A read-only check. No ledger number, page default, data file, benchmark or frozen input changes. Our own numbers from the datasheets' dice; unit names only, no rules text.
The short answer
- On Tyranids, the idea doesn't hold, as we pre-registered it. Ranking the 52 Tyranid units by how many points they can finish in one activation (Burst) instead of how many they remove on average (the ledger's Kill) does not move the units the reviewers rate above the ledger up, or the over-rated ones down: the correlation is 0.08 (chance alone gives that or more 29% of the time). Only 4 of the 10 named units move the predicted way (Exocrine, Raveners up; Toxicrene, Tyrannocyte down). Tyrannofex and Maleceptor go down by Burst, Haruspex and Genestealers don't move, Tervigon and Hierophant go up.
- Why: on Tyranids Burst and Kill mostly agree (rank correlation 0.87): a unit that removes a lot on average usually finishes a lot too. What Burst adds, the all-or-nothing of a whole unit, mostly marks down the units that chip at many targets and finish few (Ripper Swarms, Termagants, Carnifexes' guns, Hive Guard), not the units the reviewers under-rate. The named burst units are big guns with few shots or melee units already near the top of Kill.
- On Space Marines (not pre-registered, a transfer check) it does show up: correlation 0.26, chance 1%, and the preview with Burst in place of Kill lifts Marine agreement a lot (ρ 0.241 → 0.330, pairwise 59.2% → 63.9%), while it leaves Tyranids flat (ρ 0.589 → 0.591, pairwise 75.4% → 75.4%). Over the seven headline lists: ρ 0.490 → 0.516 (+0.026, under the 0.065 bar) and pairwise 70.8% → 72.1% (+1.4 points).
- Recommendation: don't build
burstKillon the strength of this check: the hypothesis failed where it was made and named. If the supervisor wants to chase the Marine signal, it should be built as an ordinary step judged on the seven lists, knowing the preview sits under the keep bar on ρ. Burst is a better single predictor of the consensus than Kill on both armies (Tyranids 0.28 against 0.17, Marines 0.36 against 0.30, Spearman; not pre-registered), which is the one reason to keep it in view.
The numbers
The pre-registered test (Tyranids, 52 units, all graded by the headline lists)
| Correlation with the residual (ledger rank − consensus rank) | Spearman ρ | One-sided permutation p (10,000) | 95% bootstrap (10,000) |
|---|---|---|---|
| Burst shift (Kill rank − Burst rank), primary | 0.079 | 0.288 | −0.200 to 0.346 |
| Reliability (Burst ÷ Kill), secondary | −0.015 | 0.543 | −0.282 to 0.247 |
| Burst shift, both phases (shoot then fight), reported only | 0.062 | 0.333 | −0.211 to 0.336 |
Verdict: doesn't hold (needed ρ > 0 with p < 0.05 and at least 6 of the 10 named units; got p 0.29 and 4 of 10).
The named units
| Unit | Predicted | Kill rank → Burst rank (of 52) | Moved as predicted | Ledger rank | Consensus rank | Residual |
|---|---|---|---|---|---|---|
| Exocrine | rise | 21 → 19 | yes | 20 | 6.5 | +13.5 |
| Tyrannofex | rise | 23 → 28 | no | 25 | 6.5 | +18.5 |
| Maleceptor | rise | 34 → 35 | no | 17 | 11 | +6 |
| Haruspex | rise | 12 → 13 | no | 24 | 22 | +2 |
| Genestealers | rise | 8 → 8 | no (no move) | 34 | 14 | +20 |
| Raveners | rise | 19 → 12 | yes | 35 | 14 | +21 |
| Toxicrene | fall | 11 → 18 | yes | 23 | 46.5 | −23.5 |
| Hierophant | fall | 44 → 43 | no | 42 | 44.5 | −2.5 |
| Tervigon | fall | 46 → 41 | no | 16 | 40 | −24 |
| Tyrannocyte | fall | 35 → 44 | yes | 48 | 46.5 | +1.5 |
Tyrannofex and Maleceptor are big single models with good guns but few shots: they rarely clear a whole squad or a whole monster, so their chance to finish is lower than their average suggests. Genestealers and Haruspex already sit high on Kill and stay there. Tervigon's and Hierophant's Kill is poor and their Burst a little less poor than the units around them.
