Football Prediction Optimized

使用优化后的提示词对足球比赛进行预测，支持批量筛选(stage1)和单场比赛预测(stage2)

Security Scan

The skill appears to perform the football prediction work it claims, but the package metadata omits the required DeepSeek API credential and the code will send input data to an external API (configurable via DEEPSEEK_BASE_URL), which is an important mismatch the user should review before installing.

Purpose & Capability

The name, description, SKILL.md and the included Python code are coherent: the skill does exactly football match selection and prediction and uses an LLM-like DeepSeek service to produce JSON outputs. There are no unrelated binaries or surprising dependencies. However, the registry metadata claimed no required environment variables while both SKILL.md and the code require DEEPSEEK_API_KEY (and optionally DEEPSEEK_BASE_URL). This metadata omission is inconsistent and worth flagging.

Instruction Scope

The runtime instructions and code confine activity to: reading provided match feature text files, constructing prompts, calling the DeepSeek API, and returning structured JSON. The skill does not attempt to read unrelated system files or extra environment variables. It will transmit input match feature text to an external service (DeepSeek) which is expected for an API-based model call, but users should be aware this is external network I/O.

Install Mechanism

There is no install spec that downloads arbitrary artifacts. The bundle includes only a Python script and a requirements.txt with 'requests'. No high-risk install behavior (no remote archives, no obscure download URLs) was found.

Credentials

The skill legitimately requires one service credential (DEEPSEEK_API_KEY) and optionally DEEPSEEK_BASE_URL; that is proportional to calling an external model API. The problem: the registry metadata listed 'Required env vars: none' while SKILL.md and the code require DEEPSEEK_API_KEY and will abort if it's missing. Additionally, DEEPSEEK_BASE_URL is user-overridable — changing it could redirect data to a different endpoint. This mismatch and the ability to override the endpoint are security considerations.

Persistence & Privilege

The skill does not request 'always: true', does not modify other skills or system-wide config, and has normal runtime behavior. It will make outbound API calls when invoked but does not persistently elevate privileges.

What to consider before installing

Key points before installing or running this skill: - Metadata mismatch: the registry says no env vars are required, but both SKILL.md and football_prediction.py require DEEPSEEK_API_KEY (and optionally DEEPSEEK_BASE_URL). Expect to provide an API key for the external DeepSeek service; the skill will refuse to run without it. - External network calls: the script sends the supplied match feature text to the DeepSeek API. Only send non-sensitive data. If you have privacy concerns about match data or any contextual data you include in the fet_txt, do not use this skill or sanitize inputs first. - Endpoint override risk: DEEPSEEK_BASE_URL can be set to a custom URL. If you set it (or if a malicious actor controls that env var in your environment), input data could be sent to a non-DeepSeek endpoint. Prefer leaving it at the default unless you trust the alternate endpoint. - Verify the API key handling: the code uses the DEEPSEEK_API_KEY directly in Authorization headers. Treat that key as sensitive, restrict its scope/permissions if possible, and rotate it after testing. - Review and test locally: because the package contains executable code, inspect the Python file locally and run it in a controlled environment (with a throwaway API key or network restrictions) before granting it access to production secrets or broad agent invocation. - Ask the publisher: the skill lacks a homepage and the registry owner info is minimal. If you need higher assurance, ask the publisher for provenance of the DeepSeek integration and why the metadata omitted required env vars. Given these points, the skill's functionality is coherent with its purpose, but the metadata omission and the ability to redirect the API endpoint are non-trivial concerns — treat the skill as suspicious until you verify the API key and endpoint settings and are comfortable with external data transmission.

足球预测优化技能

这个技能使用经过自我优化系统优化的最佳提示词，对足球比赛进行专业预测。包含两个阶段：

1. Stage1 批量筛选：对同一比赛日的多场比赛进行横向比较，筛选出最适合预测的高质量场次。
2. Stage2 单场比赛预测：对单场比赛进行深入分析，预测胜平负结果，并计算CLV收益预估。

功能特性

Stage1 批量筛选

• 多维评估：从数据完整性、市场清晰度、基本面优势、预测潜力、风险等级5个维度评分
• 智能筛选：根据综合评分自动筛选评分>6.0的比赛，数量控制在3-8场
• 结构化输出：JSON格式的筛选结果，包含详细评分和筛选理由

Stage2 单场比赛预测

• 综合分析：结合基本面分析、市场分析、数据质量评估、风险因素识别
• CLV收益预估：预测主胜、平局、客胜三个选项的CLV值（收盘赔率变化值）
• 置信度评估：提供0-1的置信度评分
• 风险提示：列出主要风险因素

