##Functions to implement the Glicko-2 rating system in R
The Glicko-2 rating system is an improvement over the Elo system. It provodes estimates of individual player ratings, as well as rating deviations (uncertainties in the rating). Additionally, the Glicko-2 system estimates the rating volitility of each player, which estimates how much a player's rating is expected to fluctuate based on how eratic their performances are.
Custom glicko2 function.R includes two functions.
glicko2() implements the Glicko-2 algorithm as described by Mark Glickman https://www.glicko.net/glicko/glicko2.pdf
Rating_calculator() processes input data and calls on the glicko2() functin to calculate and update the ratings, deviations, and volitilities of all players.
Rating_calculator() requires two arguments as inputs (data and state).
data should be a data frame with the current player information and should contain the following columns:
-
Player, Rating, DeViation, Volatility, Last_comp .
-
Player should be the identity of the player, and LAst_comp should be the datetime of their most recent competition.
state should be a data frame with the competition outcomes and should include the following columns:
- timestamp, Player1, Player2, Score
Competition outcomes should be stored in the Score column, and should reflect the outcome from the perspective of Player1. 1 indicates that Player1 was the winner. 0 indicates that Player1 was the looser. 0.5 indicates that the contest was a draw.
JavaScript translation of Glicko-2 R implementation.js includes both of the above functions, translated from R into JavaScript. These functions have not yet been tested.
Example implementation in R
initstate <- data.frame(Player=c("a","b","c","d","e","f","g","h"), Rating = c(1500,1400,1550,1700,1500,1400,1550,1700),
DeViation = c(200,30,100,300,200,30,100,300), Volatility = 0.06, Last_comp = "2023-02-05 15:14:35")
games <- data.frame(timestamp = c(rep("2023-02-06 15:14:35",3),rep("2023-02-07 15:14:35",3)), Player1 = "a", Player2 = c("b","c","d","b","c","d"), Score = c(1,0,0,1,0,0))
Rating_calculator(data=games, state=initstate)