The preview (not a step)
Each solo form's Kill replaced by its Burst in its natural mode's slot, × 2.21 for Tyranids (× 2.90 for Marines) so the median ranking form's Kill is unchanged; everything else at the step-6 baseline; scored as tools/ledgerstep.js does.
| List | ρ base → Burst | Pairwise base → Burst |
|---|---|---|
| Tyranids auspexFull | 0.688 → 0.692 | 80.8% → 80.5% |
| Tyranids hivemindFull | 0.664 → 0.662 | 80.2% → 79.5% |
| Tyranids secondFull | 0.644 → 0.610 | 78.6% → 76.5% |
| Tyranids maelstrom | 0.571 → 0.594 | 71.4% → 73.5% |
| Tyranids astrategas | 0.381 → 0.394 | 65.9% → 67.2% |
| Tyranids, mean | 0.589 → 0.591 | 75.4% → 75.4% |
| Space Marines auspexFull | 0.284 → 0.373 | 63.4% → 66.8% |
| Space Marines tobias | 0.197 → 0.287 | 54.9% → 61.0% |
| Space Marines, mean | 0.241 → 0.330 | 59.2% → 63.9% |
| The seven, mean | 0.490 → 0.516 | 70.8% → 72.1% |
Moved most, Tyranids: Ripper Swarms 37 → 50, Winged Tyranid Prime 31 → 23, Toxicrene 23 → 31, Psychophage 27 → 35, Hive Guard 28 → 36, Raveners 35 → 28, Mawloc 41 → 34, Carnifexes 30 → 37. Marines: Apothecary 10 → 77, Rhino 20 → 70, Firestrike Servo-Turrets 65 → 22, Captain in Phobos Armour 37 → 76, Lieutenant in Phobos Armour 44 → 78, Company Heroes 25 → 52, Techmarine 13 → 38, Gladiator Reaper 54 → 30.
Space Marines (a transfer check, no named units)
84 units, all graded by auspexFull or tobias. Burst shift with the residual: ρ 0.263, p 0.009, 95% 0.050 to 0.461; reliability 0.157, p 0.084 (−0.066 to 0.373); both phases 0.219, p 0.023. Kill and Burst agree less here (0.75 against the Tyranids' 0.87): the Marines' cheap characters and transports chip at many targets and finish few.
Every Tyranid unit (by ledger rank)
Kill and Burst per 100 points, β 6 for Kill. Shift: Kill rank − Burst rank (+ rises by Burst). Residual: ledger rank − consensus rank among the graded (+ the reviewers rate it better than the ledger). Best targets: its top three by points finished, the chance it finishes each in one activation.
| Unit | Mode | Kill (rank) | Burst (rank) | Shift | Reliability | Ledger rank | Consensus letter (rank) | Residual | Its best targets (chance finished) |
|---|---|---|---|---|---|---|---|---|---|
| The Swarmlord | fight | 125.3 (22) | 49.4 (26) | −4 | 0.39 | 1 | 4.50 (4) | −3 | Hyperadapted Raveners ×5 96%; Morvenn Vahl ×1 75% |
| Lictor | fight | 133.6 (18) | 65.6 (15) | +3 | 0.49 | 2 | 4.80 (3) | −1 | Painboy ×1 93%; Inquisitor Draxus ×1 85%; Infernal Master ×1 72% |
| Hive Tyrant | fight | 98.8 (33) | 50.9 (23) | +10 | 0.51 | 3 | 4.40 (6.5) | −3.5 | Pteraxii Sterylizors ×5 90%; Morvenn Vahl ×1 73%; Hyperadapted Raveners ×5 91% |
