Bayesian team-strength estimates. Each shape is a posterior distribution — its width is the model's uncertainty, its position is relative to the league average (0) over all team-seasons.
above league averagebelow league averageleague middle 50%shaded width = 95% credible interval