""" uj_diagnostic.py ================ Answers Bennett's specific doubts about the USDJPY leg: 1. WHERE does UJ make its money? P&L broken down by exit reason (tp / stop / hold_expiry / zscore). If profit is TP-driven, structure works. If it leans on hold-expiry or a few outliers, the edge is shaky. 2. THE EXPIRY PROBLEM: what % of UJ trades actually hit TP within 24h vs time out at hold-expiry? Bennett suspects the 0.70% target is too far for a 24h window. We measure it exactly. 3. OUTLIER DEPENDENCE: how concentrated is UJ's profit? Top-10 trades' share of total P&L. If a handful of trades carry it, edge is fragile. 4. INTERVENTION SENSITIVITY: UJ P&L by year, flagging known BoJ intervention periods (2022 H2, 2024, 2026). If losses concentrate there, structural risk. 5. DOES UJ EARN ITS PLACE? Basic stats: win rate, profit factor, avg win/loss, expectancy per trade, total P&L. Compare to make an EU-only decision. Reads usdjpy_trades_real_costs.csv. Columns (confirmed earlier): entry_time, exit_reason, entry_price, exit_price, direction, zscore, zscore_abs, mult, cost_pips, cost_usd, pnl_bt, pnl_real, mfe_pct, mae_pct, year, month """ import csv, sys from pathlib import Path from collections import defaultdict from datetime import datetime def resolve_base(): for b in [Path(r"C:\Users\Administrator\OneDrive\fx_macro_intraday"), Path(r"C:\Users\paul_\OneDrive\fx_macro_intraday")]: if b.exists(): return b sys.exit("base not found") BASE=resolve_base() p = BASE/"data"/"processed"/"usdjpy_trades_real_costs.csv" rows=[] with open(p, newline="", encoding="utf-8-sig") as f: for r in csv.DictReader(f): try: r["pnl_real"]=float(r["pnl_real"]) r["mae_pct"]=float(r.get("mae_pct",0) or 0) r["mfe_pct"]=float(r.get("mfe_pct",0) or 0) r["year"]=int(float(r.get("year",0) or 0)) r["zscore_abs"]=float(r.get("zscore_abs",0) or 0) except (ValueError,KeyError): continue rows.append(r) n=len(rows) total=sum(r["pnl_real"] for r in rows) wins=[r for r in rows if r["pnl_real"]>0] losses=[r for r in rows if r["pnl_real"]<=0] print("="*70) print(f"USDJPY DIAGNOSTIC ({n} trades, 2003-2026)") print("="*70) # ---- 5. headline stats ---- gross_win=sum(r["pnl_real"] for r in wins) gross_loss=sum(r["pnl_real"] for r in losses) pf = (gross_win/abs(gross_loss)) if gross_loss else float('inf') print(f"\n[HEADLINE]") print(f" Total P&L : ${total:,.0f}") print(f" Win rate : {100*len(wins)/n:.1f}% ({len(wins)}W / {len(losses)}L)") print(f" Avg win : ${gross_win/len(wins):,.0f}" if wins else " Avg win: n/a") print(f" Avg loss : ${gross_loss/len(losses):,.0f}" if losses else " Avg loss: n/a") print(f" Profit factor : {pf:.2f} (gross win / gross loss)") print(f" Expectancy/trade : ${total/n:,.1f}") # ---- 1. P&L by exit reason ---- print(f"\n[1. WHERE THE MONEY COMES FROM - P&L by exit reason]") by_reason_pnl=defaultdict(float); by_reason_n=defaultdict(int); by_reason_w=defaultdict(int) for r in rows: er=r.get("exit_reason","?") by_reason_pnl[er]+=r["pnl_real"]; by_reason_n[er]+=1 if r["pnl_real"]>0: by_reason_w[er]+=1 print(f" {'reason':<14}{'count':>7}{'totP&L':>12}{'avgP&L':>10}{'win%':>7}{'%ofProfit':>11}") for er in sorted(by_reason_pnl, key=lambda