| Zoanthropes | shoot | 107.6 (27) | 57.2 (20) | +7 | 0.53 | 4 | 3.60 (19) | −15 | Militarum Tempestus Command Squad ×5 79%; Painboy ×1 93%; Technomancer ×1 90% |
| Biovores | shoot | 63.8 (49) | 8.0 (49) | 0 | 0.13 | 5 | 5.00 (1.5) | +3.5 | Warlock ×1 60%; Painboy ×1 21%; Benefictus ×1 23% |
| Von Ryan's Leapers | fight | 170.2 (5) | 84.4 (3) | +2 | 0.50 | 6 | 3.40 (22) | −16 | Swooping Hawks ×10 96%; Pteraxii Sterylizors ×5 96%; Mandrakes ×10 96% |
| Norn Assimilator | fight | 100.3 (28) | 39.1 (33) | −5 | 0.39 | 7 | 3.60 (19) | −12 | Morvenn Vahl ×1 71%; Hyperadapted Raveners ×5 86% |
| Deathleaper | fight | 100.2 (29) | 49.2 (27) | +2 | 0.49 | 8 | 3.40 (22) | −14 | Painboy ×1 93%; Inquisitor Draxus ×1 85%; Infernal Master ×1 72% |
| Hyperadapted Raveners | fight | 170.8 (4) | 63.3 (16) | −12 | 0.37 | 9 | 4.40 (6.5) | +2.5 | Pteraxii Sterylizors ×5 99%; Morvenn Vahl ×1 73%; Tempestus Scions ×10 91% |
| Winged Hive Tyrant | fight | 82.7 (40) | 46.4 (29) | +11 | 0.56 | 10 | 3.20 (25) | −15 | Hyperadapted Raveners ×5 78%; Marneus Calgar in Armour of Antilochus ×1 81%; Azrael ×1 86% |
| Old One Eye | fight | 99.9 (32) | 50.2 (24) | +8 | 0.50 | 11 | 3.00 (28.5) | −17.5 | Morvenn Vahl ×1 47%; Infernal Master ×1 86%; Scourges with Heavy Weapons ×5 71% |
| The Red Terror | fight | 134.6 (16) | 68.3 (11) | +5 | 0.51 | 12 | 4.33 (9) | +3 | Pteraxii Sterylizors ×5 88%; Scourges with Heavy Weapons ×5 97%; Marneus Calgar in Armour of Antilochus ×1 83% |
| Tyranid Warriors with Ranged Bio-Weapons | fight | 236.2 (1) | 73.6 (6) | −5 | 0.31 | 13 | 1.80 (43) | −30 | Scourges with Heavy Weapons ×5 82%; Militarum Tempestus Command Squad ×5 82%; Wolf Scouts ×6 96% |
| Tyranid Prime with Lash Whip | fight | 158.2 (13) | 86.2 (2) | +11 | 0.54 | 14 | 3.75 (16) | −2 | Scourges with Heavy Weapons ×5 89%; Retributor Squad ×5 89%; Militarum Tempestus Command Squad ×5 98% |
| Broodlord | fight | 134.0 (17) | 78.4 (5) | +12 | 0.58 | 15 | 3.80 (14) | +1 | Infernal Master ×1 95%; Solitaire ×1 89%; Inquisitor Draxus ×1 97% |
| Tervigon | fight | 68.6 (46) | 25.3 (41) | +5 | 0.37 | 16 | 2.00 (40) | −24 | Painboy ×1 83%; Technomancer ×1 77%; Infernal Master ×1 59% |
| Maleceptor | shoot | 98.7 (34) | 35.2 (35) | −1 | 0.36 | 17 | 4.20 (11) | +6 | Hyperadapted Raveners ×5 61%; Azrael ×1 65% |
| Tyranid Warriors with Melee Bio-Weapons | fight | 165.7 (6) | 72.8 (7) | −1 | 0.44 | 18 | 3.00 (28.5) | −10.5 | Swooping Hawks ×10 100%; Necron Warriors ×20 96%; Incubi ×10 100% |
| Norn Emissary | fight | 91.3 (37) | 35.0 (36) | +1 | 0.38 | 19 | 4.20 (11) | +8 | Pteraxii Sterylizors ×5 83%; Lhykhis ×1 94%; Asurmen ×1 94% |
