#Load in required packages
library(PlayerRatings)
library(fitzRoy)
library(tidyverse)## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.3 ✔ readr 2.1.4
## ✔ forcats 1.0.0 ✔ stringr 1.5.0
## ✔ ggplot2 3.4.4 ✔ tibble 3.2.1
## ✔ lubridate 1.9.3 ✔ tidyr 1.3.0
## ✔ purrr 1.0.2
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(elo)
library(dplyr)#Scrape afl data using FitzRoy package
elo_afl2023 <- fitzRoy::fetch_results_afltables(season = 2023)There are multiple rounds in the finals series. Each game will impact a teams elo rating so we must update and create new data frames for each round we want #to predict
#The Elimination Finals (EF) and Qualyfying finals (QF) are all based on rounds 1 to 24.
#We will also create a new column to determine if the home team result was a win,loss or draw
afl2023_rating_EF_QF <- elo_afl2023 %>%
filter(!(Round %in% c('EF', 'QF', 'SF', 'PF', 'GF'))) %>%
mutate(score = ifelse(Home.Points > Away.Points, 1,
ifelse(Home.Points < Away.Points, 0,
0.5)))#The semi-finals (SF) are based on rounds 1 to 24, the EF and the QF
#We will also create a new column to determine if the home team result was a win,loss or draw
afl2023_rating_SF <- elo_afl2023 %>%
filter(!(Round %in% c('SF', 'PF', 'GF'))) %>%
mutate(score = ifelse(Home.Points > Away.Points, 1,
ifelse(Home.Points < Away.Points, 0,
0.5)))#The pleliminary finals (PF) are based on rounds 1 to 24, the EF, the QF and the SF
#We will also create a new column to determine if the home team result was a win,loss or draw
afl2023_rating_PF <- elo_afl2023 %>%
filter(!(Round %in% c('PF', 'GF'))) %>%
mutate(score = ifelse(Home.Points > Away.Points, 1,
ifelse(Home.Points < Away.Points, 0,
0.5)))#The Grand-finals (GF) are based on every round except for the GF itself
#We will also create a new column to determine if the home team result was a win,loss or draw
afl2023_rating_GF <- elo_afl2023 %>%
filter(!(Round %in% c('GF'))) %>%
mutate(score = ifelse(Home.Points > Away.Points, 1,
ifelse(Home.Points < Away.Points, 0,
0.5)))Using these data frame we can figure out the elo ratings of every team at certain points in the finals series
#Calculate elo ratings for EF and QF
eloratings_EF_QF <- elo(afl2023_rating_EF_QF[c("Round.Number", "Home.Team", "Away.Team", "score")])
#Calculate elo ratings for SF
eloratings_SF <- elo(afl2023_rating_SF[c("Round.Number", "Home.Team", "Away.Team", "score")])
#Calculate elo ratings for PF
eloratings_PF <- elo(afl2023_rating_PF[c("Round.Number", "Home.Team", "Away.Team", "score")])
#Calculate elo ratings for GF
eloratings_GF <- elo(afl2023_rating_GF[c("Round.Number", "Home.Team", "Away.Team", "score")])We will do one example on how to use these rating to predict a win probability.
Lets take the game in the EF with Carlton vs Sydney
#Use the predict function and include the elorating df which the game round belongs to (i.e. for this game it was in the EF)
#Choose the Home team and Away team
elo_pred_carlton_sydney <- predict(eloratings_EF_QF,
newdata = data.frame(
Week = NA,
Home.Team = "Carlton",
Away.Team = "Sydney"
))
print(elo_pred_carlton_sydney)## [1] 0.5827805This produces the probability.
We can sim this over 1000 games to see what happens. We’d expect the ratio to be similar to our probability.
#We will use rbinom to create either a 1 (Home/Carlton win) or 0 (Home/Carlton loss)
#The probability of getting a 1 will be determined by our win probability determined earlier
data.frame(
Carlton = rbinom(n = 1000, size = 1, prob = elo_pred_carlton_sydney)
) %>%
summarise(Carlton = sum(Carlton),
Sydney = 1000 - Carlton)## Carlton Sydney
## 1 583 417This shows that out of 1000 games simulated Carlton won 559 and Sydney won 441.