Residuals, round 9: the Maleceptor, the Red Terror and Hyperadapted Raveners (2026-10-02)

For the tier test (T-523), on round 9 (a5fa1fb9a9): three Tyranids Auspex grades well above us. What-ifs ran in a scratch copy of tools/tiertest.js. A report: no published score changes.

Method. Round 9's command with a probe at its end. A Tyranid-only change re-scores the Tyranids against round 9's pool, so the Marine and Ork ρ can't move ("–"); the banner.js fix (H1) re-ran everything. "Tyr share": Tyranid ρ with list share. "Letters": all-armies letters that change, of 1161. Baseline: Tyranids 0.494, Space Marines −0.099, Orks −0.118, every unit 0.077, Tyr share 0.453. The brief's Maleceptor C (Anvil 78) is round 8's; round 9 has it B (85.2, Anvil 85), C on the proposed cuts.

Summary

All together (H1, M3, R1 + R2, H2): Tyranids 0.573, Space Marines −0.099, Orks −0.115, every unit 0.078, Tyr share 0.514; four Tyranid letters move, plus H1's ten elsewhere.

1. The Maleceptor (180 pts; Auspex S, 8 of 23 lists)

a. The engine check

Banner counts the debuff at the 50% share, but on median stand-ins and against leaders' multipliers (the Prime's 2.70): it can't be its best job.

b. The raw row beside its peers

Damage: enemy points destroyed per 100 pts over the game (best way in) into GEQ / MEQ / TEQ / light / heavy vehicle. Hits to destroy (absolute): small arms / special / plasma / elite / anti-tank / mortal. All four hold 0.80 to 0.90 of the game (position).

Differences. It is by far the hardest (plasma 54 hits against 31 to 35, anti-tank 17 against 8 or 9) and does less per point than the Exocrine or the Haruspex. Our order inside the group matches Auspex's (only the Exocrine and Haruspex swap): the residual is the level. All four sit at Anvil 76 to 85 (B or C); Auspex has them S, A, B, B.

c. What the winning lists do with it

Six of its eight lists take two (14 copies; only the Neurolictor and Genestealers are doubled in more lists). Its lists carry two to six other monsters (a Norn in five, the Tyrannofex in four). Detachments named: Subterranean Assault twice, nothing readable in six. Its lists place at the 49th percentile of their events (all Tyranid lists 47.5th); 2 of 8 in the top quarter (all: 7 of 23).

d. Hypotheses

S needs a raw hold of 410 (it has 246); A needs 288. Only M3b reaches S, by picking its best partners, which is grading it in its best list (M4). Players doubling it argues against mere popularity, but its lists don't place better than other Tyranid lists.

e. Verdict

Missing raw quantity. The aura's main effect, making the squads beside it harder to shift, is hold, booked as a lend on two median stand-ins. Recommend M3 (Banner's own partner rule): A (88.0), Tyranid ρ 0.501, no other letter. To be accepted: run on every army's defensive auras (unitfx, def side), the Marine and Ork ρ must not fall below −0.10 and −0.12, the Tyranid ρ must stay at or above 0.494, and no army's S count may rise by more than one.

2. The Red Terror (130 pts; Auspex A, 10 of 23 lists)

a. The engine check

The re-arrival, measured. As written, re-arriving loses to a plain Deep Strike (Hammer raw 81 against 103): the model sends it back after every unengaged fight, a failed 9" charge included, so it charges from 9" (28%) every round. Letting it stay when staying charges better (R1) makes re-arrival its best way in: 174 in Subterranean Assault (7" tunnels), Hammer 45 → 59.

b. The raw row beside its peers

Differences. Against the burrowing monsters it has half their hits to destroy under plasma and anti-tank (16 and 6 against 32 and 8), lives 2.5 typical turns against 4.9 and delivers 40% less per point. Our order inverts Auspex's: Mawloc and Trygon (Anvil, via Crusher Stampede's OC) above the Red Terror and Raveners; Auspex has Raveners S, the Red Terror A, the monsters B. We pay for bulk that holds; the panel for killers that pick their fight.

