| {rank} | "
f"{html.escape(filename)} | "
f"{html.escape(str(row.get('ea_name','—')))} | "
f"{html.escape(str(row.get('symbol','—')))} | "
f"{html.escape(str(row.get('period','—')))} | "
f"{row['net_profit']:+.2f} | "
f"{row['profit_factor']:.2f} | "
f"{row['win_rate']:.1f}% | "
f"{row['max_dd_pct']:.0f}% | "
f"{row['sharpe']:.2f} | "
f"{row['n_trades']} | "
)
return (
""
"| 排名 | 报告 | EA | 品种 | 周期 | "
"净盈利 | PF | 胜率 | 回撤% | 夏普 | 笔数 | "
"
" + rows + "
"
)
# =========================================================================== #
# HTML 报告
# =========================================================================== #
def build_html(df: pd.DataFrame, anomalies: List[Dict[str, Any]]) -> str:
n = len(df)
pf_stats = {
"mean": float(df["profit_factor"].mean()),
"median": float(df["profit_factor"].median()),
"std": float(df["profit_factor"].std()),
"min": float(df["profit_factor"].min()),
"max": float(df["profit_factor"].max()),
}
net_stats = {
"mean": float(df["net_profit"].mean()),
"std": float(df["net_profit"].std()),
}
# 排序选择：PF 降序（默认）、净盈利降序、胜率降序
pf_rank = build_ranking_table(df, "profit_factor", ascending=False)
net_rank = build_ranking_table(df, "net_profit", ascending=False)
# 分布图
hist_svg = svg_hist([
("PF", df["profit_factor"].to_numpy()),
("净盈利", df["net_profit"].to_numpy()),
("胜率", df["win_rate"].to_numpy()),
])
scatter_svg = svg_scatter(df, "profit_factor", "win_rate", "PF vs 胜率")
# 异常告警
anomaly_html = ""
if anomalies:
anomaly_rows = ""
for a in anomalies:
color = "#fee2e2" if a["type"] in ("过拟合嫌疑", "回撤过大") else "#fef3c7"
fname = str(a.get("filename", ""))
anomaly_rows += (f"| {html.escape(a['type'])} | "
f"{html.escape(fname)} | "
f"{html.escape(a['detail'])} |
")
anomaly_html = (
"异常检测报告
"
"| 类型 | 报告 | 详情 |
|---|
"
+ anomaly_rows + "
"
)
return f"""
批量回测汇总报告
批量回测汇总报告
共解析 {n} 份报告 · {pd.Timestamp.now().strftime('%Y-%m-%d %H:%M')}
PF 均值
{pf_stats['mean']:.2f}
PF 中位数
{pf_stats['median']:.2f}
PF 标准差
{pf_stats['std']:.2f}
PF 范围
{pf_stats['min']:.2f}~{pf_stats['max']:.2f}
净盈利均值
{net_stats['mean']:+.2f}
1. 按 PF 排名
{pf_rank}
2. 按净盈利排名
{net_rank}
3. 指标分布
{anomaly_html}
使用说明：点击报告名称可跳转至单份详细分析（需配合 run_analysis.py）。
异常检测标记了可能过拟合或统计不显著的报告，请优先复核。
"""
# =========================================================================== #
# 单份报告 HTML（辅助跳转）
# =========================================================================== #
def write_single_htmls(
reports: List[Tuple[str, MT5Report]],
base_dir: str = OUT_DIR,
) -> None:
"""
为每份报告生成单份 mini 报告（可被主报告链接），避免生成 1MB+ 大文件。
仅在批量报告中提供索引跳转，不重复生成完整 HTML。
此处仅输出文件名索引信息供参考。
"""
pass
# =========================================================================== #
# CLI
# =========================================================================== #
def main(argv: List[str]) -> int:
ap = argparse.ArgumentParser(description="大批量报告汇总分析")
ap.add_argument("paths", nargs="+", help="报告目录或 glob 路径")
ap.add_argument("--pattern", default="*.xlsx", help="glob 模式（默认 *.xlsx）")
args = ap.parse_args(argv)
os.makedirs(OUT_DIR, exist_ok=True)
files = scan_reports(args.paths, args.pattern)
if not files:
print(f"未找到任何报告文件，路径: {args.paths}")
return 1
print(f"找到 {len(files)} 份报告，开始解析...")
reports = parse_batch(files)
if not reports:
print("所有文件解析失败，退出。")
return 1
df = build_summary_df(reports)
anomalies = detect_anomalies(df)
print(f"\n汇总统计:")
print(f" PF 均值={df['profit_factor'].mean():.2f} 中位数={df['profit_factor'].median():.2f}")
print(f" 净盈利均值={df['net_profit'].mean():+.2f}")
print(f" 异常报告: {len(anomalies)} 份")
html_doc = build_html(df, anomalies)
with open(BATCH_HTML_PATH, "w", encoding="utf-8") as f:
f.write(html_doc)
print(f"\n批量汇总报告已生成: {BATCH_HTML_PATH}")
return 0
if __name__ == "__main__":
sys.exit(main(sys.argv[1:]))