comperes offers a pipe (%>%) friendly set of tools for storing and managing competition results. Understanding of competition is quite general: it is a set of games (abstract event) in which players (abstract entity) gain some abstract scores (typically numeric). The most natural example is sport results, however not the only one. For example, product rating can be considered as a competition between products as “players”. Here a “game” is a customer that reviews a set of products by rating them with numerical “score” (stars, points, etc.).

This package leverages dplyr’s grammar of data manipulation. Only basic knowledge is enough to use comperes.

Overview

comperes provides the following functionality:

  • Store and convert competition results:
  • Summarise:
  • Compute Head-to-Head values (a summary statistic of direct confrontation between two players) with functions also using dplyr’s grammar:
    • Store output in long format as a tibble with one row per pair of players. Function: h2h_long().
    • Store output in matrix format as a matrix with rows and columns describing players and entries - Head-to-Head values. Function: h2h_mat().
    • Use common Head-to-Head functions with rlang’s unquoting mechanism. Example: . %>% h2h_mat(!!!h2h_funs["num_wins"]).

Installation

You can install comperes from CRAN with:


install.packages("comperes")

To install the most recent development version from GitHub use:


# install.packages("devtools")
devtools::install_github("echasnovski/comperes")

Examples

Store and Convert

We will be using ncaa2005, data from comperes package. It is an example competition results (hereafter - results) of an isolated group of Atlantic Coast Conference teams provided in book “Who’s #1” by Langville and Meyer. It looks like this:


library(comperes)
ncaa2005
#> # A longcr object:
#> # A tibble: 20 x 3
#>    game player score
#>   <int> <chr>  <int>
#> 1     1 Duke       7
#> 2     1 Miami     52
#> 3     2 Duke      21
#> 4     2 UNC       24
#> 5     3 Duke       7
#> 6     3 UVA       38
#> # … with 14 more rows

This is an object of class longcr which describes results in long form (each row represents the score of particular player in particular game). Because in this competition a game always consists from two players, more natural way to look at ncaa2005 is in wide format:


as_widecr(ncaa2005)
#> # A widecr object:
#> # A tibble: 10 x 5
#>    game player1 score1 player2 score2
#>   <int> <chr>    <int> <chr>    <int>
#> 1     1 Duke         7 Miami       52
#> 2     2 Duke        21 UNC         24
#> 3     3 Duke         7 UVA         38
#> 4     4 Duke         0 VT          45
#> 5     5 Miami       34 UNC         16
#> 6     6 Miami       25 UVA         17
#> # … with 4 more rows

This converted ncaa2005 into an object of widecr class which describes results in wide format (each row represents scores of all players in particular game). All comperes functions expect either a data frame with results structured in long format or one of supported classes: longcr, widecr.

Summarise

With compere the following summaries are possible:


ncaa2005 %>%
  summarise_player(min_score = min(score), mean_score = mean(score))
#> # A tibble: 5 x 3
#>   player min_score mean_score
#>   <chr>      <int>      <dbl>
#> 1 Duke           0       8.75
#> 2 Miami         25      34.5 
#> 3 UNC            3      12.5 
#> 4 UVA            5      18.5 
#> 5 VT             7      33.5

# Using list of common summary functions
library(rlang)
ncaa2005 %>%
  summarise_game(!!!summary_funs[c("sum_score", "num_players")])
#> # A tibble: 10 x 3
#>    game sum_score num_players
#>   <int>     <int>       <int>
#> 1     1        59           2
#> 2     2        45           2
#> 3     3        45           2
#> 4     4        45           2
#> 5     5        50           2
#> 6     6        42           2
#> # … with 4 more rows

Supplied list of common summary functions has 8 entries, which are quoted expressions to be used in dplyr grammar:


summary_funs
#> $min_score
#> min(score)
#> 
#> $max_score
#> max(score)
#> 
#> $mean_score
#> mean(score)
#> 
#> $median_score
#> median(score)
#> 
#> $sd_score
#> sd(score)
#> 
#> $sum_score
#> sum(score)
#> 
#> $num_games
#> length(unique(game))
#> 
#> $num_players
#> length(unique(player))

ncaa2005 %>% summarise_player(!!!summary_funs)
#> # A tibble: 5 x 9
#>   player min_score max_score mean_score median_score sd_score sum_score num_games num_players
#>   <chr>      <int>     <int>      <dbl>        <dbl>    <dbl>     <int>     <int>       <int>
#> 1 Duke           0        21       8.75          7       8.81        35         4           1
#> 2 Miami         25        52      34.5          30.5    12.3        138         4           1
#> 3 UNC            3        24      12.5          11.5     9.40        50         4           1
#> 4 UVA            5        38      18.5          15.5    14.0         74         4           1
#> 5 VT             7        52      33.5          37.5    19.9        134         4           1

