continue first full balancing round

docs/content_design.md

@@ -440,6 +440,73 @@ typical loadouts double as the `default_modules` for enemy waves):
   scrap-free recipe produces that item, mirroring the reprocessing-path
   rule (see progression_design.md action items).
 
+### Combat stat anchors — first pass
+
+All combat stats derive from a few anchors so that power-per-threat holds
+by construction; the arena suite in `bin/balancing/data/balancing.toml`
+verifies it empirically. Change the anchors first, re-derive stats second.
+
+- **Weapon DPS per threat pays a concentration tax that grows with gun
+  size**: railgun_s 2 dmg × 2.0 Hz = 4.0 DPS at 6.5 threat (0.62),
+  range 50; railgun_m 14 × 1.5 = 21 at 43.5 (0.48), range 70;
+  railgun_l 52 × 0.8 = 41.6 at 102 (0.41), range 100. Arena round 1
+  (see below) showed that at equal threat, concentrated fleets win
+  flawlessly with a flat-ish DPS curve — concentration itself (focus
+  survivability, range, no losses-to-attrition of own DPS) is worth
+  paying for, so bigger guns get less raw DPS per threat.
+- **Hull HP = 15 per threat of hull contribution** (incl. ship base
+  time): drone 60, frigate 300, destroyer 550, cruiser 1500,
+  battlecruiser 2700, battleship 7500, dreadnought/carrier 15500.
+- **Armor HP ≈ 31 per threat** — a strong premium over hull HP (15)
+  because armor is pure HP with no capability, and fights snowball:
+  killing enemies removes their DPS, while surviving merely delays —
+  so a point of HP must be cheaper than a point of DPS.
+  armor_plates adds 1000 HP at 32 threat (arena round 1: at 640 the
+  all-guns destroyer beat the armored one).
+- **TTK check:** a mirror cruiser duel (51 DPS vs 2140 EHP) lasts ~42 s —
+  inside the 30–60 s parity-fight band implied by the 25-ship fleet
+  target. Capital mirrors run longer (~90 s), deliberately.
+- **Repair ≈ 1 HP/s per threat:** repair_tool heals 12 HP × 1 Hz at 13
+  threat (arena round 1: at 2 HP/s-per-threat the escorted team won
+  flawlessly — in-combat sustain effectively removes enemy DPS and must
+  be priced like DPS, not like HP). Salvager: collection range 60 (was
+  a nonsensical 500, beyond any sensor), cargo 20, 0.5 collections/s.
+  `scrap_despawn_seconds` raised 30 → 120: a capital kill drops
+  hundreds of scrap collected one per cycle, so 30 s despawned nearly
+  everything.
+- **Mobility ladder is monotone in size** — small = fast and nimble
+  (drone 45 m/s, 60 accel) down to capitals (10 m/s, 8 accel); the old
+  placeholder had the drone as the least agile ship. Sensor ranges keep
+  the existing 150→350 ladder; all weapon ranges stay below sensor
+  ranges.
+- **Weapon modifiers are capital economy:** ×1.2 damage at 22.5 threat
+  beats adding a gun once a ship carries more than ~68 threat of weapons
+  — so modifiers pay off on gun-heavy big hulls and are a waste on small
+  ones. Kept at ×1.2/×1.2/(×1.5 range, ×0.8 rate) deliberately.
+- **Stations:** a fresh player station (3000 HP, 25 dmg × 1 Hz,
+  range 120) holds one early parity wave unaided; the enemy station at
+  level 0 matches it exactly and scales per push level
+  (+1500 HP, +12 dmg, +0.1 Hz per level). HQ 5000 HP. Range reduced
+  from 200 in round 1: at 200 (4× a small gun's range) two stations
+  annihilated a 3× threat swarm through approach fire alone — range,
+  not HP, is the station dominance lever.
+- **Known imbalance, accepted:** the carrier loses its equal-threat
+  arena until the drone-launching capability exists — the hangar is 224
+  threat of dead weight. Do not compensate with stats; fix by
+  implementing drones.
+
+**Arena round 1 results (2026-07-04):** mirrors snowball as expected
+(40–50% of the winner survives) but team 1 won all three mirrors —
+possible sim-side bias, under investigation (progression_design.md
+action items 6–7). All four equal-threat cross-tier matchups were
+flawless wins for the concentrated side; glass destroyers beat armored;
+repair escort beat raw numbers flawlessly; two stations shrugged off a
+3× threat swarm. Round-2 knob changes: concentration tax on m/l gun
+DPS (17→14, 70→52 damage), armor 640→1000, repair 25→12, station range
+200→120. Gun ranges deliberately untouched — if concentrated fleets
+still win flawlessly in round 2, the range ladder (kiting) is the next
+suspect, ahead of further DPS cuts.
+
 ## Balancing targets (first pass, July 2026)
 
 The six root numbers for the balancing pass. Every derived value (threat

docs/progression_design.md

@@ -373,3 +373,15 @@ in `requirements.md` and the git history). Still open:
    one matching deposit tile). Touches REQ-BLD-MINER ("every asteroid
    tile is equivalent" no longer holds), REQ-GW-ASTEROID-EXPAND /
    REQ-EXP-*, `world.toml`, and `visuals.toml`.
+6. **Investigate team-1 bias in balancing arenas.** In arena round 1
+   (July 2026), team 1 won all three mirror matches with large margins
+   (8/20, 3/6, 1/2 survivors). Re-run mirrors repeatedly; if team 1
+   wins consistently, find and fix the asymmetry (spawn geometry, tick
+   processing order, target-selection tie-breaking). Mirror matches
+   must be statistically symmetric or every arena result carries a
+   side-dependent error.
+7. **Log remaining HP in the balancing arena results.** Win/loss alone
+   hides the margin ("1/1 battleship" can mean 95% or 5% HP left). Add
+   the surviving entities' remaining HP fraction (or a per-team total
+   EHP percentage) to the arena end state, so equal-threat matchups can
+   be tuned toward near-draws quantitatively.