# TWENTY-NO-TV (PREREG 20:31 IST 27-Oct-2026): ~20 premium that is almost all intrinsic (ITM by ~20). Real bars; spot>=15:15 missing, not filled.
import sys; sys.path.insert(0,'/workspace/research/popular-strategies/scripts')
from ec_lib import *
from multiprocessing import Pool
WN=[('W1',20,105),('W2',105,225),('W3',225,315),('W4',315,345)]
def wof(t): return next(w for w,a,b in WN if a<=t<=b)
SHL={}
for ix in ('NIFTY','SENSEX'):
    d=pd.read_parquet(f'{CD}/{ix.lower()}_1m_clean.parquet'); d['mi']=d.timestamp.dt.hour*60+d.timestamp.dt.minute-555
    for dt,g in d.groupby('date'):
        h=np.full(375,np.nan); l=np.full(375,np.nan); ok=(g.mi>=0)&(g.mi<360)&(~g.missing); h[g.mi[ok].values]=g.high[ok].values; l[g.mi[ok].values]=g.low[ok].values; SHL[(ix,dt)]=(h,l)
e=pd.read_parquet('/workspace/market-live-stack/bots/upstox/eod/2026-10-27/nifty_1m_2026-10-27.parquet'); ts=pd.to_datetime(e.timestamp); m=(ts.dt.hour*60+ts.dt.minute-555).values; k=m<360
h=np.full(375,np.nan); l=np.full(375,np.nan); h[m[k]]=e.high.values[k]; l[m[k]]=e.low.values[k]; SHL[('NIFTY',pd.Timestamp('2026-10-27'))]=(h,l)
def trendo(s,OPEN):
    tr=np.zeros(375,int)
    for t in range(20,346):
        if np.isnan(s[t]) or np.isnan(s[t-20]): continue
        tr[t]=1 if (s[t]>OPEN and s[t]-s[t-20]>0) else -1 if (s[t]<OPEN and s[t]-s[t-20]<0) else 0
    return tr
def outb(ix,dt,r,t,slip,R):
    f=None
    for y in range(t+1,min(t+6,375)):
        if not np.isnan(r[0,y]): f=y; break
    if f is None: return np.nan,None
    fill=r[0,f]+slip; last=max([y for y in range(f,375) if not np.isnan(r[3,y])]); peak=0; ex=None; why='close'
    for y in range(f,last+1):
        if np.isnan(r[2,y]): continue
        if r[2,y]<=fill-R: ex=fill-R-slip; why='stop'; break
        g=r[3,y]-fill; peak=max(peak,r[1,y]-fill if r[1,y]-fill>=2*R else g)
        if peak>=2*R and g<=0.5*peak: ex=r[3,y]-slip; why='trail'; break
    if ex is None: ex=r[3,last]-slip
    ex=max(ex,0.05); return ex-fill-chg(ix,dt,fill,[(1,ex)]),why
def fchg(ix,dt,e,x):
    Q=lot(ix,dt); stt=(0.0005 if dt>=pd.Timestamp('2026-04-01') else 0.0002)*x; tx=(0.0000173 if ix=='NIFTY' else 0.000005)*(e+x); st=0.00002*e; sebi=1e-6*(e+x); gst=0.18*(40/Q+tx+sebi)
    return 40/Q+stt+tx+st+sebi+gst
def spotctl(ix,dt,s,t,d,f,slip):
    h,l=SHL[(ix,dt)]; e=s[t]+d*slip; ys=[y for y in range(t+1,360) if not np.isnan(s[y])]
    if not ys: return dict(sa=np.nan,sb=np.nan,sc=np.nan)
    last=ys[-1]; Sx=DH[ix].close.get(dt,np.nan); out={}
    def G(p): return d*(p-e)
    # a'
    s1=s2=None; stop=False
    for y in ys:
        adv=G(l[y] if d>0 else h[y]) if not np.isnan(h[y]) else G(s[y]); fav=G(h[y] if d>0 else l[y]) if not np.isnan(h[y]) else G(s[y])
        if adv<=-20*f: g1=-20*f if s1 is None else s1; s1=g1; s2=-20*f; stop=True; break
        if s1 is None and fav>=20*f: s1=20*f
        if fav>=40*f: s2=40*f; break
    if s1 is None: s1=G(s[last])-slip
    if s2 is None: s2=G(s[last])-slip
    if stop and s1==-20*f: s1=-20*f-slip; s2=-20*f-slip
    out['sa']=0.5*(s1+s2)-fchg(ix,dt,e,e)
    # b'
    peak=0; ex=None
    for y in ys:
        adv=G(l[y] if d>0 else h[y]) if not np.isnan(h[y]) else G(s[y]); fav=G(h[y] if d>0 else l[y]) if not np.isnan(h[y]) else G(s[y]); g=G(s[y])
        if adv<=-10*f: ex=-10*f-slip; break
        peak=max(peak,fav if fav>=20*f else g)
        if peak>=20*f and g<=0.5*peak: ex=g-slip; break
    if ex is None: ex=G(s[last])-slip
    out['sb']=ex-fchg(ix,dt,e,e)
    # c'
    ex=None
    for y in ys:
        adv=G(l[y] if d>0 else h[y]) if not np.isnan(h[y]) else G(s[y])
        if adv<=-20*f: ex=-20*f-slip; break
