[Afgan0r]

Problem

The deaths recalc reads the per-player death counter signal from the parser_events table, but the bulk full-run (full-run-recalculation.ts) never re-persists parser_events from raw_snapshot — only the single-replay audit path does. For ~90% of current sg replays (1850 / 2059 with artifact d players) parser_events has 0 player_counter rows while raw_snapshot carries the compact counters, so the counter death path finds nothing.

Player resolution and games are built from raw_snapshot, so the game is counted, while a death with no independent victim kill-row (null-killer / suicide / environmental / bleed-out d>0) is dropped. This is the platform-wide ~11024 death-game undercount (sg all-time 282720 vs the parser's own 293744; exemplar Zero 621 vs 653).

Root cause — confirmed by instrumentation on real staging data

.planning/quick/260615-f13b-deaths-undercount/instrument.ts runs the real repository/service code (buildPlayerIdentityIndex, bestPlayerIdentityIndexed, playerIdentityMatchPriority, calculatePlayerAndSquadAggregates) on real exported rows. Across 16 of Zero's d>0 replays he resolves in all 16 — refuting the F12-timestamp / isNicknameActive hypothesis (he matches via the time-ungated display_name priority-1 branch even for pre-2026-05-10 replays) — yet pre-fix only 6 deaths are credited (3 counter, 3 victim) and 10 are dropped because parser_events lacks a counter event and no kill-row names his eid as victim. Post-fix: 16/16, 0 dropped.

Concrete dropped replay: 9eb1a2ad-… (2021-05-14), prid 3662f461-…; raw_snapshot Zero = {d:1, nkd:1, eid:203}; parser_events = 140 kill / 18 destroyed_vehicle / 6 teamkill / 1 diagnostic, 0 player_counter; no kill-row with victim_entity_id=203. (Diagnostics in findings.sql.)

Fix

Read the death counter from the authoritative raw_snapshot — the same source already used for player resolution and games — instead of the stale events table.

• service.ts: AggregatePlayerEvidence gains optional counterDeaths?: DeathStats; new tallyArtifactCounterDeaths folds it into the per-replay scratch (player + squad). Exported artifactCounterDeaths(d, td) builds the DeathStats with the same semantics as the event-based counterDeaths (total = max(d, td), by_teamkills = td); counterDeaths now delegates to it.
• repository.ts: resolvedPlayers attaches counterDeaths = artifactCounterDeaths(player.d, player.td).

Invariants preserved

• One-life cap (foldCappedDeaths): ≤1 total and ≤1 by_teamkills per replay — additive with the event paths, so overlap collapses to 1 (no inflation).
• by_teamkills ≤ total (the max cap).
• Games unaffected; parser/OCAP contract, schema, and API untouched.

Tests

• aggregates.test.ts: 4 new unit cases (3 fail without the fix).
• postgres.test.ts: 1 integration case — raw_snapshot d:1, no counter event, no victim row → death credited through the real DB.

Gates

• pnpm lint, pnpm typecheck, pnpm format, pnpm test (661 unit, +4) — green.
• service.ts 100% unit coverage. The full pnpm test:coverage gate needs Docker/testcontainers (unavailable in the authoring env); the repository.ts change is covered by the new integration test, which runs under Docker in CI.

🤖 Generated with Claude Code

[Afgan0r]

Superseded by a clean PR (fix/deaths-counter-from-artifact) — #20 inadvertently reverted the Sentry feature (#19). Same deaths fix, Sentry preserved.