Commit graph

9 commits

Author SHA1 Message Date
bot-backend
42b0ebe338 fix(tradein/estimate): залипший anchor_tier молча глушил IMV-blend и quarter-index
All checks were successful
CI Trade-In / browser-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Has been skipped
CI Trade-In / changes (pull_request) Successful in 9s
CI / changes (pull_request) Successful in 9s
CI Trade-In / frontend-checks (pull_request) Has been skipped
CI / backend-tests (pull_request) Has been skipped
CI Trade-In / backend-tests (pull_request) Successful in 4m32s
Флаг anchor_tier оставался равным anchor_tier_fetched ("A"/"C"), когда якорь
фактически НЕ строился — сброс делал только low-conf гейт (#audit-1), но не
Tier C corridor-гейт (#1795) и не сама _compute_same_building_anchor, когда
она отклоняет кандидата (комплов меньше estimate_sb_min_comps). Дальше по коду
залипший флаг читается как «headline построил якорь» и молча глушит IMV/Yandex
blend (#651, гейт `anchor_tier is None`) и quarter-index correction (#764
Guard-1a) — притом что радиусный headline их не получал.

Замер: 154 из 996 сделок теряют tier-флаг этой правкой, и у всех 154 изменение
цены ровно 0.000% — чинится именно залипший ФЛАГ, не ценообразование (баланс
метрик бэктеста подтверждает: единственная дельта в baseline — новая канарейка
unrecorded_lookup_calls, все остальные метрики побитово те же).

- estimator.py: сброс `anchor_tier = None` единой веткой `if anchor is None`
  после всех трёх гейтов (Tier C / low-conf / _compute_same_building_anchor);
  display-only IMV-карточка больше не гейтится по `anchor_tier is not None`
  (иначе терялась в щели «тир добыт, якорь не построен, headline подавлен»).
- backtest_estimator.py: quarter_index_lookup/quarter_indexes_lookup в реплее
  отвечают «промах» (None/{}), если сброс флага открыл путь, которого не было
  в замороженной фикстуре, вместо падения с RuntimeError; счётчик таких промахов
  уходит в baseline как unrecorded_lookup_calls (точное целое, канарейка на
  расхождение реплея с захватом). Заодно пиннится estimate_dedup_analogs_enabled
  = False внутри replay_fixture (было только в самом гейте) — иначе штатная
  регенерация baseline (--from-fixture --update-baseline) писала baseline,
  который тест не совпадал бы никогда.
- backtest_baseline.json: перегенерирован штатным путём, unrecorded_lookup_calls=0.

Выделено из #2656/PR #2661 — фильтры свежести (scraped_at) в якоре дома и в
знаменателе коэффициента выкупа остаются в исходном PR как отдельная, более
спорная правка (двигает деньги: знаменатель просаживается на ~1.3% по бакетам).
2026-08-17 10:54:07 +03:00
bot-backend
25a865f27b feat(tradein/backtest): --city per-city валидация + corridor city-parity
All checks were successful
CI / changes (pull_request) Successful in 8s
CI Trade-In / changes (pull_request) Successful in 9s
CI Trade-In / frontend-checks (pull_request) Has been skipped
CI / backend-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Has been skipped
CI Trade-In / backend-tests (pull_request) Successful in 51s
D validation harness для oblast-точности:
- --city '<точное deals.city>' → scope выборки по deals.city; default None = байт-идентично (тот же _SAMPLE_SQL object, literal identity).
- per-city PPM2 band из deal_city_price_bands (fallback global, 0 доп-запросов при None).
- corridor city-parity: backtest _fetch_dkp_corridor теперь передаёт city=_resolve_target_city(deal.address) — зеркалит прод C2 (был cross-city contaminated). NB: меняет и default-run live-числа (не fixture-replay) — intended catch-up к проду.
- _compute_confidence подтверждён comp-count-based (без EKB-хардкода).
- +10 unit-тестов. Регресс-гейт байт-зелёный (replay не гоняет corridor).

code-review  APPROVE. Frozen fixture/baseline не тронуты. Live per-city backtest прогоню после деплоя.
2026-07-12 20:49:04 +03:00
d22e0e00ae tech-debt(tradein/estimator): collapse won estimate_* flags into defaults (#1970) (#2475)
All checks were successful
Deploy Trade-In / changes (push) Successful in 10s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / test (push) Successful in 45s
Deploy Trade-In / build-backend (push) Successful in 58s
Deploy Trade-In / deploy (push) Successful in 47s
2026-07-12 12:46:18 +00:00
d15ff3d99d feat(tradein/estimator): сегментная поправка headline+выкуп по ценовому бэнду за флагом (#2255) (#2293)
All checks were successful
Deploy Trade-In / changes (push) Successful in 9s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / test (push) Successful in 1m29s
Deploy Trade-In / build-backend (push) Successful in 56s
Deploy Trade-In / deploy (push) Successful in 44s
2026-07-03 18:43:38 +00:00
ec0f113d4f chore(tradein/backtest): --resolve-house-id flag to measure Tier-S + IMV anchor (#2002)
All checks were successful
Deploy Trade-In / changes (push) Successful in 10s
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / test (push) Successful in 1m27s
Deploy Trade-In / build-backend (push) Successful in 56s
Deploy Trade-In / deploy (push) Successful in 53s
2026-06-27 19:44:38 +00:00
ec9ed56fad feat(tradein/backtest): hermetic estimator regression gate — frozen fixture + baseline (#1966 PR 3/3)
All checks were successful
Deploy Trade-In / changes (push) Successful in 11s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / test (push) Successful in 1m25s
Deploy Trade-In / build-backend (push) Successful in 55s
Deploy Trade-In / deploy (push) Successful in 54s
2026-06-27 12:55:35 +00:00
57bf8ffb81 feat(tradein/backtest): full-spine prediction + range-coverage/calibration/segment metrics (#1966 PR 2/3)
All checks were successful
Deploy Trade-In / changes (push) Successful in 10s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-browser (push) Has been skipped
Deploy Trade-In / test (push) Successful in 1m24s
Deploy Trade-In / build-backend (push) Successful in 53s
Deploy Trade-In / deploy (push) Successful in 48s
Full-spine backtest harness via _price_from_inputs + range-coverage/calibration/per-segment MAPE. code-reviewer APPROVE. Real prod numbers: overall MAPE 18.4%/bias -1.2%; segments эконом +16% / бизнес -23% / элит -38%; range-coverage 56.6%. Read-only; full suite 2488 passed. Refs #1966.
2026-06-27 11:21:57 +00:00
d021fc9231 feat(backtest): per-rooms asking→sold correction block (#648 S1) (#650)
All checks were successful
Deploy Trade-In / changes (push) Successful in 6s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-backend (push) Successful in 43s
Deploy Trade-In / deploy (push) Successful in 34s
2026-05-29 13:11:55 +00:00
228d12eab1 feat(backtest): read-only estimator-vs-ДКП accuracy harness (#648) (#649)
All checks were successful
Deploy Trade-In / changes (push) Successful in 6s
Deploy Trade-In / build-frontend (push) Has been skipped
Deploy Trade-In / build-backend (push) Successful in 41s
Deploy Trade-In / deploy (push) Successful in 33s
2026-05-29 12:54:00 +00:00