zlt
/
akg-factor-bridge
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

复盘汇总加样本加权三列（ret_w/excess_w/beat_w），与方案第一节手工读数同口径对账；主口径仍按计划日等权 

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
zlt 2026-09-02 17:03:33 +08:00
parent cb819198ea
commit ff2a201aff
1 changed files with 11 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
plan_review.py
View File

@ -262,10 +262,17 @@ def run(since: str, until: str | None, horizons: tuple, start: str, out_dir: str
# 汇总：按名单 × 期限（全部）；按分组 × 期限；按事后环境 × 名单（只作解释）
def agg(g: pd.DataFrame) -> pd.Series:
return pd.Series({"days": g["date"].nunique(), "n": int(g["n"].sum()),
# 主口径按日等权：每个计划日先算名单均值，再跨日平均——一份名单一天一票；
# 带 _w 的三列按样本加权（大名单的日子权重大），与方案 1.8b/1.8c 手工读数同口径，只作对账。
n = g["n"].astype(float)
w = n / n.sum() if n.sum() else n
return pd.Series({"days": g["date"].nunique(), "n": int(n.sum()),
"ret": round(g["ret"].mean(), 2),
"excess_all": round(g["excess_all"].mean(), 2),
"beat": round(g["beat"].mean(), 1)})
"beat": round(g["beat"].mean(), 1),
"ret_w": round(float((g["ret"] * w).sum()), 2),
"excess_w": round(float((g["excess_all"] * w).sum()), 2),
"beat_w": round(float((g["beat"] * w).sum()), 1)})
lists_tbl = df[df["group"] == "全部"].groupby(["list", "h"]).apply(agg).reset_index()
groups_tbl = df[df["group"] != "全部"].groupby(["list", "group", "h"]).apply(agg).reset_index()
regime_tbl = df[df["group"] == "全部"].groupby(["regime_post", "list", "h"]).apply(agg).reset_index()
@ -275,7 +282,8 @@ def run(since: str, until: str | None, horizons: tuple, start: str, out_dir: str
md = [f"# 候选单复盘 · {days_plan[0]} 至 {stamp}（起点价 {start}）", "",
f"计划日 {len(days_plan)} 期；样本纪律：分组样本少于一百只或覆盖计划日少于二十个只看方向；"
f"口径切换的总样本少于六十个计划日只写方向。", "",
"## 一、六份名单 × 期限（超额=相对全池等权，命中=跑赢全池比例）", "",
"## 一、六份名单 × 期限（超额=相对全池等权，命中=跑赢全池比例；"
"不带后缀的列按计划日等权，带 _w 的列按样本加权，后者只用于与方案第一节手工读数对账）", "",
_md(lists_tbl), "",
"## 二、分组读数", "", _md(groups_tbl), "",
"## 三、按事后环境分组（未来 h 日全池涨跌，只作解释，不作交易前置）", "",