| Exocrine | shoot | 127.7 (21) | 57.4 (19) | +2 | 0.45 | 20 | 4.40 (6.5) | +13.5 | Hyperadapted Raveners ×5 78%; Azrael ×1 78% |
| Neurolictor | fight | 75.2 (42) | 27.5 (40) | +2 | 0.37 | 21 | 5.00 (1.5) | +19.5 | Painboy ×1 88%; Militarum Tempestus Command Squad ×5 70%; Cadian Command Squad ×5 91% |
| Trygon | fight | 162.4 (10) | 67.9 (14) | −4 | 0.42 | 22 | 3.00 (28.5) | −6.5 | Pteraxii Sterylizors ×5 84%; Hyperadapted Raveners ×5 84%; Azrael ×1 94% |
| Toxicrene | fight | 161.5 (11) | 58.6 (18) | −7 | 0.36 | 23 | 1.33 (46.5) | −23.5 | Pteraxii Sterylizors ×5 84%; Scourges with Heavy Weapons ×5 84%; Solitaire ×1 94% |
| Haruspex | fight | 160.8 (12) | 67.9 (13) | −1 | 0.42 | 24 | 3.40 (22) | +2 | Pteraxii Sterylizors ×5 86%; Scourges with Heavy Weapons ×5 94%; Morvenn Vahl ×1 60% |
| Tyrannofex | shoot | 122.4 (23) | 46.9 (28) | −5 | 0.38 | 25 | 4.40 (6.5) | +18.5 | Morvenn Vahl ×1 59%; The Red Terror ×1 90%; Storm Speeder Hammerstrike ×1 79% |
| Hormagaunts | fight | 110.2 (26) | 60.6 (17) | +9 | 0.55 | 26 | 4.20 (11) | +15 | Swooping Hawks ×10 80%; Scourges with Heavy Weapons ×5 99%; Mandrakes ×10 80% |
| Psychophage | fight | 115.7 (25) | 39.9 (32) | −7 | 0.34 | 27 | 3.00 (28.5) | −1.5 | Infernal Master ×1 87%; Inquisitor Draxus ×1 94%; Solitaire ×1 75% |
| Hive Guard | shoot | 153.5 (14) | 49.8 (25) | −11 | 0.32 | 28 | 3.00 (28.5) | −0.5 | Scourges with Heavy Weapons ×5 84%; Pteraxii Sterylizors ×5 64%; Eldrad Ulthran ×1 86% |
| Neurogaunts | fight | 55.2 (50) | 13.9 (47) | +3 | 0.25 | 29 | 2.20 (36.5) | −7.5 | Benefictus ×1 74%; Kroot War Shaper ×1 87%; Reductus Saboteur ×1 74% |
| Carnifexes | shoot | 163.7 (9) | 31.5 (37) | −28 | 0.19 | 30 | 2.00 (40) | −10 | Scourges with Heavy Weapons ×5 62%; Benefictus ×1 91%; Painboy ×1 82% |
| Winged Tyranid Prime | fight | 127.8 (20) | 56.5 (21) | −1 | 0.44 | 31 | 2.60 (34.5) | −3.5 | Solitaire ×1 78%; Painboy ×1 92%; Canoness with Jump Pack ×1 78% |
| Screamer-killer | fight | 165.4 (7) | 70.4 (10) | −3 | 0.43 | 32 | 2.80 (32.5) | −0.5 | Hyperadapted Raveners ×5 89%; Azrael ×1 89% |
| Harpy | shoot | 73.7 (43) | 29.8 (38) | +5 | 0.40 | 33 | 0.80 (50.5) | −17.5 | Scourges with Heavy Weapons ×5 85%; Pteraxii Sterylizors ×5 59%; Militarum Tempestus Command Squad ×5 85% |
| Genestealers | fight | 164.8 (8) | 71.5 (8) | 0 | 0.43 | 34 | 3.80 (14) | +20 | Swooping Hawks ×10 100%; Incubi ×10 100%; Pteraxii Sterylizors ×5 97% |
| Raveners | fight | 130.1 (19) | 68.1 (12) | +7 | 0.52 | 35 | 3.80 (14) | +21 | Pteraxii Sterylizors ×5 90%; Azrael ×1 87%; Scourges with Heavy Weapons ×5 90% |
| Termagants | shoot | 86.0 (39) | 13.3 (48) | −9 | 0.16 | 36 | 3.00 (28.5) | +7.5 | Scourges with Heavy Weapons ×5 37%; Benefictus ×1 62%; Ratlings ×5 78% |
