Repository navigation
auto/queue → main: autopilot window 2 (slices 7-10, Phase 3 code) - #24
Merged
Merged
Conversation
…uit fit, Miami quali fallback) (#20) * slice-7: Phase 2 carry-forwards (overtaking call-site tests, per-circuit fit, Miami quali fallback) - build_overtaking_data(event_name=...) fits only the forecast circuit; cli.forecast_weekend and backtest.forecast_race_challenger pass it. - tests/test_overtaking_wiring.py fails if either call site stops building the fit, or fits more than one circuit per forecast. - collect_v2 falls back to the fastest non-deleted session lap when a classified quali row has no Q1/Q2/Q3 time; flagged q_segment='laps' and logged. Refs #19 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EiJgo5GgDN14Q7LndX14Qv * queue: slice-7 in-review (PR #20) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EiJgo5GgDN14Q7LndX14Qv --------- Co-authored-by: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LnhCkGvJgZEMRYLWEWTFA1
* slice-8: season chaining and gp season Chains rounds r+1..R within each trial, carrying the car random walk forward; sprints via simulate_weekend; drivers' and constructors' tables from the same trials. gp season writes drivers_championship.csv and constructors_championship.csv; the remaining calendar comes from FastF1 through a stubbable loader and the command fails if it cannot be read. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QRJHJJN7CD3ym1wMdmWnsM * queue: slice-8 in-review (PR #21) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QRJHJJN7CD3ym1wMdmWnsM --------- Co-authored-by: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01QuFzDH4gT5WLYUkweqAXpV
) * slice-9: season-level backtest harness (gp backtest --season-level) Fits at cutoffs 2/6/12 of each held-out season, chains the rest with slice-8's simulator, and scores constructor-share Spearman (next to a rolling legacy number at the same cutoff), title Brier for both championships and 80% interval coverage on final points. Writes a season_level block to backtest_summary.json with gate_valid / null verdict discipline. Tested end to end on the fixture with ADVI; the real backtest has not been run. Refs #19 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JyCVWAmeHjqML4h9ugdbGQ * queue: slice-9 in-review (PR #22) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JyCVWAmeHjqML4h9ugdbGQ --------- Co-authored-by: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01F4qkQYA3HBDbujUy5MWTH6
* slice-10: post-quali podium calibration, ECE for both arms, confident-miss listing Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015bZTZNWERoJTnjfEPcTL8U * queue: slice-10 in-review (PR #23) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015bZTZNWERoJTnjfEPcTL8U --------- Co-authored-by: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Eih57hVHzqtaRFaQqnXdmG
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Brings the four slices from autopilot window 2 (Oct 20-26, log in #19) onto
main. The review routine merged each one intoauto/queueafter re-running the suite and its own mutations. It merges cleanly. The only commit onmainthat isn't here,11de384, makes the same AUTOPILOT.md change asd197145.What's in it
cli.forecast_weekend,backtest.forecast_race_challenger). Each forecast fits only its own circuit.collect_v2falls back to each driver's fastest session lap when a quali row has positions but no Q times (2025 Miami); those rows are flaggedq_segment="laps"src/sim/season.pyandgp season: chains rounds r+1..R with the team random walk carried forward, includes sprints, and builds drivers' and constructors' totals from the same trialsgp backtest --season-level [--cutoffs 2,6,12]: constructor-share Spearman against a rolling-Ridge legacy number, title-winner Brier, 80% coverage, and aseason_levelblock inbacktest_summary.jsonwith thegate_valid/ null-verdict rulecalibration_podium_postquali.csv, binned calibration and ECE forpre,pre_pairedandpostquali, pluspostquali_worse_rounds(more than 1.0 nat worse, with the pole-sitter and the winner)It also brings the window-2
docs/autopilot/queue.mdstatuses anddecisions.mdentries (2026-10-20 → 10-25). 14 files, +1829 / −36. The suite onauto/queuewas 335 passed (#23's review).What merging does not mean
repair_round.py 2025 6 Qagainst the hand-repaired Miami rows (slice-7: Phase 2 carry-forwards (overtaking call-site tests, per-circuit fit, Miami quali fallback) #20), andgp seasoncurrent_pointsagainst the real 2026 standings (slice-8: season chaining and gp season #21).confidence.pyand the pkl are untouched and stay until Phase 3 passes.forecast-data(jolpica race pace, accent-free event names,--grid-file) are not included. They still need porting by hand.Once this merges, the CLAUDE.md / AUTOPILOT.md lines saying "
auto/queueisn't merged" go stale. The doc refresh onfeat/eager-turing-3zk4wbshould be updated to match.Generated by Claude Code