The detachment audit (T-690, 5 Oct)

Jordan asked: "are the detachments all loaded and their math implemented correctly?" This report answers for the 16 armies with reviewer panels. It is a report only: no rebuild, and no ledger number moves. The tool is tools/detaudit.js; its numbers are in 2026-10-05-detachment-audit.json. It read the ledger's frozen inputs (snapshot eb3590a, BSData 374f505, hash 30f0056e6103), the built ledgers (docs/lab/ledger/ for Tyranids and Space Marines, build/ledger/ for the rest, all from before T-687), and Track A (step 34b) through the ledger page's compute(). Rules were read on Wahapedia's 11th-edition pages to understand them; nothing here is their wording.

The short answer

• Loaded: yes for 15 of the 16 armies. Every matched-play detachment is built, with no other chapter's detachments and no Boarding Actions ones. The exception is the Drukhari, built with no detachment at all (already known; T-687). The Genestealer Cults are not missing anything: Infestation Swarm is a Boarding Actions detachment, so it was right to leave it out. T-687 should not add it.
• The math is applied as our data says. In 40 hand checks, the tool worked out each weapon's modifiers on its own from the effects data, rebuilt the form and ran the ledger's own kernel. It matched the ledger on every weapon, and the stored Kill came back to the decimal every time.
• The data and the ledger's design still get the choice of detachment wrong in many armies. There are two main causes:
  1. Some "choices" in the data are really dice rolls, ladders or once-a-battle picks. The ledger takes the best option as if it lasted all game. This gives Emperor's Children's Coterie of the Conceited 23 of 23 units at +75%, Space Wolves' Saga of the Great Wolf 19 of 21, Ordo Xenos +30% for every Deathwatch unit, and Madcap Meks +41% for the Big Mek.
  2. Most detachment dice are situational, and the ledger counts a situational condition as never met. Of the 114 rules with dice effects, 82 add nothing. Meanwhile any non-dice rule of a listed kind (heal, aura, mark, Battle-shock and so on) gives every unit in that detachment the fitted ×1.065. So in seven armies the "best detachment" is simply the one whose non-dice rule happens to carry that multiplier. In Death Guard, four detachments tie on every unit.
• Enhancements and stratagems are never applied. The ledger reads only a detachment's rules. 212 of the 16 armies' 658 enhancements have dice effects in our data, but none reach a form. Stratagems are names only (950).

All armies

How to read the columns:

• Built / matched-play. The detachments the ledger's forms carry, against the matched-play list in docs/data/detachments.json. A detachment counts as matched play when it has a Detachment Points cost and no Boarding Patrol rule, and when its only field (the chapter or army it is reserved for) allows this army.
  • A chapter takes its own detachments plus the 16 Codex ones.
  • For the Drukhari, the list is the nine detachments reserved for them. The 11th-edition Drukhari page lists only those nine; the other 15 under the same key are the Craftworlds'.
  • The pre-dataslate detachments.json (git show bd8cdfccc1^) gives the same lists.
  • The ledger itself never reads detachments.json. It works its list out from the frozen detfx.json (tools/ledger.js matchedOf).
• Rules with dice. Detachment rules our data marks modelled or partly.
• Rules the ledger applies. Rules with at least one effect the ledger uses. That leaves out effects whose condition is vouched (when.manual, counted as never met), effects switched off, and chosen-target or aura effects.
• Hand checks. "Math" means the ledger's modifiers and Kill were reproduced. "Rule read right" means our data matches the rule for that unit.
• ⚑ marks an army where the winning detachment wins implausibly often, or only on a tie or a flat multiplier.

Frozen against live inputs. Of the built detachments, only Emperor's Children's Frenzied Host differs between the frozen detfx.json and the live one. That is the 30 Sep update's Frantic Focus change. Its effect is vouched either way, so it changes no number.

