Plot football abilities from Stan and MLE models

Description

Depicts teams' abilities either from the Stan models fitted via the stan_foot function or from MLE models fitted via the mle_foot function. Supported MLE models include "double_pois", "biv_pois", "dixon_coles", "neg_bin", "skellam", and "student_t". Models with separate attack and defence parameters (all except "student_t") display attack and defence effects; the "student_t" model displays combined global abilities.

Usage

foot_abilities(object, data, type = "both", teams = NULL)

Arguments

object	An object either of class stanFoot, CmdStanFit, stanfit, or class list containing the Maximum Likelihood Estimates (MLE) for the model parameters fitted with mle_foot.
data	A data frame containing match data with columns: periods: Time point of each observation (integer >= 1). home_team: Home team's name (character string). away_team: Away team's name (character string). home_goals: Goals scored by the home team (integer >= 0). away_goals: Goals scored by the away team (integer >= 0).
type	Type of ability for models with separate attack/defence parameters: one among "defense", "attack" or "both". Default is "both". Ignored for the "student_t" model which uses combined abilities.
teams	An optional character vector specifying team names to include. If NULL, all teams are included.

Value

A ggplot object showing each selected team's ability estimates:

• For static Bayesian or MLE fits, horizontal error bars (95% intervals) and point estimates.

• For dynamic Bayesian fits, ribbon and line plots over periods.

Author(s)

Leonardo Egidi legidi@units.it and Roberto Macrì Demartino roberto.macridemartino@deams.units.it.

Examples

## Not run: 
if (instantiate::stan_cmdstan_exists()) {
  library(dplyr)

  data("italy")
  italy <- as_tibble(italy)

  ### no dynamics, no prediction

  italy_2000_2002 <- italy %>%
    dplyr::select(Season, home, visitor, hgoal, vgoal) %>%
    dplyr::filter(Season == "2000" | Season == "2001" | Season == "2002")

  colnames(italy_2000_2002) <- c("periods", "home_team", "away_team", "home_goals", "away_goals")

  fit1 <- stan_foot(
    data = italy_2000_2002,
    model = "double_pois"
  ) # double poisson

  fit2 <- stan_foot(
    data = italy_2000_2002,
    model = "biv_pois"
  ) # bivariate poisson

  fit3 <- stan_foot(
    data = italy_2000_2002,
    model = "skellam"
  ) # skellam

  fit4 <- stan_foot(
    data = italy_2000_2002,
    model = "student_t"
  ) # student_t

  foot_abilities(fit1, italy_2000_2002)
  foot_abilities(fit2, italy_2000_2002)
  foot_abilities(fit3, italy_2000_2002)
  foot_abilities(fit4, italy_2000_2002)

  ### seasonal dynamics, predict the last season

  fit5 <- stan_foot(
    data = italy_2000_2002,
    model = "biv_pois",
    predict = 180,
    dynamic_type = "seasonal"
  ) # bivariate poisson
  foot_abilities(fit5, italy_2000_2002)
}

### MLE models (no Stan required)

data("italy")
italy_2000 <- italy[italy$Season == "2000", ]
italy_2000 <- data.frame(
  periods = italy_2000$Season,
  home_team = italy_2000$home,
  away_team = italy_2000$visitor,
  home_goals = italy_2000$hgoal,
  away_goals = italy_2000$vgoal
)

## Dixon-Coles MLE
mle_dc <- mle_foot(data = italy_2000, model = "dixon_coles")
foot_abilities(mle_dc, italy_2000)

## Negative Binomial MLE
mle_nb <- mle_foot(data = italy_2000, model = "neg_bin")
foot_abilities(mle_nb, italy_2000, type = "attack")

## End(Not run)