A Bayesian model for myself to debug - PyData Berlin

Bayesian_Model.py

g = simuls.groupby('Team')
season_hdis = pd.DataFrame({'points_lower': g.points.quantile(.05),
                            'points_upper': g.points.quantile(.95),
                            'goals_for_lower': g.gf.quantile(.05),
                            'goals_for_median': g.gf.median(),
                            'goals_for_upper': g.gf.quantile(.95),
                            'goals_against_lower': g.ga.quantile(.05),
                            'goals_against_upper': g.ga.quantile(.95),
                            })
season_hdis = pd.merge(season_hdis, df_observed, left_index=True, right_on='Team')
column_order = ['Team', 'points_lower', 'Points', 'points_upper', 
                'goals_for_lower', 'Points_For', 'goals_for_median', 'goals_for_upper',
                'goals_against_lower', 'Points_Against', 'goals_against_upper',]
season_hdis = season_hdis[column_order]
season_hdis['relative_goals_upper'] = season_hdis.goals_for_upper - season_hdis.goals_for_median
season_hdis['relative_goals_lower'] = season_hdis.goals_for_median - season_hdis.goals_for_lower
season_hdis = season_hdis.sort_index(by='Points_For')
season_hdis = season_hdis.reset_index()
season_hdis['x'] = season_hdis.index + .5
season_hdis

fig, axs = plt.subplots(figsize=(10,6))
axs.scatter(season_hdis.x, season_hdis.GF, c=sns.palettes.color_palette()[4], zorder = 10, label='Actual Goals For')
axs.errorbar(season_hdis.x, season_hdis.goals_for_median, 
             yerr=(season_hdis[['relative_goals_lower', 'relative_goals_upper']].values).T, 
             fmt='s', c=sns.palettes.color_palette()[5], label='Simulations')
axs.set_title('Actual Points For, and 90% Interval from Simulations, by Team')
axs.set_xlabel('Team')
axs.set_ylabel('Points Scored')
axs.set_xlim(0, 20)
axs.legend()
_= axs.set_xticks(season_hdis.index + .5)
_= axs.set_xticklabels(season_hdis['Team'].values, rotation=45)