    if ex is None: ex=G(Sx) if not np.isnan(Sx) else np.nan
    out['sc']=ex-fchg(ix,dt,e,e); return out
TVS={'tv2':lambda tv,c,f:-2*f<=tv<=2*f,'tv1':lambda tv,c,f:abs(tv)<=1*f,'tv10pct':lambda tv,c,f:abs(tv)<=0.1*c}
def day(fp):
    ix,ds=os.path.basename(fp)[:-8].split('_'); dt=pd.Timestamp(ds); s=SP.get((ix,dt))
    if s is None or (ix,dt) not in SHL: return [],[],[]
    OPEN=DH[ix].open.get(dt,np.nan); OPEN=s[0] if np.isnan(OPEN) else OPEN; tr=trendo(s,OPEN)
    A,src=load(ix,dt); slip=0.5 if ix=='NIFTY' else 1.5
    sc=DH['SENSEX'].open.get(dt,np.nan)/DH['NIFTY'].open.get(dt,np.nan) if ix=='SENSEX' else 1.0
    sets=[('raw',1.0)]+([('scaled',sc)] if ix=='SENSEX' and not np.isnan(sc) else [])
    EV=[];RN=[];NEG=[]
    for sn,f in sets:
        Lv=20*f; ST={}
        for (typ,K),r in A.items():
            c=r[3]; v=r[4]; first={k:None for k in TVS}; negseen=False
            for t in range(20,346):
                if np.isnan(c[t]) or np.isnan(s[t]): continue
                I=(s[t]-K) if typ=='CE' else (K-s[t])
                if not (0.9*Lv<=I<=1.1*Lv and 0.8*Lv<=c[t]<=1.2*Lv) or (not np.isnan(v[t]) and v[t]<=0): continue
                tv=c[t]-I
                if tv<-2*f and not negseen: negseen=True; NEG.append(dict(ix=ix,date=dt,set=sn,typ=typ,K=K,t=t,tv=tv))
                for k,fn in TVS.items():
                    if first[k] is None and fn(tv,c[t],f): first[k]=(t,tv,I)
                if TVS['tv2'](tv,c[t],f): ST.setdefault(typ,[]).append((t,K))
            for k,x in first.items():
                if x is None: continue
                t,tv,I=x; o=outcome(ix,dt,A,typ,K,t,slip,0.05*Lv)
                if o.get('status')!='ok': o.update(ix=ix,date=dt,set=sn,tvdef=k,typ=typ,K=K,t=t); EV.append(o); continue
                o['pnl_b'],o['why_b']=outb(ix,dt,r,t,slip,10*f); d=1 if typ=='CE' else -1
                lab=tr[t]*(1 if typ=='PE' else -1); mode='counter' if lab==1 else 'with' if lab==-1 else 'neutral'
                o.update(spotctl(ix,dt,s,t,d,f,slip)); o.update({'o'+k:v for k,v in spotctl(ix,dt,s,t,-d,f,slip).items()})
                o.update(ix=ix,date=dt,set=sn,tvdef=k,typ=typ,K=K,t=t,tv=tv,intr=I,spot=s[t],mode=mode,win=wof(t),src=src,f=f); EV.append(o)
        # R1 random: for primary trades (earliest per side), done in stats via these state lists -> simulate here
        prim={}
        for o in EV:
            if o['set']==sn and o['tvdef']=='tv2' and o.get('status')=='ok' and (o['typ'] not in prim or o['t']<prim[o['typ']]['t']): prim[o['typ']]=o
        for typ,o in prim.items():
            w,a,b=next(x for x in WN if x[0]==o['win']); cand=[(u,K) for u,K in ST.get(typ,[]) if a<=u<=b]
            if not cand: continue
            rg=np.random.default_rng(abs(hash((ds,ix,sn,typ)))%2**32)
            for rep in range(30):
                u,K=cand[int(rg.integers(len(cand)))]; q=outcome(ix,dt,A,typ,K,u,slip,0.05*Lv)
                if q.get('status')!='ok': continue
                q['pnl_b'],_=outb(ix,dt,A[(typ,K)],u,slip,10*f); q.update(ix=ix,date=dt,set=sn,typ=typ,rep=rep,mode=o['mode']); RN.append(q)
    return EV,RN,NEG
if __name__=='__main__':
    fps=sorted(glob.glob(f'{C}/*.parquet')); out=[[],[],[]]
    with Pool(7) as p:
        for i,res in enumerate(p.imap_unordered(day,fps)):
            for k in range(3): out[k]+=res[k]
            if i%30==0: print(i,len(out[0]),flush=True)
    for k,n in enumerate(['ntv_events','ntv_random','ntv_negtv']):
        D=pd.DataFrame(out[k])
        if len(D): D['split']=np.where(D.date<=pd.Timestamp('2025-12-31'),'train','test')
        D.to_csv(f'{L}/results/{n}.csv',index=False); print(n,len(D))