What the ledger does with a detachment

Each form takes one detachment (or one way of answering its choices), and the ledger applies the pieces like this:

• Dice effects from the detachment's rules reach the weapons through docs/detfx.js effects(), which includes kindfx.json's numbers.
• The conditions are handled one by one:
  • Unit keywords, melee or ranged, a weapon name and Rapid Fire are settled per weapon.
  • The target's keywords and Strength against Toughness are settled per target.
  • A vouched condition counts as never met. This is the "no shares" rule; the site's Best matchup now takes such conditions at half, but the ledger doesn't.
• A choice becomes one variant per option, and each form keeps whichever variant scores best for that unit.
• A noted rule of a listed kind goes on the form's rule list:
  • On a Value line it takes Track A's detachment multiplier, ×1.065, applied to every unit in the detachment unless the rule's own effects name keywords the unit lacks.
  • Arrival and movement rules go to the charge model.
• Enhancements, stratagems and the Detachment Points budget (up to three detachments an army) are not used.

Mismatches, army by army

Only armies with a mismatch are listed. Each line gives the mismatch, its cause and its size in the ledger's numbers.

Drukhari

• Nothing loads (cause: loading). detfx.json files the nine Drukhari detachments under "Xenos - Aeldari", and matchedOf reads only the army's own key, so all 37 units sit on "No detachment". For T-687:
  • Take the nine detachments.json marks only: Drukhari, not all 24 under the Aeldari key. The other 15 are Craftworld detachments, illegal for the Drukhari.
  • None of the nine has a Boarding Patrol rule, so the Boarding filter won't catch strays.
  • Once loaded, only Reaper's Wager has an unconditional dice effect (re-roll hit 1s for every unit). Expect it to win most units, the way Ironstorm Spearhead does for the Marines.
  • Exhibition of Slaughter's and Kabalite Agonysts' "target not MONSTER/VEHICLE" conditions will count as never met (see Astra Militarum).

Genestealer Cults

• Infestation Swarm is not missing (cause: loading). It is a Boarding Actions detachment: Mustering a Boarding Patrol, Rules Adaptions, no Detachment Points. It sits in the Boarding Actions section of the 11th-edition page, and our data files it under the Tyranids. The Cults' nine matched-play detachments are all built.
• Xenocult Masses wins 77% of units, 52 of them alone, on a ×1.065 it shouldn't give them (cause: wrong units).
  • Its rule, Hordes of the Faithful, heals Neophyte Hybrids in terrain and does nothing else.
  • The Lictor and the Winged Hive Tyrant take it as their best detachment, tied with Final Day (whose noted aura carries the same ×1.065). The Attilan Rough Riders take it alone.
  • Each margin over the next detachment is exactly +6.5%: the multiplier and nothing else.

Emperor's Children

• Coterie of the Conceited is everyone's best, at +75% (cause: always-on should be conditional).
  • Pledges to the Dark Prince is a ladder: Pact points earned over the battle unlock re-roll hit 1s, then re-roll wound 1s, then Lethal and Sustained in melee, then crits on 5+, cumulatively.
  • The data offers each step as a choice, and the ledger keeps the 7+ step all game.
  • Daemonettes ×10: melee Kill 143 → 262 (+83%). Best detachment for 23 of 23 units; median margin +74.5% Value over the runner-up.
  • A share of the game per step (the 7+ step only late, if at all) would fit the rule.

Space Wolves

• Saga of the Great Wolf is best for 19 of 21 units (cause: always-on should be conditional).
  • Master of Wolves lets the army pick one Hunting Pack a round, each only once a battle (Logan Grimnar allows one repeat).
  • The data says "pick one each turn", and the ledger keeps the best pack all game: +1 to hit shooting, or Lethal or Sustained in melee.
  • Bjorn: melee Kill 75 → 89 (+18%), median margin +5.4%. The right share is about one round in five.

Agents of the Imperium

• Ordo Xenos Alien Hunters gives every Deathwatch unit +27% to +34% Value (cause: always-on should be conditional).
  • Mission Tactics: one tactic a round, each once a battle.
  • The ledger keeps Furor (Sustained 1) or Malleus (Lethal) on all game.
  • Affected: Watch Master, Deathwatch Kill Team, Watch Captain Artemis, Aquila Kill Team, Corvus Blackstar.
  • Ordo Malleus (re-roll hit 1s, and wound 1s against daemons) checks out exactly on Coteaz.

