Changelog

footBayes 2.1.0

New models

  • Add the Dixon-Coles model ("dixon_coles") to stan_foot(), static and dynamic, and to mle_foot().
  • Add the negative binomial model ("neg_bin") to mle_foot().

Dynamic models in stan_foot()

  • Add dynamic_par to choose the evolution variance: a common variance (common_sd, Owen, 2011) or a variance inflation after the summer break (kl_variance and periods_per_season, Koopman and Lit, 2015).
  • Add dynamic_weight for the weighted dynamic models with commensurate priors (Macrì Demartino, Egidi and Torelli, 2026), with spike and slab hyperpriors set through dynamic_par$spike and dynamic_par$slab.

Other changes

  • print.stanFoot() with teams keeps all the global parameters and shows the team names in the team-indexed parameters.
  • foot_prob() supports the "dixon_coles" and "neg_bin" MLE models.
  • Add an identifiability constraint to the Bradley-Terry-Davidson Stan models.
  • Extend the confidence intervals of mle_foot() to the model-specific parameters and move the MLE simulations to the internal simulate_goals_mle().
  • Rewrite the vignette.

Bug fixes

  • Fix the likelihood of the static bivariate Poisson model.
  • Draw y_rep and y_prev of the bivariate Poisson models from the bivariate Poisson distribution.
  • Rewrite the diagonal-inflated bivariate Poisson model as in Karlis and Ntzoufras (2003), with the new parameter draw_dist for the draws.
  • Return diff_y_prev in the bivariate Poisson and diagonal-inflated bivariate Poisson models.
  • Fix the AIC and BIC of mle_foot() to count only the free parameters of each model.
  • Fix foot_prob() for MLE models, which used object$home instead of object$home_effect.
  • Fix compare_foot() with NA rows in a probability matrix, which misaligned the outcomes of the following elements of source.
  • Match ranking in stan_foot() to the teams by name instead of by position, with informative errors for missing or duplicated teams.
  • Fix the observed frequencies in pp_foot(type = "aggregated") for models fitted with predict > 0.

Documentation

  • Update foot_abilities() for the six MLE models.
  • Fix the documentation of italy, priors, btd_foot(), compare_foot(), plot_logStrength() and of norm_method and rho in stan_foot().

footBayes 2.0.1

  • Updated vignette.
  • Correct typo in foot_compare() description.
  • Correct typo in pp_foot() description.

footBayes 2.0.0

  • Migration from rstan interface to CmdStanR interface.
  • Add support for the VI, pathfinder and laplace algorithms.
  • Add pre-compiled CmdStan models using the package instantiate.
  • Add AIC and BIC output elements in mle_foot().
  • Add Bayesian static/dynamic Negative Binomial model in stan_foot().
  • Minor ggplot2 edits on foot_rank(), foot_abilities() and pp_foot().
  • Updated vignette.

footBayes 1.0.0

  • Bradley-Terry model for abilities.
  • Dynamic ranking in the models.
  • Updated vignette.
  • Probabilistic predictive performance (pseudo-R squared, Brier score, etc.).

footBayes 0.2.0

  • Inclusion of diagonal-inflated bivariate Poisson and zero-inflated Skellam models.
  • Minor edits and drop engsoccerdata dependence.
  • Updated vignette.

footBayes 0.1.0

  • First submission to CRAN.