R/elo_rank.R

In Chang-Yu-Chang/MigrationCommunity: Microbial Consumer-Resource Model

#' Calculate competition ranking by elo
#' @description Calculate the competitive ranking
#' @name elo_rank
#' @param network igraph object generated by `network_make()`
#' @return ...

elo_rank <- function(network) {
  net <- network

  # Remove duplicate coexistence
  ## Coexistence links
  edges_coext <- subgraph.edges(net, eids = which(E(net)$interaction == "coexistence"), delete.vertices = F) %>%
    as.undirected() %>%
    as.directed(mode = "arbitrary") %>%
    get.edgelist()

  ## Competition links
  net_compt <- subgraph.edges(net, eids = which(E(net)$interaction == "exclusion"), delete.vertices = F)
  net_comb <- add.edges(net_compt, as.vector(t(edges_coext)))
  E(net_comb)$interaction[is.na(E(net_comb)$interaction)] <- "coexistence"

  # Elo input table
  pairs_elo <- as.data.frame(cbind(1, get.edgelist(net_comb), E(net_comb)$interaction)) %>% setNames(c("period", "from", "to", "result"))
  pairs_elo$period <- as.numeric(pairs_elo$period)
  pairs_elo$from <- as.character(pairs_elo$from)
  pairs_elo$to <- as.character(pairs_elo$to)
  pairs_elo$result <- as.character(pairs_elo$result)
  pairs_elo$result[pairs_elo$result == "exclusion"] <- 1 # Player 1 wins
  pairs_elo$result[pairs_elo$result == "coexistence"] <- 0.5 # Draw
  pairs_elo$result <- as.numeric(pairs_elo$result)

  # Elo rank
  elo_ranking <- PlayerRatings::elo(pairs_elo)$rating %>% dplyr::select(-Lag)

  # Standardize rating
  min_rate <- min(elo_ranking$Rating)
  max_rate <- max(elo_ranking$Rating)
  elo_ranking$Rating <- (elo_ranking$Rating - min_rate) / (max_rate - min_rate)

  return(elo_ranking)
}


if (FALSE) {

  # Split by community and drop community column. Calculate Elo score
  pairs1 <- pairs1[!is.na(pairs1$score), ] %>% # Remove missing value
    split.data.frame(., f = .$replicate) %>%
    lapply(function(x) {
      PlayerRatings::elo(dplyr::select(x, replicate, from, to, score)) %>%
        `[[`(1) %>%
        mutate(replicate = x$replicate[1])
    })

  # Standardize the ranking score
  # The most competitive isolate in the community scores 1, and the least competitive isolate scores 0.
  pairs1 <- lapply(pairs1, function(x) {
    min_rate <- min(x$Rating)
    max_rate <- max(x$Rating)
    x$Rating <- (x$Rating - min_rate) / (max_rate - min_rate)
    return(x)
  })


  # rbind result
  pairs1 <- do.call("rbind", pairs1)

  #
  pairs1 <- pairs1 %>%
    dplyr::select(-Lag) %>%
    setNames(c("isolate", "rating", "games", "win", "draw", "loss", "replicate")) %>%
    dplyr::select(replicate, isolate, rating, games, win, draw, loss)
  row.names(pairs1) <- NULL

  return(pairs1)
}