Orks

• Madcap Meks (cause: always-on should be conditional).
  • Unpredictable Genius is a D6 rolled each time: 4 gives +1 to hit, 5 takes away cover, 6 gives −1 T, and a 1 hands the target a 4+ invulnerable save.
  • The ledger takes the best roll every time.
  • Big Mek in Mega Armour: ranged Kill 112 → 144 (+28%), Value +41% over War Horde. Mek Gunz: +37%.
• Dread Mob (cause: always-on should be conditional).
  • Try Dat Button! is a roll: one D6 gives one of +1 A, +2 S or +1 AP; two D6 give two different results and a hazard roll.
  • The ledger takes the best pair every time and skips the hazard.
  • Deff Dread: melee Kill 89 → 94 (+5.5%). 6 units alone.
• War Horde (Sustained Hits 1 in melee for every Ork) is right, and its 66% is earned.

World Eaters

• Vessels of Wrath (cause: wrong units). Wrath of Khorne excludes Epic Heroes, but our data scopes it to CHARACTER only, so Khârn and Angron get +1 AP in melee. Khârn: melee Kill 113 → 120 (+5.6%); it is his and Angron's best detachment.
• Brazen Engines (cause: wrong units). Rampaging Terrors gives only Daemon Vehicles a Battle-shock test, yet the ×1.065 reaches every unit.
• Cult of Blood (cause: wrong units, a code bug). tools/ledger.js detRulesOf checks a rule's own effects for unit keywords but not its choice options' effects. So Idols of Khorne, whose options name only Jakhals and Goremongers, counts as every unit's rule. Brazen Engines and Cult of Blood tie for best on 21 of 30 units.
• Khorne Daemonkin (cause: always-on should be conditional; Soak only). Blood Tithe's abilities are bought with tithe points during the battle, but the ledger keeps Boon of Blood (4+ invulnerable for Blood Legions) on all game. 1 unit alone.

Astra Militarum

• Combined Arms loses half its rule (cause: conditional should be always-on).
  • Born Soldiers gives REGIMENT units Lethal Hits on ranged attacks against anything but monsters and vehicles.
  • Our data writes that as a vouched condition ("target not MONSTER/VEHICLE"), because the effects vocabulary has no "not these keywords" condition. The ledger therefore never counts it.
  • Only the SQUADRON half (Lethal against monsters and vehicles) is live, so Combined Arms wins just 12 vehicle units.
• Designation Force and Siege Regiment tie on 60 of 72 units (cause: wrong units). Neither has a dice effect the ledger applies; each carries one noted rule (a mark, an enemy debuff), so each gives every unit ×1.065. Designated Targets only adds 3" of detection range to one enemy unit.

Dark Angels

• Ironstorm Spearhead wins 80% (cause: structural; the reading is right). Armoured Wrath really does reach every Adeptus Astartes unit: one re-roll of a hit, wound or damage roll a phase. It is the only unconditional effect in the 24 detachments apart from Librarius Conclave's, for psykers. Every other Codex and Dark Angels detachment's dice are vouched and add nothing.
• Dark Age Arsenal (cause: conditional should be always-on). +1 S for plasma weapons is written as vouched ("Plasma weapons only"), although when.weapon: ["plasma"] exists for exactly this.
• Wrath of the Rock (Soak only). "−1 to wound when the attack's S is above the unit's T" is vouched, though sVsT: "gt" exists.

Space Marines, Blood Angels

• Space Marines: Ironstorm Spearhead wins 65% (cause: structural), as above.
• Blood Angels: Rage-Cursed Onslaught (re-roll wound 1s in melee for everyone; +1 A after charging, vouched) checks out. One oddity outside the detachment data:
  • Chief Librarian Mephiston's Value is 4× in Rage-Cursed Onslaught (2,051 against 509 elsewhere) from +12% melee Kill.
  • The form's burst, prefer and life inputs switch from ranged to melee and double.
  • With burstKill off, Rage-Cursed is 19% below Librarius Conclave for him.
  • The same switch decides one other big case (an Emperor's Children Sorcerer). It is rare (9 units in all), but worth a look in tools/burst.js.

