Skip to content

v0.4.7 — mark lethal-threat onset and harden Renju rule correctness

Choose a tag to compare

@github-actions github-actions released this 20 May 01:20
· 319 commits to main since this release

Play this build: https://gomoku2d.byebyebryan.com/

Theme: mark lethal-threat onset and harden Renju rule correctness.

0.4.7 started as the lethal-threat pass: mark the point where a replay becomes
effectively lost, so the analyzer can explain the setup corridor instead of
spending attention on the obvious final conversion. That forced combo threats
and Renju forbidden moves into the same model. The result is a hardening release:
the analyzer now has a clearer endpoint, and the rule engine has a more reliable
Renju legality foundation.

Rules and analysis correctness

  • Added lethal-threat detection for terminal coverage and one-step combo/fork
    threats, including open fours, four-three coverage, double-three coverage,
    shared-block rejects, and Renju forbidden-block cases.
  • Split replay explanation into setup corridor, lethal onset, and lethal tail so
    the UI/report can show how the loser reached an already-lethal state instead
    of treating every final conversion ply as equally meaningful.
  • Replaced the old shape-only Renju forbidden shortcut with a recursive
    legality-aware checker, then added extracted Renju.net advanced examples,
    handwritten fixtures, and external-reference validation notes.
  • Added metrics and prefilters for the corrected Renju checker so the legality
    fix remains practical for bot search, hints, tournament reports, and replay
    analysis.
  • Refreshed the curated bot and analysis reports after the Renju-rule and
    lethal-onset changes.

Web

  • Refined Replay status copy around lethal onset: terminal frames read as a won
    state, post-onset loser frames read as guaranteed loss, setup-corridor frames
    read as no viable escape, and last-escape frames use focused player-facing
    language.
  • Updated the replay timeline to separate traceback progress, setup corridor,
    lethal tail, last escape, and lethal onset instead of treating the whole
    analyzed suffix as one red block.
  • Made Replay turn navigation and board animation calmer by pacing the
    opponent's last-move idle animation and fixing animation listener ownership so
    stone-destroy callbacks cannot get stuck on the last frame.

Repo and docs

  • Promoted lethal-threat terminology and examples into docs/lethal_threats.md
    and aligned docs/game_analysis.md / docs/corridor_search.md around setup
    corridor versus lethal tail.
  • Documented the Renju validation process and corpus so future legality changes
    can be tested against known hard examples instead of relying on intuition.
  • Added a focused v0.4.7 replay screenshot review for the updated timeline,
    status, and last-escape presentation.
  • Set up mistake detection as the narrow remaining analyzer follow-up, which
    shipped in 0.4.8; broader product cleanup moves to the 0.5 polish line.

Full changelog: CHANGELOG.md