| Ripper Swarms | fight | 100.2 (30) | 6.6 (50) | −20 | 0.07 | 37 | 2.00 (40) | −3 | Ratlings ×5 51%; Kroot War Shaper ×1 46%; Benefictus ×1 33% |
| Pyrovores | shoot | 146.0 (15) | 71.0 (9) | +6 | 0.49 | 38 | 3.33 (24) | +14 | Scourges with Heavy Weapons ×5 99%; Swooping Hawks ×10 67%; Retributor Squad ×5 91% |
| Neurotyrant | shoot | 88.3 (38) | 29.6 (39) | −1 | 0.34 | 39 | 3.67 (17) | +22 | Solitaire ×1 68%; Infernal Master ×1 68%; Technomancer ×1 82% |
| Barbgaunts | shoot | 222.0 (2) | 97.9 (1) | +1 | 0.44 | 40 | 2.00 (40) | 0 | Scourges with Heavy Weapons ×5 87%; Swooping Hawks ×10 50%; Militarum Tempestus Command Squad ×5 87% |
| Mawloc | fight | 100.0 (31) | 56.0 (22) | +9 | 0.56 | 41 | 2.60 (34.5) | +6.5 | Scourges with Heavy Weapons ×5 98%; Tempestus Scions ×10 78%; Retributor Squad ×5 97% |
| Hierophant | shoot | 72.0 (44) | 19.7 (43) | +1 | 0.27 | 42 | 1.50 (44.5) | −2.5 | Venatari Custodians ×6 79%; Chosen ×10 99%; Einhyr Hearthguard ×10 93% |
| Parasite of Mortrex | fight | 92.3 (36) | 37.6 (34) | +2 | 0.41 | 43 | 2.20 (36.5) | +6.5 | Painboy ×1 79%; Benefictus ×1 72%; Kroot War Shaper ×1 94% |
| Harridan | shoot | 67.8 (47) | 23.9 (42) | +5 | 0.35 | 44 | 1.50 (44.5) | −0.5 | Chosen ×10 97%; Exalted Eightbound ×6 89%; Venatari Custodians ×6 66% |
| Venomthropes | fight | 173.9 (3) | 84.2 (4) | −1 | 0.48 | 45 | 2.00 (40) | +5 | Scourges with Heavy Weapons ×5 79%; Solitaire ×1 79%; Wolf Scouts ×6 96% |
| Hive Crone | shoot | 117.6 (24) | 40.3 (31) | −7 | 0.34 | 46 | 0.80 (50.5) | −4.5 | Scourges with Heavy Weapons ×5 98%; Swooping Hawks ×10 63%; Pteraxii Sterylizors ×5 68% |
| Gargoyles | shoot | 65.1 (48) | 14.0 (46) | +2 | 0.22 | 47 | 3.60 (19) | +28 | Benefictus ×1 85%; Wolf Scouts ×6 69%; Painboy ×1 67% |
| Tyrannocyte | fight | 93.7 (35) | 17.2 (44) | −9 | 0.18 | 48 | 1.33 (46.5) | +1.5 | Painboy ×1 53%; Kroot War Shaper ×1 79%; Benefictus ×1 53% |
| Tyrant Guard | fight | 79.2 (41) | 43.0 (30) | +11 | 0.54 | 49 | 2.80 (32.5) | +16.5 | Scourges with Heavy Weapons ×5 93%; Pteraxii Sterylizors ×5 73%; Infernal Master ×1 96% |
| Sporocyst | shoot | 69.1 (45) | 16.3 (45) | 0 | 0.24 | 50 | 1.25 (48) | +2 | Technomancer ×1 69%; Infernal Master ×1 52%; Painboy ×1 65% |
| Mucolid Spores | shoot | 0.0 (51.5) | 0.0 (51.5) | 0 | – | 51 | 0.80 (50.5) | +0.5 | C'tan Shard of the Void Dragon ×1 0%; Allarus Custodians ×6 0%; Defiler ×1 0% |
| Spore Mines | shoot | 0.0 (51.5) | 0.0 (51.5) | 0 | – | 52 | 0.80 (50.5) | +1.5 | C'tan Shard of the Void Dragon ×1 0%; Allarus Custodians ×6 0%; Defiler ×1 0% |
Mucolid Spores and Spore Mines have no attacks of their own: no Kill and no Burst.
