# Evidence items for the likelihood model (draft)

Candidates: Traversette (Durance-Guil), Clapier (Isère-Arc-Savine, incl. Petit Mont-Cenis variant), Mont-Cenis (Isère-Arc), Montgenèvre (Durance), Larche (Durance-Ubaye), Petit-Saint-Bernard (Isère-Tarentaise), Grand-Saint-Bernard, Other col (Agnel, Croix, Échelle, Autaret, etc.).

Source families (items within a family are NOT independent; the collapsed run treats each family as one item):
  P = Polybius book 3 narrative (and Livy where he repeats it): distances, days, Island, Allobroges, ambush terrain, summit, view, snow, descent, destination.
  L = Livy's own geography (Tricastini, Vocontii, Tricorii, Druentia) — probably Coelius/annalistic; may be contaminated.
  T = the testimonia on the pass: Polybius 34.10.18 via Strabo ("through the Taurini"), Varro's five passes (Servius), Pompey's letter (Sallust), Coelius' Cremonis iugum, Nepos' saltus Graius, Ammianus.
  G = physical: Mahaney's mire and rockfall; Iron Age/Roman finds at the cols.
  F = feasibility: col height, exposure in early November, practicability for 20,000+ men, thousands of animals, 37 elephants; known use of the col in antiquity.

Items (provisional):
 P1 Island at the Isère-Rhône confluence, 4 days (about 600 stades) from the crossing, "like the Delta", Allobroges adjacent  → favours Isère routes (Clapier, Cenis, PSB); Durance routes need the Island at the Aygues (Skaras) or a doubling back.
 P2 800 stades in ten days "along the river" through flat country escorted through Allobrogian land to the foot of the ascent → Isère routes fit (Valence→Aiguebelle 888 st.); Durance-direct (Avignon→Sisteron 908 st.) fits only with an Orange Island; the Drôme approach reaches only Crest (512) or Die/Luc (790).
 P3 First ambush by ALLOBROGES at a defile with a town near, on the first day of the ascent → Isère routes (Allobrogian territory ends at the Isère/Arc gate); Durance routes have Vocontii/Tricorii/Caturiges, not Allobroges.
 P4 Nine days from the ascent's start to the summit, with two fights, a one-day halt and a night in the gorge → route length from the ascent gate to the col: Clapier 85 km (9.5 km/day), Cenis 87, PSB 63, Montgenèvre-from-Sisteron 132 (14.7), Traversette-from-Sisteron 139 (15.4), de Beer's Drôme-Cabre-Guil 180-230 km (20-26 km/day: implausible with the incidents), Larche 113-143.
 P5 View of the Po plain from the summit (Polybius) / a promontory (Livy) → DEM viewshed: Traversette yes (17% of azimuths), Clapier yes (0.6% from the col, 5-8% from points 1 km away), all others no.
 P6 Old snow from the previous winter under new snow on the descent, in late Oct/early Nov (Pleiades) → Traversette (2947) strongly; Clapier (2482) moderately; Cenis (2083), PSB (2188) weakly; Montgenèvre (1854), Larche (1991) weakly-to-no. (Snow-line data from the terrain dossier; 3rd-century BC glaciers were larger than today's: Holzhauser 2005.)
 P7 Descent narrow and steep, steeper than the ascent (Livy), a broken cliff-path of 1.5 stades / 1000 feet; three days from the cliffs to the plain → Traversette (very steep 2381 m descent in 28 km), Clapier (2015 m in 25 km), Cenis (1675 m in 32 km), Montgenèvre (1496 m in 51 km, gentle), PSB (1885 m in 72 km to Ivrea, gentle), Larche (79 km, gentle). 3 days to the plain: Traversette 28 km, Clapier 25, Cenis 32 (fit); Montgenèvre 51, PSB 72, Larche 79 (strained).
 P8 Descent among the Taurini (Polybius 3.60.8 first action; Livy 21.38.5 "proxima gens"), "into the plain of the Po and the Insubres" (3.56.3) → Susa-valley passes (Clapier, Cenis, Montgenèvre) best; Traversette (Saluzzo: Taurini/Vagienni border) good; Larche (Cuneo: Ligurian Vagienni) weaker; PSB/GSB (Salassi, Libui) poor: Livy's own argument.
 L1 Livy's Tricastini–Vocontii–Tricorii–Druentia → favours Durance routes (Montgenèvre, Traversette, Larche); but "campestri maxime itinere ad Alpes" from the Druentia and "ad laevam" from an Isère Island are contradictory; Walbank et al. treat as contamination. Weight modest.
 T1 Polybius/Strabo: Hannibal's pass = "through the Taurini", distinct from "through the Salassi" → excludes PSB/GSB (strong, Polybius' own words if Strabo quotes him accurately), neutral among Cottian passes.
 T2 Varro's list (Servius Aen. 10.13): sea-Ligurian, Hannibal's, Pompey's, Hasdrubal's, Graian: if south-to-north, Hannibal's lies between the coast and Pompey's (Montgenèvre): Traversette/Larche/Agnel favoured, Montgenèvre excluded, Cenis/Clapier disfavoured (they'd be "Hasdrubal's"?). If the order is not geographic, weak. Also Pompey (Sallust Hist. 2.98.4) "iter aliud atque Hannibal": Pompey's pass ≠ Hannibal's; Pompey's is generally Montgenèvre.
 T3 Coelius' Cremonis iugum and Nepos' saltus Graius (PSB) and the popular Poeninus (GSB) → the annalistic/Roman tradition put him in the Graian/Pennine Alps; Livy rejects it on the Taurini argument. Small weight for PSB.
 G1 Mahaney's Traversette mire: consistent, does not test; tiny weight (LR ~1.0-1.3 for Traversette).
 G2 Finds and known antiquity of the passes: Montgenèvre, Cenis, PSB, GSB were major routes; Clapier a local pass; Traversette a minor col (medieval salt route; 1480 tunnel because mules could not pass loaded) → feasibility.
 F1 Practicability for the army and 37 elephants in early November; Hoyte 1959 took an elephant over Mont-Cenis (after being turned back from Clapier by a rockfall?); Traversette headwall.

Discriminating structure: the Isère-versus-Durance approach question (P1-P3, L1, T2) and the summit question (P5, P6, P7) cross-cut. Clapier is the Isère route with a view and moderate snow; Traversette is the Durance route with the best view and the surest snow; Cenis/Montgenèvre fail the view; PSB/GSB fail the Taurini.

Prior: roughly equal across the six named, lower for GSB, some mass for "other col".