c. What the winning lists do with it

One a list (Epic Hero), in 10 of 23; with Biovores (9), Raveners, the Neurotyrant and the Neurolictor (8 each), Hyperadapted Raveners (7): the deep-strike package. Detachments named: Subterranean Assault, Vanguard Onslaught and Ambush Predators, 3 each. Its lists place at the 52nd percentile (Tyranid average 47.5th), 2 of 10 in the top quarter.

d. Hypotheses

A needs a Hammer raw of 360 (103 now, 174 with R1) or Runner 88, which the points gate caps near 70 at 130 pts. Precision and the heal don't close a two- to three-fold gap. R3 has some support (its lists place below average, fewer top-quarter finishes), but ten lists can't separate that from chance.

e. Verdict

Engine simplification plus a missing raw quantity; past C, unmeasured or popular. R1 makes the rule help instead of hurt; R2f gives Runner the survival only the Fight-phase version of the rule earns. D → C (75.5), Tyranid ρ flat (0.493), Tyr share 0.459. Recommend R1 + R2f; to be accepted: the Tyranid ρ within 0.01 of 0.494, and no unit whose re-deploy comes after enemy shooting (Raveners; other armies' end-of-turn rules) lifted by it. R2 for Raveners (ρ 0.526) claims survival their rule doesn't give.

3. Hyperadapted Raveners (165 pts; Auspex S, 11 of 23 lists)

a. The engine check

The bug. tools/banner.js adds a character's own damage and durability as one model (dmg(rec, 1, …), htk(rec, 1, …)) while charging all its points. Four of these five models go uncounted, and bodies win (0.37). At its minimum size: 0.68.

b. The raw row beside its peers

Differences. Its damage per point is the group's best into vehicles, but it lives 2.5 typical turns at five bodies. Its Banner is a seventh of the Prime's for the same rule (Sustained Hits 1): the Prime lends to a gaunt squad far bigger than itself; the Hyperadapted Raveners are compared with 6.6 more Raveners and (bug) count as one body. Auspex all but inverts our order: its C is our S, and two of its three S units are our two lowest.

c. What the winning lists do with it

Joined to Raveners in 9 of its 15 copies (8 of 11 lists); four lists take two; always five models. Detachments named: Vanguard Onslaught 5, Ambush Predators 5, Subterranean Assault 2. Its lists place at the 51st percentile, 3 of 11 in the top quarter.

d. Hypotheses

Why H2 moves so much. Joined, hits to destroy double and turns alive go 2.5 → 4.6, so it fights through turns 3 and 4: Hammer raw 204 → 307, though damage per turn per point rises only about 15%. That is critique #1 (scored alone at minimum size), and partly a property of absolute survival: a fixed typical enemy turn shifts a bigger unit more slowly, where a real opponent would aim more at it. H2 is right in kind; its size is an upper bound. H4 is the honest alternative: if the panel grades the pair, H2's shared letter is what it means.

e. Verdict

Engine bug plus missing raw quantity. H1 is neutral on every yardstick (Tyranids 0.496, Orks −0.115, every unit 0.075) and should go in on correctness alone. Then recommend H2's rule: a Leader that is itself a multi-model squad (or any pair joined in most lists taking both) is scored as the joined unit, both datasheets showing the pair's letter "with <partner>". Both units D → B, Tyranid ρ 0.560, Tyr share 0.503. To be accepted: the same rule on every army's most-joined pairs (the Marine panel first) must not lower the Marine or Ork ρ; the Tyranid ρ must stay at or above 0.53; and the lift must survive a fire budget that scales with the target's points (if it vanishes there, it was the size artifact).

4. Engine bugs and corrections

1. tools/banner.js counts a character as one model in its join (3a): 26 characters in 12 armies. Fix: its minimum size in both calls.
2. tiertest.js goRe forces a fight-phase re-deployer back after a failed charge (2a). Fix: stay when staying charges better, as the "enemy" branch already decides.
3. The Red Terror's heal is modelled (×1.06); the brief's "no healing" is out of date.

5. Every what-if

Letter agreement barely moves (raw 18 to 19, within one 31 to 32). Every Tyranid gain is inside one SE (0.13) and on the panel we tune on: diagnoses, not evidence. The clean panels can only judge M3, R2f and H2 once each runs for every army; that run is each one's acceptance test.