Update backtest.py

backtest.py

@@ -6,90 +6,83 @@ import numpy as np
 import matplotlib.pyplot as plt
 import os
 
-# --- PARAMETERS (Tweak these to find the best fit) ---
+# --- UNIVERSAL PHYSICS OF FINANCE CONFIG ---
 START_DATE = "2004-01-01"
-END_DATE = "2026-02-26"
-CONFIG = {
-    "TECH_WEIGHT": 0.4,       # Weight of Price/Stochastics
-    "FUND_WEIGHT": 0.4,       # Weight of Quarterly/Financial Health
-    "SENTIMENT_WEIGHT": 0.2,   # Weight of News/Headlines
-    "VOL_TARGET": 0.18        # Risk ceiling (18% volatility)
-}
+INITIAL_CAPITAL = 100000
 
-def backtest_engine():
-    # 1. Load Universe (EQUITY_L.csv)
-    try:
-        uni_df = pd.read_csv("EQUITY_L.csv")
-        uni_df.columns = [c.strip() for c in uni_df.columns]
-        tickers = [f"{s}.NS" for s in uni_df['SYMBOL'].head(50).values] # Top 50 for speed
-        tickers.append("^NSEI")
-    except:
-        tickers = ["RELIANCE.NS", "TCS.NS", "BAJFINANCE.NS", "TITAN.NS", "^NSEI"]
-
-    # 2. Fetch Multi-Source Data
-    data = yf.download(tickers, start=START_DATE, end=END_DATE, progress=False)['Close']
-    data = data.ffill().bfill()
-    benchmark = data['^NSEI']
-    stocks = data.drop(columns=['^NSEI'])
+def fetch_macro_data():
+    """Fetches Equity and Commodity proxies for Crisis Monitoring."""
+    assets = {"NIFTY": "^NSEI", "GOLD": "GC=F", "SILVER": "SI=F"}
+    data = yf.download(list(assets.values()), start=START_DATE, progress=False)['Close']
+    if isinstance(data.columns, pd.MultiIndex):
+        data.columns = [col[0] for col in data.columns]
+    return data.rename(columns={v: k for k, v in assets.items()}).ffill().bfill()
 
-    # 3. Factor Engines
-    # Technical: Z-Score (Mean Reversion)
-    z_score = (stocks - stocks.rolling(20).mean()) / stocks.rolling(20).std()
+def run_simulation(data, mode="Renaissance"):
+    """
+    Modes: 
+    'Aggressor' (Momentum-Heavy), 
+    'Stoic' (Mean Reversion + Gold), 
+    'Renaissance' (Multi-Factor Hybrid)
+    """
+    df = data.copy()
     
-    # Fundamental Proxy: (Using P/E & ROE data would be ideal, 
-    # but here we simulate 'Sentiment' as a 60-day acceleration)
-    sentiment = stocks.pct_change(60) 
-
-    # 4. Simulation Loop (Handling 2008 & 2020)
-    portfolio_val = [100000]
-    cash = 100000
-    holdings = {}
+    # 1. CORE LOGIC: Universal Factors
+    df['SMA_50'] = df['NIFTY'].rolling(50).mean()
+    df['SMA_200'] = df['NIFTY'].rolling(200).mean()
+    df['VOL'] = df['NIFTY'].pct_change().rolling(20).std()
     
-    dates = data.index
-    for i in range(200, len(dates), 5):
-        date = dates[i]
-        curr_val = cash + sum(holdings[t] * stocks.at[date, t] for t in holdings)
-        portfolio_val.append(curr_val)
-
-        # EXIT LOGIC: Sell if score drops or Vol spikes
-        # In 2008 and 2020, SMA200 cross-under triggered exits
-        to_sell = [t for t in holdings if stocks.at[date, t] < stocks[t].rolling(200).mean().at[date]]
-        for t in to_sell:
-            cash += holdings[t] * stocks.at[date, t]
-            del holdings[t]
+    # 2. PHASE 1: CRISIS PREDICTOR (Advance-Decline Proxy & Vol Acceleration)
+    # Volatility Acceleration: Is 'Fear' growing faster than 'Price'?
+    df['VOL_ACCEL'] = df['VOL'].pct_change(5) 
+    # Crisis Signal: Price below 50SMA + Volatility Spiking
+    df['CRISIS_STRESS'] = (df['NIFTY'] < df['SMA_50']) & (df['VOL_ACCEL'] > 0.1)
 
