Trading Audit Dashboard Skill v2.0

Produces a full forensic trading performance audit as a single self-contained HTML dashboard. Covers 7 phases: Executive Dashboard, Strategy Reverse-Engineering, Behavioral Forensics, Advanced Analytics (NEW), Monte Carlo Forecasting, Professional Recommendations, and JSON Export.

Step 1 — Identify Input Format and Extract Trade Data

Read the uploaded file(s). Supported formats:

• PDF (broker statements like Trading 212, eToro, IBKR) → use pdfplumber
• CSV → parse with pandas/csv
• XLSX → use openpyxl or pandas
• JSON → load directly

For PDF Statements (most common)

import pdfplumber

with pdfplumber.open('file.pdf') as pdf:
    all_text = {}
    for i, page in enumerate(pdf.pages):
        all_text[i+1] = page.extract_text() or ''

Scan pages for CFD / derivatives trade tables. Look for pages containing:

• Keywords: executed trades, CFD, closed, realized, settlement
• Column headers: DIRECTION, BUY/SELL, QUANTITY, PRICE, RESULT, P&L, PROFIT

Skip Invest/equity account pages — only process CFD/derivatives/futures/forex closed trades.

Key Parsing Rules

For Trading 212 format:

• Opening entries have - in the Average Price column → skip these
• Closing entries have a populated Average Price → include these
• TOTAL RESULT is the last EUR/USD amount on the line (after FX Fee, Result, Div Adj, OI)
• Sign is literal in the text: €64.25 = profit, €-32.10 = loss

For generic CSV/XLSX:

• Look for columns: profit, pnl, realized_pnl, net_result, closed_pnl
• Parse open_time/close_time for hold duration
• Extract volume/lots/quantity for position size analysis
• Extract swap / commission / spread fields if present (used by cost drag analysis)

Minimum Required Fields Per Trade

trade = {
    'date': 'YYYY-MM-DD',
    'time': 'HH:MM:SS',          # or '00:00:00' if unavailable
    'datetime': 'YYYY-MM-DD HH:MM:SS',
    'direction': 'Buy' | 'Sell',
    'instrument': 'XAUUSD',      # ticker/symbol
    'size': 2,                   # lots/units/quantity
    'total_result': 64.25,       # signed float, positive=profit
    'order_type': 'Limit',       # optional
    'swap_cost': 0.0,            # optional — overnight financing (negative = cost)
    'spread_cost': 0.0,          # optional — spread/commission drag
}

Save to /home/claude/parsed_trades.json.

Step 2 — Compute All Statistics

Run scripts/compute_stats.py on the parsed trades. If the script is unavailable, compute inline using the formulas in references/statistics.md.

The script now computes all v2.0 metrics automatically:

Core Metrics (unchanged)

• Win rate, avg win/loss, profit factor, expectancy, net P&L, max win/loss, trades/day

Session / DOW / Size / Order Type (unchanged)

• Asian/London/NY session stats; Mon–Fri day-of-week; position size buckets; order type comparison

Sharpe Ratio & Max Drawdown (unchanged)

• Annualised Sharpe; max drawdown in currency and percent

Monte Carlo (unchanged)

• 10,000 simulations × remaining months of year; P10/P50/P90 projections

v2.0 NEW Metrics

Wilson Confidence Interval on Win Rate

# Formula in references/statistics.md → 'win_rate_ci' key
# Output: lower, upper, label, note
# Always display alongside win rate — critical for small samples (< 100 trades)

Kelly Criterion

# 'kelly' key: full_kelly_pct, half_kelly_pct, note
# Recommend Half-Kelly always; Full Kelly capped at 50% display

Risk of Ruin

# 'risk_of_ruin' key: ror_pct, ror_2x_pct (at 2x lot size), note
# Ruin threshold = 50% drawdown from current equity

Drawdown Recovery Modelling

# 'dd_recovery' key: recovery_trades, recovery_days, note

Loss Streaks & Tilt Detection

# 'loss_streaks' key: max_loss_streak, tilt_detected, tilt_detail

Streak & Momentum Analytics

# 'streak_analytics' key: current_streak, current_streak_type,
#   overconfidence_detected, hot_10_trade_pnl, cold_10_trade_pnl

Rolling Period Comparison (30/60/90 day)

# 'rolling_periods' key: windows {30d, 60d, 90d}, trend (improving/stable/declining)

Hourly P&L Heatmap

# 'hourly_heatmap' key: heatmap[0..23], best_hour, worst_hour
# Only populated when time data available for ≥50% of trades

Multi-Instrument Matrix (when > 1 instrument detected)

# 'instrument_matrix' key: instruments{}, best_by_sharpe, worst_by_sharpe

Transaction Cost / Swap Drag

# 'cost_drag' key: total_swap_cost, total_spread_cost, total_cost_drag,
#   cost_pct_of_gross, estimated (bool), note

Step 3 — Determine Starting Equity

Look in the broker statement for:

• Trading 212: Account value in the CFD section of the Overview
• IBKR: Net Asset Value or Starting Cash
• eToro: Account Balance at period start
• MetaTrader: Balance at start of statement period
• If unavailable: estimate as abs(net_pnl) / 0.20 or ask user

Step 4 — Generate the HTML Dashboard

Read assets/dashboard_template.html for the full visual design system.

Substitute all {{PLACEHOLDER}} values with computed statistics.

Existing Placeholders (unchanged from v1.0)

v2.0 New Placeholders

New Dashboard Sections to Add

After the existing Behavioral Forensi