Delete app.py

app.py

@@ -1,133 +0,0 @@
-import gradio as gr
-import pandas as pd
-import numpy as np
-import xgboost as xgb
-import plotly.graph_objects as go
-from scipy.stats import gaussian_kde
-from pybaseball import pitching_stats, batting_stats, cache
-
-cache.enable()
-
-class AdvancedProEngine:
-    def __init__(self):
-        # Data Pull
-        p_raw = pitching_stats(2015, 2025, ind=1)
-        h_raw = batting_stats(2015, 2025, ind=1)
-        
-        self.p_metrics = ["ERA", "SO", "WHIP", "Stuff+"]
-        self.h_metrics = ["AVG", "HR", "OPS", "wRC+"]
-        self.p_feats = ['Age', 'IP', 'FIP', 'SwStr%']
-        self.h_feats = ['Age', 'PA', 'OBP', 'ISO']
-        
-        # Safe Data Cleaning
-        self.p_db = p_raw[self.p_feats + self.p_metrics + ['Name', 'Season']].dropna()
-        self.h_db = h_raw[self.h_feats + self.h_metrics + ['Name', 'Season']].dropna()
-        
-        self.models = {}
-        self._train_models()
-
-    def _train_models(self):
-        # Training with raw numpy arrays to prevent DataFrame attribute errors
-        for m in self.p_metrics:
-            self.models[f"P_{m}"] = xgb.train({'objective': 'reg:squarederror'}, xgb.DMatrix(self.p_db[self.p_feats].values, label=self.p_db[m].values))
-        for m in self.h_metrics:
-            self.models[f"H_{m}"] = xgb.train({'objective': 'reg:squarederror'}, xgb.DMatrix(self.h_db[self.h_feats].values, label=self.h_db[m].values))
-
-    def simulate_2026(self, name, role, metric, n_sims=2500):
-        db = self.p_db if role == "Pitcher" else self.h_db
-        feats = self.p_feats if role == "Pitcher" else self.h_feats
-        player_latest = db[db['Name'] == name].sort_values('Season').iloc[-1:]
-        if player_latest.empty: return None
-        
-        base_vals = player_latest[feats].values[0]
-        base_vals[0] += 1 # 2026 Aging
-        
-        # Monte Carlo: Injecting 6% variance noise into physical traits
-        noise = np.random.normal(1.0, 0.06, (n_sims, len(base_vals)))
-        sim_inputs = base_vals * noise
-        return self.models[f"{role[0]}_{metric}"].predict(xgb.DMatrix(sim_inputs))
-
-engine = AdvancedProEngine()
-
-def run_ui(role, name):
-    if not name: return "", go.Figure()
-    
-    metrics = engine.p_metrics if role == "Pitcher" else engine.h_metrics
-    primary_metric = metrics[0]
-    
-    # 1. KPI Scorecards
-    kpi_html = "<div style='display: flex; flex-wrap: wrap; gap: 8px; justify-content: center;'>"
-    for m in metrics:
-        sims = engine.simulate_2026(name, role, m, n_sims=300)
-        mu = np.mean(sims)
-        val = f"{mu:.2f}" if mu < 10 else f"{int(mu)}"
-        kpi_html += f"<div style='background:#fff; border:1px solid #eee; padding:12px; border-radius:10px; min-width:80px; text-align:center;'>"\
-                    f"<div style='font-size:0.7em; color:#777;'>{m}</div><div style='font-size:1.3em; color:#1A73E8; font-weight:bold;'>{val}</div></div>"
-    kpi_html += "</div>"
-    
-    # 2. Monte Carlo Pulse with CONFIDENCE INTERVALS
-    full_sims = engine.simulate_2026(name, role, primary_metric)
-    low, med, high = np.percentile(full_sims, [5, 50, 95]) # The 90% Confidence Interval
-    
-    kde = gaussian_kde(full_sims)
-    x = np.linspace(min(full_sims), max(full_sims), 250)
-    y = kde(x)
-    
-    fig = go.Figure()
-    # The Probability Mountain
-    fig.add_trace(go.Scatter(x=x, y=y, fill='tozeroy', line_color='#1A73E8', fillcolor='rgba(26, 115, 232, 0.15)', name="Probability"))
-    
-    # The Confidence Interval (The Ribbon)
-    fig.add_vrect(
-        x0=low, x1=high, 
-        fillcolor="#1A73E8", opacity=0.15, 
-        layer="below", line_width=0,
-        annotation_text=f"90% Zone: {low:.2f} - {high:.2f}", 
-        annotation_position="top left"
-    )
-    
-    # The Midpoint Pill (Projected Value)
-    fig.add_vline(x=med, line_dash="dash", line_color="#1A73E8", annotation_text="Projected")
-
-    fig.update_layout(
-        title=f"2026 {primary_metric} Confidence Pulse",
-        paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(0,0,0,0)',
-        height=300, margin=dict(l=10,r=10,t=50,b=10),
-        xaxis=dict(showgrid=False, zeroline=False), yaxis=dict(visible=False)
-    )
-    
-    return kpi_html, fig
-
-# --- UI with Algorithm Breakdown ---
-with gr.Blocks(theme=gr.themes.Soft(), css=".gradio-container {max-width: 550px !important;}") as demo:
-    gr.Markdown("## 💎 2026 Player Intelligence")
-    
-    with gr.Row():
-        role_select = gr.Radio(["Pitcher", "Hitter"], value="Pitcher", label="Role")
-        player_select = gr.Dropdown(choices=sorted(engine.p_db['Name'].unique().tolist()), label="Search Player", interactive=True)
-
-    scorecards = gr.HTML()
-    pulse_plot = gr.Plot()
-
-    # Dynamic Updates
-    role_select.change(lambda r: gr.update(choices=sorted(engine.p_db['Name'].unique().tolist() if r=="Pitcher" else engine.h_db['Name'].unique().tolist())), inputs=role_select, outputs=player_select)
-    player_select.change(run_ui, inputs=[role_select, player_select], outputs=[scorecards, pulse_plot])
-
-    # Algorithm Breakdown
-    with gr.Accordion("📚 ALGORITHM BREAKDOWN", open=False):
-        gr.Markdown("""
-        ### 🧠 1. The Brain: XGBoost
-        We use an **Extreme Gradient Boosting** ensemble to model player performance.
-        - **Math:** We minimize the Objective Function $$Obj(\\theta) = \\sum L(y_i, \\hat{y}_i) + \\sum \\Omega(f_k)$$
-        - **Logic:** This handles the non-linear relationship between physical traits (like Velocity) and results (like Strikeouts).
-        """)
-        
-        gr.Markdown("""
-        ### 🎲 2. The Stress Test: Monte Carlo
-        A single guess is risky. We run **2,500 parallel 2026 seasons** for every player.
-        - **The Probability Pulse:** The mountain chart above shows the distribution of all outcomes.
-        - **Confidence Intervals (The Ribbon):** We highlight the **5th to 95th percentile**. This defines the realistic 'Floor' and 'Ceiling' for 2026 based on historical volatility.
-        """)
-        
-
-demo.launch()