hltv-cs2-deep-analysis

name	hltv-cs2-deep-analysis
description	Download and analyze detailed HLTV CS2 match-map data for any team or player, then generate both Markdown and native HTML reports in Chinese. Use when users ask for deep diagnosis (weak maps, strong maps, opponent-map matchup issues) and transparent missing-data handling.
metadata	{"openclaw":{"requires":{"bins":"[Truncated]","env":"[Truncated]"}}}

HLTV CS2 Deep Analysis

Use this skill for end-to-end analysis: collect detailed rows online, compute deep indicators, handle missing values transparently, and output Chinese dual-format reports.

Quick Start (One-liner)

curl -fsSL https://raw.githubusercontent.com/ulinwang/hltv-cs2-deep-analysis/main/scripts/install_deps.sh | bash

Or clone and run locally:

git clone https://github.com/ulinwang/hltv-cs2-deep-analysis.git
cd hltv-cs2-deep-analysis
./scripts/install_deps.sh

Workflow

0. Install dependencies (recommended once, includes report + Playwright runtime check).

skills/hltv-cs2-deep-analysis/scripts/install_deps.sh

若不安装，pipeline 会自动回退生成 quick report（stdlib 版），并给出安装提示。 注意：采集依赖 node/npm/npx 与 @playwright/cli（通过 npx 拉取）。

1. Collect detailed rows online (real browser, not static HTML).

python3 skills/hltv-cs2-deep-analysis/scripts/collect_hltv_detailed.py \
  --subject-type team \
  --subject-id 11283 \
  --subject-slug falcons \
  --subject-label "Falcons (kyousuke era)" \
  --start-date 2025-06-23 \
  --end-date 2026-02-20 \
  --max-rows 30 \
  --max-pages 5 \
  --output-csv output/hltv/falcons_deep/detailed_maps.csv

默认建议 --headed --persistent（有头 + 持久上下文），更适合 HLTV。 脚本会自动翻页抓取（直到无下一页或达到 --max-pages）。

2. Build deep reports (Markdown + HTML).

python3 skills/hltv-cs2-deep-analysis/scripts/build_deep_report.py \
  --input-csv output/hltv/falcons_deep/detailed_maps.csv \
  --subject-label "Falcons (kyousuke era)" \
  --output-md output/hltv/falcons_deep/report.md \
  --output-html output/hltv/falcons_deep/report.html \
  --template-html skills/hltv-cs2-deep-analysis/assets/report_template.html

如需仅分析 player 的某支队伍阶段（转会期场景），可追加：

--player-team-filter PARIVISION

3. Run one-command pipeline (online path).

skills/hltv-cs2-deep-analysis/scripts/run_deep_analysis_pipeline.sh \
  team 11283 falcons "Falcons (kyousuke era)" \
  output/hltv/falcons_deep 2025-06-23 2026-02-20 30

传递采集/报告额外参数（推荐）：

HLTV_COLLECTOR_EXTRA_ARGS="--headed --persistent --max-pages 8 --nav-timeout 180 --cli-timeout 180 --retries 3 --retry-delay 1.5" \
HLTV_REPORT_EXTRA_ARGS="" \
skills/hltv-cs2-deep-analysis/scripts/run_deep_analysis_pipeline.sh \
  team 11283 falcons "Falcons (kyousuke era)" \
  output/hltv/falcons_deep 2025-06-23 2026-02-20 30

player 转会期样例（仅保留 Jame 在 PARIVISION 的样本）：

HLTV_COLLECTOR_EXTRA_ARGS="--headed --persistent --max-pages 12 --player-team-filter PARIVISION" \
HLTV_REPORT_EXTRA_ARGS="--player-team-filter PARIVISION" \
skills/hltv-cs2-deep-analysis/scripts/run_deep_analysis_pipeline.sh \
  player 13776 jame "Jame (PARIVISION period)" \
  output/hltv/jame_parivision 2025-01-13 2026-03-02 0

4. Collectable advanced fields per map.

• Team-level: team_rating_3_subject/opponent, first_kills_subject/opponent, clutches_won_subject/opponent.
• Player-derived aggregates (subject side): rating, adr, kast, swing, opening_duel_diff, clutches_won.
• Player spotlight: best_player, best_player_rating, best_player_adr.
• Player row context: player_team（用于 player 多队时期过滤与归属校准）。

5. Missing-data handling (built into report script).

• Report script applies hierarchical imputation for advanced metrics: 地图中位数 -> 对手中位数 -> 全局中位数（仅用于分析）。
• Report outputs a dedicated quality section:
  • total missing before/after
  • per-column fill counts
• If missing remains after fill, report will still generate and expose residual missing.

6. Analysis focus.

• 哪些图打得不好/好：胜率、净回合差、rating、ADR、KAST、开局对枪、首杀差、残局差。
• 对手为什么打得好：通过 team_rating_3、首杀与残局等对抗指标解释。
• 打哪些队的什么图不好/好：对手-地图组合层面的样本诊断与原因归纳。
• 自动生成“可执行建议”列表：按 P1/P2/P3 优先级输出执行动作、量化目标、覆盖范围。

Output Expectations

• detailed_maps.csv: online-collected map rows with advanced metrics and source links.
• report.md: 中文深度诊断报告（弱图/强图、对手-地图组合、原因归因）。
• report.html: 中文原生 HTML 报告（可视化看板 + 诊断表格 + 可执行建议）。

Resources

• scripts/collect_hltv_detailed.py: data collection from HLTV team/player stats pages.
• scripts/build_deep_report.py: deep indicator computation and dual-format report generation.
• scripts/run_deep_analysis_pipeline.sh: one-command pipeline.
• references/schema-and-sources.md: schema and data-quality rules.
• assets/report_template.html: native HTML template.