is a rung whose verdict is decided by which legs survived rather than by the arms floorS/vS the floor and verdict after the steal filter; `*` marks a FLIP nmed/np5/np95 the enumerated null of floors under a random d-leg drop pctl where floorS sits in that null (share of members at or below it) n the null's size - C(8, d), the WHOLE family, not a sample P(vS) share of random d-drops that reach the steal filter's verdict P(fl) share of random d-drops that flip the verdict AT ALL ratio/floor counterfactuals: the filtered ratio against the UNFILTERED floor, and the unfiltered ratio against the FILTERED floor - which half of the filter the flip is attributable to floornull --joint: the whole table under the d-matched null (every rung's drop size held at the steal filter's, only WHICH legs randomised) cell m d | n | v0 -> vS | P(a random d-drop flips this rung) t4 96 5 | 48 | unres -> fold *| 1.0000 t4 128 1 | 8 | fold -> fold | 0.0000 t4 160 3 | 56 | fold -> fold | 0.0357 t4 192 1 | 8 | unres -> fold *| 0.2500 t4 224 1 | 8 | unres -> unres | 0.0000 t4 256 2 | 28 | unres -> unres | 0.5000 t4 288 1 | 8 | unres -> ntt *| 0.2500 t4 320 3 | 56 | unres -> ntt *| 0.8929 t4 384 1 | 8 | unres -> ntt *| 0.7500 t4 512 1 | 8 | ntt -> ntt | 0.0000 t8 96 3 | 56 | fold -> fold | 0.0000 t8 128 3 | 56 | unres -> fold *| 0.4464 t8 160 7 | 0 | unres -> GONE *| 0.0000 t8 192 1 | 8 | unres -> unres | 0.0000 t8 224 2 | 28 | unres -> ntt *| 0.3929 t8 256 0 | 1 | ntt -> ntt | 0.0000 t8 288 0 | 1 | ntt -> ntt | 0.0000 t8 320 1 | 8 | ntt -> ntt | 0.0000 t8 384 3 | 56 | ntt -> ntt | 0.0000 t8 512 3 | 56 | ntt -> ntt | 0.0000 the steal filter flips 8 of 20 rungs. under the d-matched null the EXPECTED number of flips is 4.52, sd 1.20, and P(a random same-size drop flips 8 or more) = 0.0050. distribution of the flip count under that null: 1:0.002 2:0.035 3:0.158 4:0.304 5:0.298 6:0.155 7:0.042 8:0.005 READ THIS AGAINST THE PER-RUNG TABLE AND NOT INSTEAD OF IT. The two answer different questions and both answers are real: no INDIVIDUAL flip is established (every one is reproduced by chance 21% to 82% of the time, and three of them sit in 8-member nulls whose smallest attainable P is 0.125), while the SET of them is not chance - which says steal predicts WHICH leg owns the A/A worst, and nothing about whether any one rung's verdict can now be read. AND THIS IS A CALIBRATION, NOT A SIZED HYPOTHESIS TEST. The statistic is computed on the round that raised the question, so 0.0032 prices the coincidence and does not license a decision - the same limit the guest-steal retrospective states about its own nulls.