Adeptus Mechanicus

• Rad-zone Corps wins 78% (cause: wrong units).
  • Rad-bombardment is a pre-battle bombardment of the enemy's deployment zone. It acts once for the army.
  • The ledger gives every unit's Kill line ×1.065 for it. The Onager's form has the same dice as Cohort Acquisitus; the ×1.065 is the whole margin.
• Lords of the Forge (doubt; Soak only). War-form Mantles gives TECH-PRIEST models a 4+ invulnerable save and Feel No Pain 5+. The Servitor Battleclade carries TECH-PRIEST at unit level (its Underseer), so all nine models get it.

Death Guard, Thousand Sons, Leagues of Votann

• No detachment rule of theirs reaches the dice (cause: structural). Death Guard's nine are all noted; Thousand Sons' and Votann's dice rules are all vouched. Their winners are set by the ×1.065 alone (cause: wrong units):
  • Death Guard: Mortarion's Hammer, Paragons of Putrescence, Shamblerot Vectorium and Tallyband Summoners tie on every unit.
  • Thousand Sons: Ritual of Regeneration (heals only non-monster Thousand Sons psykers) is the Maulerfiend's and Defiler's best.
  • Leagues of Votann: Persecution Prospect and Brandfast Oathband tie on 17 of 22 units.

Tyranids

• No mismatch.
  • Subterranean Assault: re-roll hit 1s for every Tyranid; checked on the Lictor and the Biovores.
  • Invasion Fleet's Hyper Adaptions: checked on the Biovores.
  • Subterranean Assault's 58% is earned. Its margin over Invasion Fleet is a few per cent.
• One caveat, shared with every army: a choice made once for the army (Hyper Adaptions, Psychic Disciplines, Wrath of Khorne) is answered per form. Two units in one detachment can each take the option best for them.

For T-687 (and the ledger's owner)

1. Drukhari: load the nine only: Drukhari detachments from the Aeldari key, not all 24 (above).
2. Genestealer Cults: don't add Infestation Swarm; it's Boarding Actions.
3. detRulesOf scoping (tools/ledger.js): check choice options' effects too. Idols of Khorne currently reaches every World Eaters unit.
4. Data fixes that would move numbers (effects files, a deliberate re-freeze):
   • Choices that are rolls, ladders or once-a-battle need a share or a mean, not a free pick: Pledges to the Dark Prince, Master of Wolves, Mission Tactics, Unpredictable Genius, Try Dat Button!, Blood Tithe.
   • Wrath of Khorne should exclude Epic Heroes.
   • Born Soldiers' Regiment half and the Drukhari's two "not MONSTER/VEHICLE" rules need a "not these target keywords" condition.
   • Dark Age Arsenal needs when.weapon.
   • Wrath of the Rock, Covenite Coterie and Tools of Torment need sVsT.
   • Blood Tithe's Feel No Pain should be fnpPsychic/fnpMortal.
5. Design questions, not bugs (for the supervisor):
   • The ×1.065 on every unit for a noted rule that names a few units or acts once for the army.
   • Vouched conditions at zero, which makes the few unconditional rules decide every army's best detachment.
   • No enhancements or stratagems.

Doubts

• The rule texts read were the current 11th-edition pages, after the 30 Sep update. The ledger is pinned before it. None of the rules checked here is on the update's list as far as I can tell, but I didn't diff them.
• How much the fixes would move the exam is not measured. A report only. The Emperor's Children and Space Wolves fixes would change their best detachments outright. The ×1.065 question would turn seven armies' best detachments into ties or into whichever detachment has real dice.
• "Wins" counts each unit's ranking row (its best solo form at Track A). A tie shares the win. "Alone" means the detachment's best form beats every other detachment's best form for that unit.
• The hand checks are 40 (2 or 3 per army), chosen by the tool: the most-winning detachment on a unit it wins and one it doesn't, then the detachment with the most live dice. The rule judgements are mine.