# Kill-rule scan 11 Oct 2026 (#1871), auditor
Rule: kill at n>=20 if mean<1.0 and upper 80% bound<1.05; fast-track at n>=40, mean>=1.3, lower 80% bound>=1.0. Mean +-1.28 se, all trades in each test's CSV (replay rows included where the CSV mixes them). Script: /opt/crew/work/auditor/killrule.py

```
grid A  190 2.53 1.96 3.11 FASTTRACK
busy60 A MAIN 7 1.66 0.91 2.42 
busy60 B  27 0.43 0.17 0.69 KILL
busy60 Z  83 0.84 0.65 1.03 KILL
busy60 P  22 0.93 0.55 1.3 
busy60 R  16 1.21 0.52 1.91 
busy60 Q  16 1.04 0.43 1.65 
busy60 N2  28 1.25 0.89 1.6 
busy60 N3  14 0.42 0.05 0.78 
v3005 tp5 MAIN 5 0.83 -0.23 1.9 
v3005 tp3  5 1.06 0.23 1.89 
a1mc BASE  33 1.17 0.53 1.82 
a1mc LATEB MAIN 33 1.24 0.19 2.28 
a1mc ENS6  33 1.51 0.22 2.79 
a1mc NPRE10  27 1.33 0.54 2.11 
a1mc NOREPEAT  28 1.35 0.59 2.11 
a1mc TOPB11  1 0.97 0.97 0.97 
fast5 A MAIN 2 1.28 -0.36 2.92 
fast5 B  2 2.01 -0.57 4.59 
fast5 C  2 1.28 -0.36 2.92 
fast5 D  1 0.0 0.0 0.0 
fast5 E  1 0.0 0.0 0.0 
fast5 Z  1 0.0 0.0 0.0 
fast5 Z5  2 2.01 -0.57 4.59 
fast5 U  2 1.09 -0.31 2.49 
fast5 NA  30 1.24 0.95 1.52 
fast5 NS  13 1.04 0.68 1.39 
fast5 NS5  11 0.75 0.25 1.25 
fast5 T120  2 0.59 -0.17 1.35 
fast5 T300  2 1.09 -0.31 2.49 
fast5 NT120  32 1.01 0.65 1.37 
fast5 AF  2 1.28 -0.36 2.92 
fast5 AT2  2 1.28 -0.36 2.92 
seed S MAIN 29 1.16 0.78 1.55 
seed L  29 1.16 0.78 1.55 
seed B  64 0.99 0.76 1.21 
v3late1m tp2 MAIN 15 1.02 0.78 1.26 
v3late1m tp3  15 1.4 1.0 1.8 
v3late1m tp5  15 1.58 0.94 2.23 
v3late1m tp2u  17 1.04 0.81 1.27 
late315 A MAIN 10 0.28 -0.08 0.64 
late315 B  136 0.13 0.07 0.2 KILL
late315 C  146 0.14 0.08 0.21 KILL
fl2m A  71 2.06 1.72 2.39 FASTTRACK
dip C  6 0.6 0.49 0.71 
dip A MAIN 3 0.47 0.47 0.47 
dip B  3 0.47 0.47 0.47 
lv A MAIN 38 1.83 1.15 2.52 
lv B  38 2.95 1.7 4.2 
lv C  41 1.73 1.1 2.37 FASTTRACK
r300 A  15 3.25 1.77 4.73 
r300 B MAIN 15 2.48 1.69 3.27 
r300 C  15 2.02 1.29 2.76 
r300 D  15 2.2 0.98 3.43 
wh30 A MAIN 3 2.14 1.27 3.01 
wh30 B  3 2.19 1.24 3.15 
wh30 C  3 2.19 1.34 3.03 
wh30 D  3 3.12 1.12 5.11 
flow60 tp5 MAIN 59 0.91 0.65 1.16 
flow60 tp10  59 1.02 0.68 1.36 
flow60 ladder  59 0.96 0.67 1.25 
nh100 A MAIN 18 1.11 0.97 1.25 
nh100 B  18 1.09 0.97 1.2 
nh100 C  18 1.09 0.97 1.2 
tinyv3 tp5 MAIN 19 1.02 0.51 1.54 
tinyv3 tp3  19 0.87 0.53 1.21 
tinyv3 tp10  19 1.23 0.57 1.9 
wlink A  151 2.06 1.88 2.25 FASTTRACK
breadth A MAIN 45 0.99 0.71 1.26 
breadth B  28 1.18 0.78 1.58 
breadth C  70 0.9 0.7 1.09 
lpown A MAIN 9 0.65 -0.03 1.32 
lpown B  8 0.71 -0.05 1.47 
lpown X  9 0.0 0.0 0.0 
lpown Z  63 0.84 0.54 1.13 
lpown N  266 0.98 0.83 1.13 
wlink1 A MAIN 112 1.73 1.51 1.94 FASTTRACK
top1 A MAIN 36 1.11 0.49 1.73 
v3ride ALL  26 1.31 0.83 1.79 
v3ride AGE60  20 1.59 0.99 2.19 
v3m1 A3  45 0.78 0.63 0.94 KILL
v3m1 B  45 0.94 0.66 1.21 
v3m1 C5 MAIN 45 1.27 0.91 1.62 
v3m1 C10  45 1.81 1.16 2.45 FASTTRACK
caps A MAIN 43 1.11 0.8 1.42 
caps B  55 0.99 0.74 1.24 
caps C  68 0.91 0.71 1.11 
momdd A MAIN 11 1.93 1.1 2.77 
momdd B  11 3.15 0.7 5.61 
seedhi TP5 MAIN 8 1.52 0.63 2.41 
seedhi TP10  8 2.27 0.7 3.85 
wide A MAIN 69 1.38 0.93 1.84 
wide B  69 1.88 1.15 2.62 FASTTRACK
wide C  69 0.97 0.74 1.19 
```

## Main-arm verdicts (what the rule says, after splitting replay vs forward)
KILL by rule (main arm): none. Non-main arms hitting KILL (control/ref arms, no action beyond dropping from page): busy60 B, Z; late315 B, C (by design must lose); v3m1 A3.
Closest to kill on main arm: flow60 tp5 (n59 0.91, hi 1.16; forward n15 0.46, hi 0.64: forward-only already below, its own kill says retire at n30 -> retire then); late315 A (n10, 0.28; reaches n20 soon, will KILL); dip (n3 main, all arms <1x); breadth A (n45 0.99, hi 1.26: not yet); tinyv3 tp5 (n19 1.02: at n20 check).
FAST-TRACK by rule on all rows, but replay-contaminated -> NOT fast-tracked until forward-only passes:
- grid (replay by design, already VERIFIED), wlink A n151 2.06 (CSV mixes replay; forward A n29 2.36 per testmgr2, needs n40 forward), wlink1 A n112 1.73 (forward-only n32 1.03, CI 0.75-1.31 -> fails), fl2m n71 2.06 (replay n71; forward ~1/day), lv A n38 (forward n6), wide B (forward n13 1.30, CI 0.04-2.56), v3m1 C10 n45 1.81 lo 1.16 (non-main arm; main C5 1.27 lo 0.91; real forward test, same entry): **candidate for audit** (sellability at 10x unproven).
Rule reading: fast-track needs forward rows only; otherwise a replay-fitted filter fast-tracks itself.
Recommendation to manager: wlink1 is a good example, all-row 1.73 vs forward 1.03. Make the scan forward-only in scoreboard.py (split on entry_t >= `since`).