-        # ENTRY LOGIC: Combine Stochastics + Sentiment
-        # Filter for stocks with low Z-score (Value) but positive sentiment (News)
-        scores = (z_score.loc[date] * -1 * CONFIG["TECH_WEIGHT"]) + (sentiment.loc[date] * CONFIG["FUND_WEIGHT"])
-        best_picks = scores.sort_values(ascending=False).head(5).index.tolist()
+    # 3. STRATEGY ALLOCATION
+    if mode == "Aggressor":
+        # 80% Momentum, Fast Exit
+        df['Weight_Equity'] = np.where(df['NIFTY'] > df['SMA_50'], 0.8, 0.0)
+    elif mode == "Stoic":
+        # 60% Mean Reversion, Permanent 30% Gold
+        df['Weight_Equity'] = np.where(df['NIFTY'] > df['SMA_200'], 0.6, 0.2)
+        df['Weight_Gold'] = 0.3
+    else: # Renaissance (The 51% Goal)
+        # Dynamic: Full exposure if trend is strong, 
+        # Aggressive De-leverage if CRISIS_STRESS is True
+        df['Weight_Equity'] = np.where(df['CRISIS_STRESS'], 0.0, 
+                              np.where(df['NIFTY'] > df['SMA_200'], 1.0, 0.2))
+        df['Weight_Gold'] = np.where(df['CRISIS_STRESS'], 0.5, 0.0)
 
-        # Allocate
-        alloc = (curr_val * 0.9) / 5 # Keep 10% cash cushion
-        for t in best_picks:
-            if t not in holdings and len(holdings) < 10:
-                price = stocks.at[date, t]
-                qty = int(alloc / price)
-                if cash > qty * price:
-                    holdings[t] = qty
-                    cash -= qty * price
-
-    # 5. Analysis & Visualization
-    strategy_series = pd.Series(portfolio_val, index=dates[200::5])
-    bench_series = (benchmark.loc[strategy_series.index] / benchmark.loc[strategy_series.index[0]]) * 100000
+    # 4. BACKTEST MATH
+    df['Ret_Nifty'] = df['NIFTY'].pct_change()
+    df['Ret_Gold'] = df['GOLD'].pct_change()
+    
+    # Strategy Return
+    df['Strat_Ret'] = (df['Weight_Equity'].shift(1) * df['Ret_Nifty']) + \
+                      (df.get('Weight_Gold', 0).shift(1) * df['Ret_Gold'])
+    
+    return INITIAL_CAPITAL * (1 + df['Strat_Ret'].fillna(0)).cumprod()
 
-    plt.figure(figsize=(12, 8))
-    plt.subplot(2, 1, 1)
-    plt.plot(strategy_series, label="Meta-Strategy (Hybrid)", color='blue')
-    plt.plot(bench_series, label="Nifty 50 (Benchmark)", color='gray', alpha=0.5)
+def backtest_engine():
+    data = fetch_macro_data()
+    
+    # Run all three
+    res_agg = run_simulation(data, "Aggressor")
+    res_sto = run_simulation(data, "Stoic")
+    res_ren = run_simulation(data, "Renaissance")
+    
+    # Plotting
+    output_file = "backtest_result.png"
+    plt.figure(figsize=(12, 7))
+    plt.plot(res_ren, label="The Renaissance (Hybrid - 51% Goal)", color='purple', linewidth=2.5)
+    plt.plot(res_agg, label="The Aggressor (Momentum)", color='red', alpha=0.6)
+    plt.plot(res_sto, label="The Stoic (Mean Reversion)", color='blue', alpha=0.6)
+    
     plt.yscale('log')
-    plt.title("2004-2026 Simulation: Crisis Performance (2008 & 2020)")
+    plt.title("Meta-Simulation (2004-2026): Anticipating 2008 & 2020 Crises")
+    plt.ylabel("Wealth (INR)")
     plt.legend()
-
-    plt.subplot(2, 1, 2)
-    drawdown = (strategy_series - strategy_series.cummax()) / strategy_series.cummax()
-    plt.fill_between(drawdown.index, drawdown, 0, color='red', alpha=0.3)
-    plt.title("Portfolio Drawdown (Risk Control)")
-    
-    plt.tight_layout()
-    plt.savefig("backtest_result.png")
-    return "backtest_result.png"
+    plt.grid(True, which="both", alpha=0.2)
+    plt.savefig(output_file)
+    plt.close()
+    return output_file