To modify scores based on the rest of results one can use join_*_summary() functions:


suppressPackageStartupMessages(library(dplyr))
ncaa2005_mod <- ncaa2005 %>%
  join_player_summary(player_mean_score = mean(score)) %>%
  join_game_summary(game_mean_score = mean(score)) %>%
  mutate(score = player_mean_score - game_mean_score)

ncaa2005_mod
#> # A longcr object:
#> # A tibble: 20 x 5
#>    game player score player_mean_score game_mean_score
#>   <int> <chr>  <dbl>             <dbl>           <dbl>
#> 1     1 Duke   -20.8              8.75            29.5
#> 2     1 Miami    5               34.5             29.5
#> 3     2 Duke   -13.8              8.75            22.5
#> 4     2 UNC    -10               12.5             22.5
#> 5     3 Duke   -13.8              8.75            22.5
#> 6     3 UVA     -4               18.5             22.5
#> # … with 14 more rows

ncaa2005_mod %>% summarise_player(mean_score = mean(score))
#> # A tibble: 5 x 2
#>   player mean_score
#>   <chr>       <dbl>
#> 1 Duke       -15.5 
#> 2 Miami       11.4 
#> 3 UNC         -5   
#> 4 UVA         -2.12
#> 5 VT          11.2

This code modifies score to be average player score minus average game score. Negative values indicate poor game performance.

Head-to-Head

Computation of Head-to-Head performance is done with h2h_long() (output is a tibble; allows multiple Head-to-Head values per pair of players) or h2h_mat() (output is a matrix; only one value per pair of players).

Head-to-Head functions should be supplied in dplyr grammar but for players’ matchups: direct confrontation between ordered pairs of players (including playing with themselves) stored in wide format:


ncaa2005 %>% get_matchups()
#> # A widecr object:
#> # A tibble: 40 x 5
#>    game player1 score1 player2 score2
#>   <int> <chr>    <int> <chr>    <int>
#> 1     1 Duke         7 Duke         7
#> 2     1 Duke         7 Miami       52
#> 3     1 Miami       52 Duke         7
#> 4     1 Miami       52 Miami       52
#> 5     2 Duke        21 Duke        21
#> 6     2 Duke        21 UNC         24
#> # … with 34 more rows

Typical Head-to-Head computation is done like this:


ncaa2005 %>%
  h2h_long(
    mean_score_diff = mean(score1 - score2),
    num_wins = sum(score1 > score2)
  )
#> # A long format of Head-to-Head values:
#> # A tibble: 25 x 4
#>   player1 player2 mean_score_diff num_wins
#>   <chr>   <chr>             <dbl>    <int>
#> 1 Duke    Duke                  0        0
#> 2 Duke    Miami               -45        0
#> 3 Duke    UNC                  -3        0
#> 4 Duke    UVA                 -31        0
#> 5 Duke    VT                  -45        0
#> 6 Miami   Duke                 45        1
#> # … with 19 more rows

ncaa2005 %>% h2h_mat(mean(score1 - score2))
#> # A matrix format of Head-to-Head values:
#>       Duke Miami UNC UVA  VT
#> Duke     0   -45  -3 -31 -45
#> Miami   45     0  18   8  20
#> UNC      3   -18   0   2 -27
#> UVA     31    -8  -2   0 -38
#> VT      45   -20  27  38   0

Supplied list of common Head-to-Head functions has 9 entries, which are also quoted expressions:


h2h_funs
#> $mean_score_diff
#> mean(score1 - score2)
#> 
#> $mean_score_diff_pos
#> max(mean(score1 - score2), 0)
#> 
#> $mean_score
#> mean(score1)
#> 
#> $sum_score_diff
#> sum(score1 - score2)
#> 
#> $sum_score_diff_pos
#> max(sum(score1 - score2), 0)
#> 
#> $sum_score
#> sum(score1)
#> 
#> $num_wins
#> num_wins(score1, score2, half_for_draw = FALSE)
#> 
#> $num_wins2
#> num_wins(score1, score2, half_for_draw = TRUE)
#> 
#> $num
#> dplyr::n()

ncaa2005 %>% h2h_long(!!!h2h_funs)
#> # A long format of Head-to-Head values:
#> # A tibble: 25 x 11
#>   player1 player2 mean_score_diff mean_score_diff… mean_score sum_score_diff sum_score_diff_… sum_score
#>   <chr>   <chr>             <dbl>            <dbl>      <dbl>          <int>            <dbl>     <int>
#> 1 Duke    Duke                  0                0       8.75              0                0        35
#> 2 Duke    Miami               -45                0       7               -45                0         7
#> 3 Duke    UNC                  -3                0      21                -3                0        21
#> 4 Duke    UVA                 -31                0       7               -31                0         7
#> 5 Duke    VT                  -45                0       0               -45                0         0
#> 6 Miami   Duke                 45               45      52                45               45        52
#> # … with 19 more rows, and 3 more variables: num_wins <dbl>, num_wins2 <dbl>, num <int>

To compute Head-to-Head for only subset of players or include values for players that are not in the results, use factor player column:


ncaa2005 %>%
  mutate(player = factor(player, levels = c("Duke", "Miami", "Extra"))) %>%
  h2h_mat(!!!h2h_funs["num_wins"], fill = 0)
#> # A matrix format of Head-to-Head values:
#>       Duke Miami Extra
#> Duke     0     0     0
#> Miami    1     0     0
#> Extra    0     0     0