app.py

import gradio as gr
import joblib
import numpy as np

# Carregar o modelo serializado
model = joblib.load("tiebreak_model.pkl")

# Função para realizar a previsão
def predict_tiebreak(win_odds_input, loser_odds_input):
    # Converter vírgulas para pontos caso necessário
    win_odds_input = str(win_odds_input).replace(',', '.')
    loser_odds_input = str(loser_odds_input).replace(',', '.')

    # Converter para float
    win_odds = float(win_odds_input)
    loser_odds = float(loser_odds_input)
    
    # Calculando as features
    odds_ratio = win_odds / loser_odds
    log_odds_w = np.log(win_odds)
    log_odds_l = np.log(loser_odds)
    prob_w = 1 / win_odds
    prob_l = 1 / loser_odds
    odds_spread = abs(loser_odds - win_odds)

    # Criando o vetor de features para o modelo
    features = np.array([[odds_ratio, log_odds_w, log_odds_l, prob_w, prob_l, odds_spread]])

    # Realizando a previsão
    prob = model.predict_proba(features)[0, 1]  # Probabilidade da classe 1 (menos de 1.5 tiebreaks)
    
    # Calculando a odds mínima
    odds_minima = 1 / prob
    
    # Formatando a probabilidade para percentual com duas casas decimais
    prob_percent = f"{round(prob * 100, 2)}%"
    
    # Retornando os valores
    return prob_percent, round(odds_minima, 2)

# Interface Gradio
inputs = [gr.Number(label="Win Odds"), gr.Number(label="Loser Odds")]
outputs = [gr.Textbox(label="Probabilidade de menos de 1.5 Tiebreaks"), gr.Textbox(label="Odds Mínima")]

# Criação da interface
gr.Interface(fn=predict_tiebreak, inputs=inputs, outputs=outputs, title="Previsão de Tiebreaks",
             description="Insira as odds de vitória e derrota para prever a probabilidade de haver menos de 1.5 tiebreaks e calcular as odds mínimas.").launch()