import pandas as pd,numpy as np,matplotlib; matplotlib.use('Agg'); import matplotlib.pyplot as plt
M=pd.read_csv('results/lows_minute_windows.csv'); R=pd.read_csv('results/lows_summary.csv')
n=len(R); cols=4; rows=int(np.ceil(n/cols))
fig,axs=plt.subplots(rows,cols,figsize=(18,3.1*rows),sharex=True)
for ax,r in zip(axs.flat,R.itertuples()):
    g=M[M.low_id==r.low_id]
    ax2=ax.twinx(); ax2.bar(g.rel,g.fut_vr.clip(upper=12),color='tab:blue',alpha=.3,width=.9); ax2.set_ylim(0,24); ax2.axhline(3,color='tab:blue',lw=.6,ls=':')
    ax.plot(g.rel,g.nifty-r.low,color='k',lw=1.4,label='Nifty − low'); ax.plot(g.rel,g.fut-g.fut[g.rel==0].values[0],color='tab:orange',lw=.9,label='Fut − fut@low')
    ax.plot(g.rel,(g.pe-g.pe[g.rel==0].values[0]),color='tab:red',lw=.8,ls='--',label='ATM PE Δ'); ax.plot(g.rel,(g.ce-g.ce[g.rel==0].values[0]),color='tab:green',lw=.8,ls='--',label='ATM CE Δ')
    ax.axvline(0,color='grey',lw=.6); ax.axhline(0,color='grey',lw=.5)
    ax.set_title(f"{r.date[5:]} {r.low_hm}  fall {r.fall:.0f}  +30m {r.rec30:+.0f}\n{r.classification[:46]}",fontsize=8.5)
    ax.tick_params(labelsize=7); ax2.tick_params(labelsize=6)
for ax in axs.flat[n:]: ax.axis('off')
axs.flat[0].legend(fontsize=6,loc='upper left')
fig.text(.5,.005,'minutes from the low (−15 … +45). Bars (right axis) = Nifty futures volume ÷ 20-day same-minute average (dotted = 3x). Spot after 15:14 missing.',ha='center',fontsize=9)
plt.tight_layout(rect=(0,.02,1,1)); plt.savefig('charts/lows_windows.png',dpi=95); plt.close()
# part 2 hit rates
Z=pd.read_csv('results/p2_config_stats.csv'); C=pd.read_csv('results/p2_outcomes_controls.csv')
fig,ax=plt.subplots(1,2,figsize=(14,4))
for k,(m,lab) in enumerate([('r40_30','P(+40 pts within 30 min) %'),('hold60','P(low not broken in next 60 min) %')]):
    xs=[];vals=[];labs=[]
    for X in [150,200,250]:
        for kind,col in [('signal','tab:blue'),('control_quiet','tab:grey'),('at_low_any','tab:orange')]:
            if kind=='signal':
                v=Z[(Z.X==X)&(Z.V==3)&(Z.win=='3bar')&(Z.oi=='any')&(Z.period=='all')][m if m!='r40_30' else 'r40_30'].values[0] if m=='r40_30' else Z[(Z.X==X)&(Z.V==3)&(Z.win=='3bar')&(Z.oi=='any')&(Z.period=='all')]['hold60'].values[0]
            else:
                c=C[(C.kind==kind)&(C.X==X)]; v=(c[m]*c.n).sum()/c.n.sum()
            xs.append(len(xs)+ (X-150)//50); vals.append(v); labs.append(col)
    ax[k].bar(xs,vals,color=labs); ax[k].set_xticks([1,5,9]); ax[k].set_xticklabels(['fall ≥150','fall ≥200','fall ≥250']); ax[k].set_title(lab,fontsize=10)
    for x,v in zip(xs,vals): ax[k].text(x,v+.5,f'{v:.0f}',ha='center',fontsize=8)
from matplotlib.patches import Patch
ax[0].legend(handles=[Patch(color='tab:blue',label='signal: 3-bar fut vol ≥3x + up bar at low'),Patch(color='tab:grey',label='control: same bar, fut vol <1.5x'),Patch(color='tab:orange',label='any first touch of low')],fontsize=7)
plt.suptitle('Oct-2024 → Oct-2026, first event per day (futures-volume era)',fontsize=10); plt.tight_layout(); plt.savefig('charts/p2_hitrates.png',dpi=95); plt.close()
T=pd.read_csv('results/p2_option_trades.csv'); T=T[(T.expiry_rule=='nearest')&(T['style']=='scale')]
s=T[(T.kind=='signal')&(T.X==150)&(T.V==3)&(T.win=='3bar')&(T.oi=='any')].pnl; r=T[T.kind=='random'].pnl; c=T[(T.kind=='control_quiet')&(T.X==150)].pnl
fig,ax=plt.subplots(1,2,figsize=(14,4))
b=np.arange(-30,80,4)
ax[0].hist(r.clip(-30,78),bins=b,density=True,alpha=.5,label=f'random entries (n={len(r)}, avg {r.mean():.1f})',color='grey')
ax[0].hist(s.clip(-30,78),bins=b,density=True,alpha=.6,label=f'signal X150/3x/3bar (n={len(s)}, avg {s.mean():.1f})',color='tab:blue'); ax[0].legend(fontsize=8); ax[0].set_title('ATM CE P&L per trade, pts after costs (SL 20, scale-out)',fontsize=10)
for lab,x,col in [('signal X150/3x/3bar',T[(T.kind=='signal')&(T.X==150)&(T.V==3)&(T.win=='3bar')&(T.oi=='any')],'tab:blue'),('control quiet X150',T[(T.kind=='control_quiet')&(T.X==150)],'grey'),('signal X200/3x/3bar',T[(T.kind=='signal')&(T.X==200)&(T.V==3)&(T.win=='3bar')&(T.oi=='any')],'tab:green')]:
    x=x.sort_values('date'); ax[1].plot(pd.to_datetime(x.date),x.pnl.cumsum(),label=lab,color=col)
ax[1].axhline(0,color='k',lw=.5); ax[1].legend(fontsize=8); ax[1].set_title('cumulative pts per 1 unit (multiply by lot size for ₹)',fontsize=10)
plt.tight_layout(); plt.savefig('charts/p2_option_pnl.png',dpi=95); plt.close(); print('ok')