import pandas as pd,numpy as np
P=pd.read_parquet('data/minute_panel.parquet')
P.loc[P.date<pd.Timestamp('2024-09-27'),['fvol','foi']]=np.nan   # front month (Sep-2024) not available before
n=P.groupby('date').close.count(); bad=n[n<300].index; print('non-standard sessions dropped',list(bad.date))
P=P[~P.date.isin(bad)]
V=P.pivot_table(index='date',columns='hm',values='fvol')
base=V.shift(1).rolling(20,min_periods=15).mean()
R=(V/base).stack().rename('fvr').reset_index()
P=P.merge(R,on=['date','hm'],how='left')
P.to_parquet('data/minute_panel_v.parquet'); print(P.fvr.describe()); print('sessions with fvr',P[P.fvr.notna()].date.nunique(),P[P.fvr.notna()].date.min())