k:-by_reason_pnl[k]): c=by_reason_n[er]; tp=by_reason_pnl[er] print(f" {er:<14}{c:>7}{tp:>12,.0f}{tp/c:>10,.0f}" f"{100*by_reason_w[er]/c:>6.0f}%{100*tp/total:>10.0f}%") # ---- 2. expiry problem ---- tp_n=by_reason_n.get("tp",0); exp_n=by_reason_n.get("hold_expiry",0); st_n=by_reason_n.get("stop",0) print(f"\n[2. THE EXPIRY PROBLEM - does the 0.70% TP get hit in 24h?]") print(f" Hit TP : {tp_n} ({100*tp_n/n:.0f}%)") print(f" Timed out (expiry): {exp_n} ({100*exp_n/n:.0f}%)") print(f" Stopped out : {st_n} ({100*st_n/n:.0f}%)") if exp_n>=tp_n: print(f" >>> CONFIRMED: more trades TIME OUT than hit target. TP likely too far") print(f" for the 24h window, OR hold too short. Worth testing tighter TP") print(f" or longer hold.") else: print(f" >>> TP is hit more often than timeout - structure is working as intended.") # how profitable are the expiry trades? exp_pnl=by_reason_pnl.get("hold_expiry",0) print(f" Hold-expiry trades net: ${exp_pnl:,.0f} " f"({'profitable on balance' if exp_pnl>0 else 'net losing'})") # ---- 3. outlier dependence ---- srt=sorted(rows, key=lambda r:-r["pnl_real"]) top10=sum(r["pnl_real"] for r in srt[:10]) bot10=sum(r["pnl_real"] for r in srt[-10:]) print(f"\n[3. OUTLIER DEPENDENCE]") print(f" Top 10 trades P&L : ${top10:,.0f} ({100*top10/total:.0f}% of total profit)") print(f" Worst 10 trades : ${bot10:,.0f}") print(f" Biggest single win: ${srt[0]['pnl_real']:,.0f} ({srt[0].get('year')})") print(f" Biggest single loss: ${srt[-1]['pnl_real']:,.0f} ({srt[-1].get('year')})") if top10/total>0.5: print(f" >>> WARNING: >50% of profit from 10 trades. Edge is outlier-dependent/fragile.") else: print(f" >>> Profit is reasonably distributed, not just a few lucky trades.") # ---- 4. intervention sensitivity (by year) ---- print(f"\n[4. P&L BY YEAR (* = known BoJ intervention period)]") INTERV={2022,2024,2026} by_year_pnl=defaultdict(float); by_year_n=defaultdict(int) for r in rows: by_year_pnl[r["year"]]+=r["pnl_real"]; by_year_n[r["year"]]+=1 for y in sorted(by_year_pnl): flag=" *" if y in INTERV else " " bar_pnl=by_year_pnl[y] print(f" {y}{flag} {by_year_n[y]:>4} trades ${bar_pnl:>10,.0f}") interv_pnl=sum(by_year_pnl[y] for y in by_year_pnl if y in INTERV) other_pnl=sum(by_year_pnl[y] for y in by_year_pnl if y not in INTERV) print(f"\n Intervention-year P&L total: ${interv_pnl:,.0f}") print(f" Other-year P&L total : ${other_pnl:,.0f}") if interv_pnl<0: print(f" >>> UJ LOSES money in intervention years. Structural vulnerability to BoJ.") else: print(f" >>> UJ survives intervention years on balance (though may be bumpy within).") # ---- decision summary ---- print(f"\n{'='*70}") print("DECISION HELP:") print(f" - If profit factor < ~1.2, win-rate-thin, AND outlier/intervention dependent") print(f" -> UJ's edge is questionable; consider EU-only or restructure.") print(f" - If profit factor healthy and distributed -> UJ earns its place; the wide") print(f" TP/timeout may just be the strategy's nature (trades exit at expiry near") print(f" breakeven, wins come from the moves that do run).") print(f" - The expiry % tells you if TP/hold structure is worth re-testing.") print("="*70)