使用方法

方法1：通过OpenClaw技能调用

当用户需要足球比赛预测时，技能会自动触发：

• "批量筛选今天的所有比赛"
• "预测这场比赛的结果"
• "分析这场比赛并给出CLV预估"

方法2：直接使用Python脚本

# Stage1 批量筛选
python3 /Users/cjy/.openclaw/workspace/skills/football-prediction-optimized/football_prediction.py stage1 --input batch_fet_txt.txt

# Stage2 单场比赛预测
python3 /Users/cjy/.openclaw/workspace/skills/football-prediction-optimized/football_prediction.py stage2 --input single_fet_txt.txt

方法3：与lota-football技能配合使用

# 1. 使用lota-football获取比赛特征文本
# 2. 将特征文本保存到文件
# 3. 使用本技能进行预测
python3 football_prediction.py stage2 --input fet_txt.txt

输入格式

Stage1 输入

多场比赛的特征文本，每场比赛的特征文本用===分隔：

{第一场比赛的特征文本}
===
{第二场比赛的特征文本}
===
...

Stage2 输入

单场比赛的特征文本（完整fet_txt格式）

输出格式

Stage1 输出

{
  "date": "比赛日期",
  "total_matches": 总比赛数量,
  "selected_matches": [
    {
      "lota_id": "比赛唯一ID",
      "home_team": "主队名称",
      "away_team": "客队名称",
      "league": "所属联赛",
      "score": 综合评分,
      "data_completeness": 数据完整性得分,
      "market_clarity": 市场清晰度得分,
      "fundamental_strength": 基本面优势得分,
      "prediction_potential": 预测潜力得分,
      "risk_level": 风险等级得分,
      "decision": "KEEP",
      "reason": "筛选理由"
    }
  ],
  "summary": {
    "total_selected": 选中比赛数量,
    "average_score": 平均综合评分,
    "selection_rate": 选中比例
  }
}

Stage2 输出

{
  "lota_id": "比赛唯一ID",
  "prediction": "H/D/A",
  "confidence": 0.85,
  "reasoning": [
    "理由1：...",
    "理由2：...",
    "理由3：..."
  ],
  "clv_estimates": {
    "home": 0.02,
    "draw": -0.01,
    "away": 0.03
  },
  "risk_factors": [
    "风险因素1",
    "风险因素2"
  ],
  "timestamp": "2026-03-28 10:30:00"
}

环境变量

• DEEPSEEK_API_KEY: DeepSeek API密钥（必需）
• DEEPSEEK_BASE_URL: DeepSeek API基础URL（可选，默认：https://api.deepseek.com）

技术实现

1. 模型调用：使用DeepSeek API（deepseek-chat模型）
2. 提示词优化：使用自我优化系统得到的最佳提示词
3. 错误处理：网络重试、格式验证、异常捕获
4. 性能优化：异步调用、缓存机制

与现有技能集成

与lota-football技能集成

1. 使用lota-football技能获取比赛列表和特征文本
2. 将特征文本传递给本技能进行预测
3. 结合两个技能的结果进行综合决策

与odds-plotter技能集成

1. 使用odds-plotter技能绘制赔率走势图
2. 使用本技能进行预测分析
3. 结合图表和预测结果进行深度分析

示例用例

用例1：每日比赛筛选

用户：筛选今天所有竞彩比赛，推荐3-8场高质量比赛
流程：
1. lota-football获取今日比赛特征文本
2. Stage1批量筛选
3. 输出推荐比赛列表

用例2：单场比赛深度分析

用户：深度分析曼联对切尔西这场比赛
流程：
1. lota-football获取该场比赛特征文本
2. Stage2单场比赛预测
3. 输出预测结果和CLV预估

用例3：批量预测与筛选

用户：批量预测今天所有比赛，并筛选出最有价值的比赛
流程：
1. lota-football获取今日所有比赛特征文本
2. Stage2逐个预测每场比赛
3. 根据置信度和CLV值进行排序筛选

错误处理

1. API调用失败：自动重试3次，使用指数退避
2. 输入格式错误：提示正确的输入格式示例
3. 模型输出格式错误：尝试解析或使用备用提示词
4. 环境变量缺失：提示设置必要的环境变量

性能考虑

• 批量处理时限制并发数，避免API限制
• 使用本地缓存存储频繁查询的比赛数据
• 支持断点续传，避免重复处理

更新日志

v1.0.0 (2026-03-28)

• 初始版本，基于自我优化系统的最佳提示词
• 支持Stage1批量筛选和Stage2单场比赛预测
• 与lota-football技能无缝集成