v0.27.0 — **PHASE 2 COMPLETE**

Added

• T-205 Prog-vs-field curve (PLAN §10 Compare). analysis/prog_trajectory.py::prog_trajectory_for_encounter() returns three things for one encounter:
  • Our sessions — derived from Fight rows whose report_code appears in WatchedReport. Per session: pulls, kills, best_phase, best_fight_percentage. Folds T-102's "M-PARSE-less prog-point tracker (auto)" AC into this — no separate T-102 ship needed.
  • Manual prog points — all ProgPoint rows with source='manual' (from T-010).
  • Field distribution — wipes for the encounter that aren't in our watchlist, binned at 10% fight_percentage resolution.
• GET /api/encounters/{id}/prog-curve.
• React ProgCurveView replaces the T-205 placeholder in CompareView — field histogram with inline bars, your-best-bucket highlighted in blue ("← you"), our-session list below.
• 6 new tests covering session aggregation, manual-point surfacing, field/our separation, bucket resolution, chronological ordering, empty encounter. 213 tests total.
• Live AC on FRU (202 field wipes ingested): distribution shape matches the real prog walls — 62 wipes at P0 Fatebreaker (90–100% remaining), 30 wipes at 70–80% (Adds intermission/P3 entry), 25 at 40–50% (Pandora's Knight), 8 at 0–10% (enrage/Fatebound late). Backbone for the "where most groups die" comparison the static needs.

Phase 2 done

All 8 Phase-2 tasks shipped: T-201 (field backfill), T-202 (consensus → fight_model), T-203 (mechanic classification), T-204 (empirical DPS check), T-205 (prog-vs-field), T-206 (failure cartography), T-207 (M-GATE diagnostic), T-208 (Encounters tab + M-GATE strip).

Note

• T-102 was listed as a Phase-1 task but never explicitly shipped. Its AC ("M-PARSE-less prog-point tracker (auto + manual): furthest phase/session over time; pulls + hours") is fully covered by T-010 (manual) + T-205 (auto from watched reports). Marking T-102 as superseded by T-205 rather than re-shipping.