run_encounter gains an opt-in collect_runs flag returning one RunOutcome
per battle (winner, rounds, per-combatant id/side/hp/hp_max/dead) next to
the EncounterReport; default behavior is unchanged. objectives.py provides
the pure optimizer primitives: P(survives all), P(TPK), P(>=1 death),
expected hp lost, expected rounds and a worst-decile (CVaR-style) tail
metric. phase2_spike.py holds the spike-level strengthened monster policy
(focus fire on the lowest-AC enemy, ties to the wounded; party side stays
on default_policy). tests/test_phase2_danger_map.py prints the danger-map
sketch: warband and lich scenarios measured default vs focus-fire at
100 runs / seed 11, pinning the delta deterministically.
Adds the S6 layer on top of the deterministic combat engine:
- metrics.py: closed-form binomial statistics (win rate, standard error,
3-sigma confidence band) with clamped degenerate proportions.
- runner.py: EncounterSpec/Side definitions, per-run SeededRng streams,
global duplicate-id disambiguation, attrition averages per combatant.
- cli.py: pf1e-sim entry point producing a French balance report with
win rates, 3-sigma bands, draws, rounds, and attrition.
- Movement fix: greedy straight-line heuristic could oscillate at walls;
_step_toward now follows the true shortest path via an unbounded
Dijkstra cost field from the target (Grid.reachable budget=None).
- Regression tests: melee unit routes around a wall and engages; grid
unbounded-budget coverage.
ctan