Update app.py

app.py

@@ -3,32 +3,44 @@ import pandas as pd
 import numpy as np
 import plotly.graph_objects as go
 
-# Live CSV URL for Median Sales Price
 CSV_URL = "https://gardenstatemls.stats.showingtime.com/infoserv/s-v1/kpou-Asg"
 
 def monte_carlo_live(T=1.0, steps=12, n_paths=100):
     try:
-        # Load and clean CSV
+        # Load CSV with metadata skipped
         df = pd.read_csv(CSV_URL, skiprows=9)
-        df = df.dropna(axis=1, how='all')  # Drop empty unnamed columns
+
+        # Drop completely empty columns
+        df = df.dropna(axis=1, how='all')
+
+        # Expect first two useful columns: Date and Median Sales Price
+        df = df.iloc[:, :2]
         df.columns = ['Date', 'Median Sales Price']
 
-        # Parse full month + year format like "January 2009"
-        df['Date'] = pd.to_datetime(df['Date'], format="%B %Y")
+        # Filter out rows where either value is NaN or clearly non-numeric
+        df = df.dropna(subset=['Date', 'Median Sales Price'])
+        df = df[df['Median Sales Price'].str.contains(r'\d', na=False)]
 
-        # Convert price to float
+        # Parse date with full month names
+        df['Date'] = pd.to_datetime(df['Date'], errors='coerce')
+        df = df.dropna(subset=['Date'])
+
+        # Clean price data
         df['Median Sales Price'] = df['Median Sales Price'].replace('[\$,]', '', regex=True).astype(float)
 
-        # Historical prices and returns
+        # Ensure sorted time series
+        df = df.sort_values(by='Date')
+
         prices = df['Median Sales Price'].values
-        returns = np.diff(np.log(prices))
+        if len(prices) < 2:
+            raise ValueError("Not enough valid price data after cleaning.")
 
-        # Calculate annualized drift (mu) and volatility (sigma)
-        mu = np.mean(returns) * 12
-        sigma = np.std(returns) * np.sqrt(12)
+        returns = np.diff(np.log(prices))
+        mu = returns.mean() * 12
+        sigma = returns.std() * np.sqrt(12)
         S0 = prices[-1]
 
-        # Monte Carlo simulation (Geometric Brownian Motion)
+        # Monte Carlo simulation
         dt = T / steps
         paths = np.zeros((steps + 1, n_paths))
         paths[0] = S0
@@ -36,31 +48,26 @@ def monte_carlo_live(T=1.0, steps=12, n_paths=100):
             rand = np.random.normal(0, 1, n_paths)
             paths[t] = paths[t - 1] * np.exp((mu - 0.5 * sigma**2) * dt + sigma * np.sqrt(dt) * rand)
 
-        # Time axis
         time = np.linspace(0, T, steps + 1)
-
-        # Create Plotly chart
         fig = go.Figure()
 
         for i in range(n_paths):
             fig.add_trace(go.Scatter(x=time, y=paths[:, i], mode='lines', line=dict(width=1), showlegend=False))
 
-        # Mean and median paths
         mean_path = paths.mean(axis=1)
         median_path = np.median(paths, axis=1)
         fig.add_trace(go.Scatter(x=time, y=mean_path, name='Mean Path', line=dict(width=3, dash='dash')))
         fig.add_trace(go.Scatter(x=time, y=median_path, name='Median Path', line=dict(width=3, dash='dot')))
 
         fig.update_layout(
-            title="Monte Carlo Simulation — Median Sales Price Forecast",
-            xaxis_title="Time (Years)",
+            title="Monte Carlo Simulation — MLS Median Sales Price Forecast",
+            xaxis_title="Years Ahead",
             yaxis_title="Simulated Price ($)",
             height=600
         )
 
-        # Text summary
-        summary = f"Simulated {n_paths} paths from latest value ${S0:.2f}.\n" \
-                  f"Annualized drift (μ): {mu*100:.2f}%, Volatility (σ): {sigma*100:.2f}%"
+        summary = f"Simulated {n_paths} paths from latest price ${S0:.2f}.\n" \
+                  f"μ = {mu*100:.2f}% / year, σ = {sigma*100:.2f}% / year."
 
         return fig, summary
 
@@ -75,14 +82,14 @@ def monte_carlo_live(T=1.0, steps=12, n_paths=100):
         }])
         return fig, f"Error: {e}"
 
-# Gradio UI
+# Gradio app
 with gr.Blocks() as demo:
-    gr.Markdown("## 📈 Monte Carlo Simulation: Live MLS Median Sales Price Forecast")
+    gr.Markdown("## 📊 Monte Carlo Simulation Based on Live MLS Data")
 
     with gr.Row():
         years = gr.Slider(0.1, 5.0, value=1.0, step=0.1, label="Forecast Horizon (Years)")
         steps = gr.Slider(12, 120, value=12, step=12, label="Time Steps")
-        paths = gr.Slider(10, 300, value=100, step=10, label="Number of Paths")
+        paths = gr.Slider(10, 300, value=100, step=10, label="Number of Simulated Paths")
 
     run_button = gr.Button("Run Simulation")
     plot = gr.Plot()