Method notes and doubts
- Exact dice, no simulation. The walk is checked against ev.js's own exact kill calculation (tests/unit/burst.test.js: the chance of slaying every model and the mean kills agree to 1e-9). So no seed or ±0.01 error for P; the permutation and bootstrap are seeded (642, 643, 644).
- Forms rebuilt and checked: each ranking form was rebuilt through tools/ledger.js on the frozen inputs; its β-6 Kill recomputed agrees with the JSON's within 5e-5. 819 Tyranid solo forms (453 dice signatures) and 2,691 Marine solo forms for the preview.
- What the walk leaves out: saves are taken attack by attack, not all together lowest first; damage never spills to the next model (Devastating Wounds included, as ev.js has it); a unit puts its whole activation into one target (no split fire); the charge always lands (the ledger's Kill assumes the same); a target's own abilities only as its defender profile carries them, as in the ledger.
- Natural mode picks one phase: 33 Tyranid units fight, 19 shoot (Tyranid Warriors with Ranged Bio-Weapons fight, since their raw Kill is larger there). The both-phases reading was close to the same (0.06).
- "Top 3 targets": the best 3 unit copies' worth of the typical army (15.2 copies across 630 rows, 611 after joining rows of the same datasheet, size and points); the best targets are mostly expensive characters and small elite squads.
- Pre-registration: the definition and prediction were committed before the run that read a letter. The earlier reports named the over- and under-rated units already (the brief's named list comes from them), so the named-unit part was never blind; the correlation test was.
- The defending side (Soak counting only enemy activations that can finish the unit) would need the walk the other way: each enemy unit of the typical army as the attacker (its side's groups are already in prepArmy) into each of our forms as the target (stagesOf's groups, with our defending modset in place of none, which volleysOf now hard-wires), then the enemy points spent until the form is finished: activations drawn best first by points per chance of finishing, with damage carried between activations (the walk takes a starting state, so one activation's leftover state can feed the next). About 630 × 819 walks for Tyranids, a few minutes.
Files and runtime
tools/burst.js (node tools/burst.js --write, Tyranids 30 s with the preview; --army space-marines --write, 2.5 min), tests/unit/burst.test.js, reports/tiers/2026-10-04-burst-check.json and -space-marines.json (every unit's Kill, Burst, ranks, consensus, residual, best targets; the statistics; the preview).
Pre-registration (written and committed before the reviewers' letters were read for this check)
Committed alone in 4c5e423693; unchanged since.
Everything below is fixed before the first run that reads a reviewer letter. The results sections are added in a later commit; any change to this section after that commit would be marked as such.
The idea being tested
Jordan's colleague (a Tyranid player, 4 Oct): what matters is finishing a target in one activation, not average damage. A unit left alive runs, hides and shoots back; partial damage on a monster is worth little. A reliable gun that clears the bar beats one with the same mean that falls short half the time.
Burst, as measured (tools/burst.js)
- The unit's form: each Tyranid unit's ranking row at the step-6 baseline (tools/ledger-benchmarks/2026-10-03-step6-baseline.json through docs/lab/ledger/ledger.js compute(), exactly as tools/ledgerstep.js scores it): its size, its loadout (Auto-equip's or the datasheet's, as the form says) and its detachment's always-on dice (the first detachment the form lists; every detachment in a form's group gives the same dice by construction). Rebuilt through tools/ledger.js (formEffects, formSide) on the frozen inputs, and checked: its β-6 Kill recomputed must match the JSON's.
- The targets: the ledger's typical 2,000-point army (tools/ledger.js typicalArmy and prepArmy: 630 rows, 15.2 unit copies), each row at the size it is fielded, default wargear, its defender profiles in the matrix's order (front group first; a fractional group count rounded, the front group taking the remainder).
