Skip to content

CF Compass v4.2.0

Choose a tag to compare

@github-actions github-actions released this 16 Sep 05:34
· 26 commits to main since this release

CF Compass v4.2.0

本次更新让「只打虚拟赛」与「自行设定训练水平」都能进入日常训练闭环,同时保留独立的官方数据。

双 Rating 与训练偏好(Issue #31)

  • 默认官方模式。在主页个人进度的 Rating 选项中切换本地估算模式;选择按账号保存。
  • 有官方历史 Rated 对照数据的虚拟场次自动按时间累计估分;缺失数据明确标记并可重试。展开估分详情可查看账本。
  • 估算模式同步影响相关训练难度、复习、赛事推荐、Rating 颜色与分析曲线,不覆盖官方缓存、实际 AC 数或真实比赛成绩。
  • 今日训练新增手动参考 Rating(800–3500)、弱项标签,并集中管理复习数量、难度范围及各档题量。手动参考分在官方模式下生效。

估算边界

这不是 Codeforces 官方 Rating,也不是正式赛与虚拟赛混合生涯。轨迹从第一场可计算虚拟赛之前的官方分开始,无历史分时采用模拟起点 1400;后续正式赛不重置这条轨迹。重复比赛和赛前做过题的场次仍计入并提示,不只保留正向涨分。同步提交中不可见的无提交参赛无法发现。缺失场次使结果保持暂定,历史新账号处理为近似,不承诺预测精度。

界面

  • 训练设置集中,估分明细默认折叠,减少重复信息。
  • 修复窄进度卡片文字截断,扩展训练布局,Rating 颜色跟随分数。
  • 活跃用户排名前 500 显示直接排名,其余显示百分比与名次。
  • 沉浸大厅保留标题栏、拖动区域和窗口按钮。

下载与验证

升级前建议导出备份。提供 Windows x64 Setup / portable、Linux x64 AppImage / deb、macOS Intel / Apple Silicon dmg,以及逐包 SHA-256 和总清单。Windows 日常使用建议 Setup。

发布流程要求逻辑回归、Windows 打包后界面检查、Linux/macOS 原生构建与启动检查、下载校验及 Windows 安装/便携冷启动通过后才公开 Release。估分测试使用合成历史数据验证实现,不是现实预测精度评估。

核心训练和估算无需 AI;个人数据留在本机,公开接口用于同步。安装包不含个人资料、密钥或第三方壁纸。macOS 使用 ad-hoc 签名,未经过 Apple Developer ID 公证。

English summary

Official Rating remains the default. An optional local estimate accumulates eligible virtual sessions with historical Rated evidence and feeds training targets, related recommendations, colors and analytics without changing official records. Daily Training now groups manual reference Rating, weak tags, review bounds and problem counts. Immersive mode keeps the titlebar visible; narrow cards and training layouts are improved.

Estimates form a separate virtual-only trajectory, not official Rating. Repeated contests and prior exposure contribute with warnings; missing evidence stays explicit, and sessions without visible submissions cannot be discovered. Back up before upgrading. Packages are gated by native builds, tests, checksums and Windows installation/cold-start verification.