From 60d3fa545847ebbf7bcbec3df776370861c98d2a Mon Sep 17 00:00:00 2001
From: MLingyu123 <125688718+MLingyu123@users.noreply.github.com>
Date: Fri, 1 May 2026 00:37:41 +0800
Subject: [PATCH] Add Futu script for NVDA selloff attribution

---
 futu_nvda_analysis.py | 158 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 158 insertions(+)
 create mode 100644 futu_nvda_analysis.py

diff --git a/futu_nvda_analysis.py b/futu_nvda_analysis.py
new file mode 100644
index 0000000..f6f33f6
--- /dev/null
+++ b/futu_nvda_analysis.py
@@ -0,0 +1,158 @@
+#!/usr/bin/env python3
+"""
+使用 Futu OpenD 拉取 NVDA 实时/分时数据，并做简单“今晚暴跌归因”分析。
+
+依赖：
+  pip install futu pandas
+
+使用前：
+  1) 启动 Futu OpenD（默认 127.0.0.1:11111）
+  2) 在 OpenD 中允许本机连接与对应市场权限
+
+示例：
+  python futu_nvda_analysis.py --date 2026-04-30
+"""
+
+from __future__ import annotations
+
+import argparse
+from dataclasses import dataclass
+from datetime import datetime
+from typing import Dict, List
+
+import pandas as pd
+from futu import OpenQuoteContext, RET_OK, SubType
+
+
+@dataclass
+class SymbolSnapshot:
+    code: str
+    name: str
+    last_price: float
+    open_price: float
+    high_price: float
+    low_price: float
+    prev_close_price: float
+    volume: float
+    turnover: float
+
+    @property
+    def pct_change(self) -> float:
+        if self.prev_close_price == 0:
+            return 0.0
+        return (self.last_price - self.prev_close_price) / self.prev_close_price * 100
+
+
+def get_snapshot(quote_ctx: OpenQuoteContext, codes: List[str]) -> Dict[str, SymbolSnapshot]:
+    ret, df = quote_ctx.get_market_snapshot(codes)
+    if ret != RET_OK:
+        raise RuntimeError(f"get_market_snapshot failed: {df}")
+
+    out: Dict[str, SymbolSnapshot] = {}
+    for _, r in df.iterrows():
+        out[r["code"]] = SymbolSnapshot(
+            code=r["code"],
+            name=r["name"],
+            last_price=float(r["last_price"]),
+            open_price=float(r["open_price"]),
+            high_price=float(r["high_price"]),
+            low_price=float(r["low_price"]),
+            prev_close_price=float(r["prev_close_price"]),
+            volume=float(r["volume"]),
+            turnover=float(r["turnover"]),
+        )
+    return out
+
+
+def get_intraday_bars(quote_ctx: OpenQuoteContext, code: str) -> pd.DataFrame:
+    ret, data = quote_ctx.get_cur_kline(code, num=390, ktype="K_1M", autype="qfq")
+    if ret != RET_OK:
+        raise RuntimeError(f"get_cur_kline failed for {code}: {data}")
+    return data.copy()
+
+
+def detect_selloff_window(kline_df: pd.DataFrame, top_n: int = 5) -> pd.DataFrame:
+    df = kline_df.copy()
+    df["pct_1m"] = (df["close"] / df["close"].shift(1) - 1) * 100
+    drops = df.nsmallest(top_n, "pct_1m")[["time_key", "open", "close", "volume", "pct_1m"]]
+    return drops.reset_index(drop=True)
+
+
+def analyze_cross_section(snaps: Dict[str, SymbolSnapshot], target: str = "US.NVDA") -> str:
+    target_snap = snaps[target]
+    peer_codes = [c for c in snaps if c != target]
+
+    peer_moves = [snaps[c].pct_change for c in peer_codes]
+    peer_avg = sum(peer_moves) / len(peer_moves) if peer_moves else 0.0
+    diff = target_snap.pct_change - peer_avg
+
+    if target_snap.pct_change < -3 and diff < -2:
+        style = "个股主导抛压（更像仓位/估值/事件驱动）"
+    elif target_snap.pct_change < 0 and peer_avg < 0:
+        style = "板块与大盘共振下跌（系统性风险偏好回落）"
+    else:
+        style = "波动较中性，需结合消息面进一步确认"
+
+    return (
+        f"NVDA当日涨跌幅: {target_snap.pct_change:.2f}%\n"
+        f"对照组平均涨跌幅: {peer_avg:.2f}%\n"
+        f"相对差值(NVDA-对照): {diff:.2f}pct\n"
+        f"归因判断: {style}"
+    )
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--host", default="127.0.0.1")
+    parser.add_argument("--port", type=int, default=11111)
+    parser.add_argument("--date", default=datetime.utcnow().strftime("%Y-%m-%d"))
+    parser.add_argument(
+        "--benchmarks",
+        nargs="*",
+        default=["US.AMD", "US.SMH", "US.QQQ"],
+        help="对照标的，默认 AMD/SMH/QQQ",
+    )
+    parser.add_argument("--output_csv", default="nvda_intraday_1m.csv")
+    args = parser.parse_args()
+
+    target = "US.NVDA"
+    codes = [target] + args.benchmarks
+
+    quote_ctx = OpenQuoteContext(host=args.host, port=args.port)
+    try:
+        ret, msg = quote_ctx.subscribe([target], [SubType.K_1M], is_first_push=False)
+        if ret != RET_OK:
+            raise RuntimeError(f"subscribe failed: {msg}")
+
+        snaps = get_snapshot(quote_ctx, codes)
+        nvda_kline = get_intraday_bars(quote_ctx, target)
+
+        nvda_kline.to_csv(args.output_csv, index=False)
+        selloff_windows = detect_selloff_window(nvda_kline, top_n=5)
+
+        print("=" * 72)
+        print(f"日期: {args.date} | 标的: {target}")
+        print("=" * 72)
+
+        print("\n[1] 实时快照")
+        for code in codes:
+            s = snaps[code]
+            print(
+                f"{code:8s} {s.name:15s} 最新 {s.last_price:8.2f} | "
+                f"开 {s.open_price:8.2f} 高 {s.high_price:8.2f} 低 {s.low_price:8.2f} | "
+                f"涨跌 {s.pct_change:6.2f}%"
+            )
+
+        print("\n[2] NVDA 1分钟最大跌幅时段（Top5）")
+        print(selloff_windows.to_string(index=False))
+
+        print("\n[3] 自动归因")
+        print(analyze_cross_section(snaps, target=target))
+
+        print(f"\n已导出1分钟K线到: {args.output_csv}")
+    finally:
+        quote_ctx.close()
+
+
+if __name__ == "__main__":
+    main()