# Option-buyer trades on every candidate (ATM CE, nearest expiry and skip-0DTE) + random-minute benchmark. Real 1m bars, no filling.
import pandas as pd,numpy as np
R='/workspace/research/flow-forensics-2026-10/recov'; B='/workspace/research/flow-forensics-2026-10'
C=pd.read_csv(f'{R}/results/candidates_scored.csv',parse_dates=['date'])
OPT=pd.read_pickle(f'{R}/data/opt_panels.pkl')
P=pd.read_parquet(f'{B}/lows/data/minute_panel_v.parquet',columns=['date','hm','close','low'])
t=pd.read_csv(f'{R}/data/today/NIFTY.csv'); t['date']=pd.to_datetime(t.ts.str[:10]); t['hm']=t.ts.str[11:16]
P=pd.concat([P[P.date>='2024-10-21'],t[['date','hm','close','low']]]); SP={d:g.set_index('hm') for d,g in P.groupby('date')}
exps=sorted({e for _,e in OPT})
def nexp(d,skip0=False):
    c=[e for e in exps if e>=d and not(skip0 and e==d)]; return c[0] if c else None
SLIP=3.0
def lot(t): return 25 if t<pd.Timestamp('2024-11-20') else (75 if t<pd.Timestamp('2026-01-01') else 65)
def hmadd(h,m): return (pd.Timestamp('2000-01-01 '+h)+pd.Timedelta(minutes=m)).strftime('%H:%M')
def charges(t,buy,sells):
    Q=lot(t); n=1+len(sells); brk=20.0*n; sv=sum(f*p for f,p in sells)*Q; bv=buy*Q
    stt=0.001*sv; txn=0.0003503*(bv+sv); sebi=1e-6*(bv+sv); stamp=3e-5*bv; gst=0.18*(brk+txn+sebi)
    return (brk+stt+txn+sebi+stamp+gst)/Q
def trade(d,e,K,hm,style,RL=None,hold=60):
    if style=='runner': hold=240
    O=OPT.get((d,e))
    if O is None or ('CE',K) not in O['close'].columns: return None
    s=pd.DataFrame({c:O[c][('CE',K)] for c in ['open','high','low','close']}).dropna()
    h1=hmadd(hm,1)
    if h1 not in s.index: return None
    fill=s.open[h1]+SLIP; end=min(hmadd(hm,hold+1),'15:25'); path=s.loc[h1:end]
    if len(path)<5: return None
    sp=SP.get(d); rem=1.0; sells=[]; stop=fill-20 if style!='spotstop' else fill-40; peak=fill; t1=t2=False
    for i,(x,b) in enumerate(path.iterrows()):
        if style=='spotstop' and sp is not None and x in sp.index and sp.low.get(x,np.inf)<RL-5:
            sells.append((rem,b.close-SLIP)); rem=0; break          # exit at that bar's close when spot breaks the low
        if style=='runner' and sp is not None and x in sp.index:
            if sp.low.get(x,np.inf)<RL-5: sells.append((rem,b.close-SLIP)); rem=0; break      # structural stop: spot breaks the low
            hs=sp.loc[:x]; 
            if sp.close.get(x,-np.inf)>=RL+200 or x>='15:14': sells.append((rem,b.close-SLIP)); rem=0; break   # target: +200 from low, or 15:14
        if style=='runner':
            if b.low<=fill-35: px=min(b.open,fill-35) if i>0 else fill-35; sells.append((rem,px-SLIP)); rem=0; break
            continue
        if b.low<=stop:
            px=min(b.open,stop) if i>0 else stop; sells.append((rem,px-SLIP)); rem=0; break
        if style in('scale','spotstop'):
            if not t1 and b.high>=fill+20: sells.append((0.4,fill+20-SLIP)); rem-=0.4; t1=True; stop=max(stop,fill)
            if not t2 and b.high>=fill+40: sells.append((0.3,fill+40-SLIP)); rem-=0.3; t2=True
            peak=max(peak,b.high)
            if t2: stop=max(stop,peak-20,fill+20)
    if rem>1e-9: sells.append((rem,path.close.iloc[-1]-SLIP))
    g=sum(f*p for f,p in sells)-fill; c=charges(d,fill,sells)
    return dict(fill=fill,gross=g,cost=c,pnl=g-c,pnl_rs_lot=(g-c)*lot(d))
rng=np.random.default_rng(11)
days=sorted(C.date.unique()); alld=sorted(SP)
rnd=[]
for _ in range(3000):
    d=alld[rng.integers(len(alld))]; sp=SP[d]; hs=[h for h in sp.index if '09:20'<=h<='14:14' and sp.close.get(h)==sp.close.get(h)]
    if not hs: continue
    h=hs[rng.integers(len(hs))]; rnd.append(dict(date=d,hm=h,entry=sp.close[h],RL=sp.low.loc[:h].min(),kind='random'))
EV=pd.concat([C.assign(kind='candidate'),pd.DataFrame(rnd)],ignore_index=True)
out=[]
for r in EV.itertuples():
    K=float(round(r.entry/50)*50)
    for lab,e in (('nearest',nexp(r.date)),('skip0DTE',nexp(r.date,True))):
        if e is None: continue
        for st in ('scale','sl_only','spotstop','runner'):
            z=trade(r.date,e,K,r.hm,st,RL=r.RL)
            if z: out.append(dict(idx=r.Index,date=r.date,hm=r.hm,kind=r.kind,expiry_rule=lab,dte=(e-r.date).days,style=st,**z))
TR=pd.DataFrame(out); TR.to_csv(f'{R}/results/trades_all.csv',index=False)
print(TR.groupby(['kind','expiry_rule','style']).pnl.agg(['count','mean','median',lambda x:(x>0).mean()*100]).round(2).to_string())