Residuals, three Tyranids: Neurogaunts, the Lash Whip Prime, Biovores (2026-10-02)

An investigation on the round-9 run (T-562, a5fa1fb9a9). A report: no score, page or tool changes. The what-ifs ran as switches in a scratch copy of tools/tiertest.js, each against round 9 in the same process; the control reproduces round 9 exactly (ρ Tyranids 0.49, Space Marines −0.10, Orks −0.12; Tyranid list share 0.45, every unit 0.077). Rules were read on Wahapedia's 11th-edition pages and are summarised in our own words. Lists: grimstat-corpus (CC BY 4.0), the 23 Tyranid lists in the window.

0. Summary

ρ triples are Tyranids / Space Marines / Orks. The dice engine has no bug. There are two modelling faults (§1e, §2e) and one data note: build/units.json lists Neurogaunts' larger size as 21 models for 90 pts, where the datasheet has 22 (two Nodebeasts). Neurogaunts have a 6+ save, not 5+.

1. Neurogaunts (Auspex B; ours A)

a. Engine check

Is the "typical fire" too soft? A tool "turn" is half the median unit's hits to destroy per class: 36 small-arms hits, 20 special, 6.5 plasma, 6 elite, 3.6 anti-tank, 2.5 mortal. Neurogaunts need 19.8 / 16.5 / 13.2 / 13.2 / 13.2 / 11 hits, so they last 0.55 / 0.81 / 2.04 / 2.20 / 3.67 / 4.40 turns against each class. Tu is the meta-weighted arithmetic mean (8 / 20 / 37 / 7 / 22 / 6%), each class floored at 1: 2.27. Plasma and anti-tank (59% of the meta's fire, overkill on one-wound models) carry it.

A real shooting phase, measured with ev.js using every winning-list unit's default ranged volley into eleven Neurogaunts with no cover: an average list kills 33.5 of them from 9.0 shooting units, which is 3.7 per shooting unit (median 3.2). The tool's turn removes 4.8, the attention of about 1.3 average shooters. Two turns is right if one enemy unit looks at it each turn, and wrong if three do. The size of a turn matters little (doubling it, N2, gives 88.2). What matters is the mix: one enemy turn fires every class at once, so the share destroyed per turn adds across classes and turns alive is the harmonic mean, 1 ÷ Σ(wᵢ ÷ turnsᵢ). That gives 1.49 (gate 0.8), against 1.51 for Termagants and Hormagaunts and still about 4 for the Exocrine.

b. Raw rows

Damage: wounds per turn per 100 pts into GEQ / MEQ / TEQ / light / heavy. Hits to destroy: small / special / plasma / elite / AT / mortal. Position: share of the game on the two objectives.

The biggest differences: the peers carry OC 2 a model and Neurogaunts OC 1, so Termagants hold more per point (0.33 OC a point against 0.24), and Neurogaunts have no second job (Hammer 24, Runner 50). What keeps Neurogaunts at A is that every horde gets the same turns alive, carried by overkill.

c. Lists

Two lists of 23, each with one unit of eleven, beside the Red Terror (both). No detachment is named. Neither list has a Neurotyrant (the leader banner.js pairs them with).

d. Hypotheses (letters moved: of 1,161, Tyranids in brackets)

The "consensus wrong" case is weak: the lists (2 of 23) agree with Auspex, and the relay, the one thing we leave out that could lift Neurogaunts, measures at almost nothing. A Tyranid player who wants cheap OC takes Termagants or Hormagaunts, which also hit.

e. Verdict

Wrong measure. Tu treats each enemy turn as a single weapon class and floors every class at one turn, so the classes that actually kill gaunts count least. Change: use the harmonic mean for Tu, in the gate and the Battle-shock share only. Hammer and Runner keep T: switching T as well (N4b) costs the Marine panel 0.18. To be accepted: Neurogaunts under 88 with Hormagaunts still at A, the Tyranid ρ at least 0.49, and both clean panels within 0.05 of round 9 (N4 meets all four). On the proposed cuts Neurogaunts are already B.

