Verdict: not adopted. A shape that puts the right transports up exists (in-sample +1.2 pooled), but chosen honestly (the shape and its strengths picked on 10 armies, scored on the other 5) it holds out at −0.0 ± 0.5. Nothing on the page changed; the work stays as a research tool, tools/carrycargo.js, and the riders list in tools/carry.js.
Why today's Carry line throws every transport up
The page's line (wCarry, T-629) is a transport's capacity per 100 of its points ÷ the army's median transport, × the Kill weight (about 500). At 1, a typical transport earns a whole typical unit's Kill in Value whatever it carries, and a cheap one more: the Trukk (60 pts, 12 models) gets +1,341 on top of its own 197 lines. It doesn't double-count the cargo's Kill (the cargo's own Value is untouched); it ignores the cargo altogether. Capacity per point is highest exactly on the cheap dedicated transports, so every one of them rises together, whether its army needs one or not.
The rebuild (sliders; 0 = today)
A transport is worth what a ride adds to the best squad of its own army that may embark (BSData's transport line via tools/transports.js: keywords, exclusions, the most models that fit), in that squad's own Value, per 100 transport points: the rider's melee Kill raised from what lands on its own to carryRide; plus carryShield × the rider's Value, × its fragility (carryFrag: reference Soak ÷ its Soak, to a power); carryPer divides by the pair's points instead; carryDedOnly keeps only Dedicated Transports; carryCap stops a transport outranking its rider; carryOnce discounts the one-use deliveries (Drop Pod, Tyrannocyte); carryNeed × the share of the army's Kill that needs delivering; carrySquads bars lone models. Two findings: lands are near saturation for most riders (0.74 to 1 at the baseline's lands 1.5), so the delivery term barely separates anyone; and what does separate is protection of fragile, valuable cargo on cheap dedicated transports (Wyches, Incubi, Infractors, Rubrics, Acolytes).
The exam (Track A, step34b baseline; 15 armies pooled, Tyranids the training set)
| Setting | Pooled Δ (SE, 95%) | Better | Tyranids | Held out (9 folds) |
|---|---|---|---|---|
| Today's line, wCarry 1 | +0.1 (exam) | – | 79.3 → 75.5 | – |
| Cargo, delivery only (w 1, ride 0.9) | −0.6 (0.6, [−1.9, +0.5]) | 6/15 | −0.3 | – |
| Shape A: w 0.5, shield 0.4, frag 2 (ref 150), dedicated only, cap 1 | +1.2 (0.8, [−0.1, +2.8]) | 7/15 | −1.7 | strengths only: +0.7 ± 0.4 |
| Shape A with carryOnce 1 (Tyrannocyte held), w 0.75 | +1.2 (0.8, [−0.1, +2.8]) | 7/15 | ±0.0 | – |
| Shape and strengths chosen on 10 (16,128 settings) | – | – | – | −0.0 ± 0.5 (seeds +0.5, −0.7, +0.1) |
Shape A per army: Drukhari +14.0, Emperor's Children +5.2, Death Guard +0.8, Thousand Sons +1.3, Genestealer Cults +0.7, AdMech +0.6; Agents −2.2, Orks −1.0, World Eaters −0.9, Space Marines −0.8. Without the Drukhari it is +0.3. The "strengths only" +0.7 picks w on 10 armies for a shape already chosen on all 15, so it leaks; the fully nested −0.0 is the honest number. The ceiling (each transport placed freely) is +4.7 pooled, so the room is there; the features aren't.
The transports (percentile in its army; reviewers' letters)
| Army | Transport | Letters | Today w 0 | Today w 1 | Shape A |
|---|---|---|---|---|---|
| Drukhari | Raider | S | 19 | 86 | 86 |
| Drukhari | Venom | S | 8 | 53 | 83 |
| Thousand Sons | Chaos Rhino | S | 3 | 76 | 38 |
| Emperor's Children | Chaos Rhino | A | 0 | 64 | 68 |
| Death Guard | Chaos Rhino | A | 9 | 54 | 31 |
| World Eaters | Chaos Rhino | B/B/A | 0 | 55 | 97 |
| Astra Militarum | Chimera | A | 32 | 90 | 41 |
| Genestealer Cults | Goliath Truck | B | 22 | 61 | 87 |
| Orks | Trukk | C | 9 | 96 | 24 |
| Orks | Battlewagon | B | 28 | 93 | 28 |
| Space Marines | Rhino | B/D | 6 | 100 | 35 |
| Space Marines | Drop Pod | A/C | 0 | 98 | 39 |
| Space Marines | Razorback | B/D | 30 | 90 | 59 |
| Space Marines | Land Raider Crusader | B/D | 31 | 89 | 28 |
| Tyranids | Tyrannocyte | C/D/D | 8 | 98 | 76 (8 with carryOnce 1) |
The shape fixes the old line's main fault (walkers, Land Raiders and the Battlewagon stay where their own lines put them; the Trukk only to the 24th) and lifts the Drukhari and Emperor's Children boats, but still raises the Space Marine Rhino and Razorback (Devastators, Scouts ride) and the World Eaters' Rhino past A units, and drops the Thousand Sons' S Rhino short. One army (Drukhari, one list of 23 units) carries most of the gain.
Reproduce
node tools/carrycargo.js --bsdata <BSData checkout> --grid '<json>' | --per '<setting>' … | --table '{}' '{"old":1}' '<setting>' (evaluates the line over the page's own Values; the riders come from tools/carry.js at run time, nothing written).
Next: the transport is an army-level question (T-676's second layer): the Drukhari want Shape A and the Orks want nothing. A per-army transport dial, held out list by list where an army has 2+ lists, is where this goes.