- One activation: the unit's natural mode only. Natural mode = the phase (shooting or fight) that gives the larger share of the form's raw ledger Kill at β 6 (ties: shooting). Shooting: every gun into the one target (each gun its best profile for that target by expected removal, as tools/v4walk.js into() picks it). Fight: the charge lands, every model its best melee weapon for that target and Extra Attacks on top (as into() assigns them); never into AIRCRAFT without FLY.
- The dice: exact, no Monte Carlo (so no seed and no ±0.01 error): per weapon, ev.js's own per-attack distribution of unsaved wounds and damage after Feel No Pain (profileVs: hit, critical hit, Sustained and Lethal Hits, wound and Anti, Devastating Wounds, save and invulnerable, the modifiers from the form's dice effects, Lance on the charge), the number of attacks from the dice (n draws of the weapon's A, Blast by the target's models), then a walk of the target's state (models slain, wounds left on the model taking damage) attack by attack: the model in front's profile decides that attack's unsaved wounds, damage beyond the model in front is lost (no spill, as ev.js killDP), the carried damage stays on the model. Weapons fire one after another, lowest average damage per unsaved wound first (a player clears the small models before the big guns); a One Shot weapon fires in a fifth of activations (the ledger's share), all or nothing. A rule the engine averages over a volley (one re-roll a volley) stays averaged.
- Finished (the colleague's rule, fixed): a target with the MONSTER or VEHICLE keyword only when every model is destroyed; any other unit when fewer than 25% of its starting models remain (slain ≥ floor(0.75 × models) + 1, so a single model, or a unit of up to 4, must be destroyed). P_j = the chance of that.
- Points finished: E_j = P_j × the target's points (one copy).
- Choosing a target as a player would: the targets sorted by E_j, best first; Burst is the mean E over the best 3 unit copies' worth of them (weighted by each row's copies in the typical army, so "its top 3 targets" of a 15-unit enemy army), × 100 ÷ the form's points. Points finished per activation per 100 points.
- Kill (the ledger's average): the same form's raw Kill at β 6, ranged + melee (every charge landing), per 100 points, from docs/lab/ledger/tyranids.json.
- Reliability: Burst ÷ Kill.
- A second reading, reported beside it but not judged: Burst with both phases (its shooting, then its fight into what is left of the same target).
The yardstick
- Consensus: each unit's mean letter over the headline Tyranid lists that grade it (auspexFull, hivemindFull, secondFull, maelstrom, astrategas; S 5, A 4, B 3, C 2, D 1, F 0), ranked best first (ties share the mean rank).
- Ledger rank: the unit's rank by Value at the step-6 baseline (1 best of 52).
- Residual: ledger rank − consensus rank (positive: the reviewers rate it better than the ledger does). Units no headline list grades are left out of every statistic.
- Burst shift: Kill rank − Burst rank among the 52 (positive: it rises when ranked by Burst instead of Kill).
The prediction (if the idea holds)
- Primary: Spearman ρ(residual, Burst shift) > 0: the units the reviewers rate above the ledger rise by Burst over Kill, the over-rated ones fall.
- Named units: Exocrine, Tyrannofex, Maleceptor, Haruspex, Genestealers, Raveners rise (Burst shift > 0); Toxicrene, Hierophant, Tervigon, Tyrannocyte fall (Burst shift < 0). A unit with no Burst and no Kill counts as not moving the predicted way.
- Secondary: Spearman ρ(residual, reliability) > 0.
Tests: one-sided permutation p (shuffle the residual over the units, 10,000 times, seeded), and a 95% bootstrap interval for ρ (resample units, 10,000 times, seeded).
Verdict: "holds" when the primary ρ is positive with p < 0.05 and at least 6 of the 10 named units move the predicted way; otherwise "doesn't hold". The secondary correlation and the both-phases reading are reported, not judged.
The preview (not a step)
The ledger with each solo form's Kill replaced by its Burst (in its natural mode's slot, so the charge model still scales a melee Burst), × one constant so the median ranking form's Kill is unchanged, everything else at step 6; joined forms unchanged (they don't rank). Scored on the headline lists (pairwise and ρ), Tyranids; Space Marines too if cheap (their ledger, auspexFull and tobias). A preview only: a step would be the burstKill switch, judged as the others.