2. Tyranid Prime with Lash Whip (Auspex C; ours S)

a. Engine check

Where Banner 97 comes from (banner.js's own functions, the leader's attacks switched on and off):

95% of the added damage is its own fists, counted at full rate every turn with no delivery and no survival, and on Termagants, the squad whose extra bodies hit least. In Hammer the same fists score 22, because alone it is a solo character (Tu 0.55) walking to melee. One quantity is counted twice, by two rules. The Winged Prime (0.28 lent of 2.56) and the Broodlord (0.92 of 4.87) have the same fault.

b. Raw rows

The biggest difference: the Prime has the best own melee per point of the four and the weakest lend. Round 9 ranks these leaders by their own damage per point, and the panel ranks them by what they lend.

c. Lists

13 copies in 11 lists (two lists take two). They join Hormagaunts 5 times, melee Warriors twice and Termagants once; 5 copies aren't joined in the list text. Detachments: Vanguard Onslaught 3, Ambush Predators 3, Subterranean Assault 2. The Prime is most often listed beside Biovores (11), Hormagaunts (10) and the Swarmlord (9).

d. Hypotheses

P3 was applied to the Prime alone. Applied to every leader, it would hand back the own-attack worth that P1 takes from the characters falling to D, and losing that worth is why P1 cuts every unit's list share to 0.046.

e. Verdict

A fault in how Banner is measured, and both sides are partly right. The lists are right that the Prime earns its place, but it earns it with its own melee while the squad shields it. The reviewer is right that its buff is weak (0.53 of the same points in bodies). Change: Banner counts only what a leader lends (its rules, plus the wounds it puts behind the squad). Its own attacks go to Hammer as part of the joined unit, delivered at the squad's turns alive (extend the Dedicated Transport bundle in tiertest.js to Leader links). To be accepted: the Prime lands at B or C, the Tyranid ρ stays at 0.52 or above with neither clean panel lower (P1 + P3: B 83.4, 0.54 / −0.09 / −0.11), and once every leader has the joined-unit Hammer, every unit's list share gets back to 0.077 or better. P1 must not ship alone.

3. Biovores (Auspex S; ours A)

a. Engine check

A Spore Mines unit-turn in points (our derivation). Screening: the army's own OC-0 screen costs 30 pts for a five-round game (a Ripper Swarm or a Mucolid Spore), 6 pts a unit-round, and a mine lives one round under our fire (Tu 1.01). The burst: 1.0 mortal wound expected × 11.4 pts a wound (the five targets' points per wound at the meta's target weights) × an even chance it reaches anything = 5.7 pts. A spawned mine ≈ 11.6 pts; four spawns ≈ 46 pts a game, 0.77 per Biovore point.

b. Raw rows

The biggest differences: Biovores have the weakest gun per point of the four and the same fragile profile as Pyrovores, yet they are the one taken. The spawn is the whole difference. The panel grades all four S or A, and the lists take only the Biovore.

c. Lists

All 22 lists take exactly one unit of one Biovore, never more models. They do so in all seven detachments named (Vanguard Onslaught 5, Ambush Predators 5, Subterranean Assault 4, Assimilation Swarm 3, Talons 3, Synaptic Nexus 1, Warrior Bioform Onslaught 1), most often beside the Neurolictor (16), the Neurotyrant (16) and Zoanthropes (15). One model means a one-mine unit a turn: the lists buy an extra unit every turn, not the gun.

d. Hypotheses

The popularity reading doesn't hold. When 96% of lists pick one model and they are free to leave it out, that is a strong signal of its worth.

e. Verdict

A missing quantity, already half-modelled; the rest is the ruler. The spawn proxy already carries 38 points of the score. Our own price for a mine (11.6 pts) is too small to matter through Banner (B4); it would take about 30 pts a mine (B5). Change: V5C.SPAWN 0.5 → 0.75, because the Biovore makes a new unit every round and the lists value exactly that (always one model). It moves Biovores alone: Tyranid ρ 0.49 → 0.51, Tyranid list share 0.45 → 0.47, and no other letter of 1,161. To be accepted: Biovores reach S at the current cuts (97.5) with nothing else moving. On the proposed cuts (S ≥ 98.79) only a share of 1 gets there (100).

4. What it says

The two over-ratings are measurement faults, not missing rules. One quantity is counted twice by two rules (a leader's own damage, undelivered in Banner and delivered in Hammer), and one is combined the wrong way (per-class survival averaged as if the enemy fired one class a turn). The under-rating is the one real missing quantity, and its weight (0.5) is a builder's guess that the lists say is low. N4, P2 and B6 each move 44 letters or fewer. The Banner fix (P1) is the only change that lifts all three panels together, and it needs the joined-unit Hammer before it can ship.