Skip to contents

Calculate attack evenness

Source: R/attack_evenness.R
ov_aev.Rd

Attack evenness is a measure of how balanced a team's attack is. Teams that rely heavily on one or two attackers will have a relatively low evenness value, whereas teams that use all of their attackers will be higher. See https://untan.gl/attack-evenness.html for further background. Evenness is calculated for each lineup used by a team and averaged (weighting by the number of attacks that each lineup made) to get an overall evenness value. Liberos are not expected to attack, and setter attacks (dumps) are ignored. Evenness can be calculated on grouped data: by default, it is calculated by match but averaged over matches when reporting the final result.

Usage

ov_aev(
  x,
  team,
  rotation,
  reference_props = NULL,
  calculate_by = "match_id",
  report_by,
  min_N_attacks = 10,
  detail = FALSE
)

Arguments

x

data.frame: the plays data.frame as returned by datavolley::read_dv() or peranavolley::pv_read()

team

string: the team to calculate attack evenness for. If not provided, it will be calculated separately for all teams in x

rotation

string: the calculation needs to know what position each player is playing, so that it can work out how many attacks that player should have made under a "perfectly even attack" scenario. This is important for middle hitters, who are usually replaced by the libero in back court and therefore cannot make attacks at those times. The rotation parameter can be one of:

  • "player_role": use the player roles as specified in the "player_role" column in x. These are typically the player roles defined in the roster, but it is left to the user to populate this column

  • "SHM": assume a setter-hitter-middle rotation order (i.e. outside hitter is in position 2 when the setter is in 1) and infer the player roles from that

  • "SMH": assume a setter-middle-hitter rotation order and infer the player roles from that

  • "none": don't assume player roles, which will mean that under a "perfectly even attack" scenario, each player (excluding the setter) will be expected to make 20% of attacks. This option is probably of limited use

reference_props

data.frame or NULL: if NULL, the default attack profile as returned by ov_aev_reference_props() will be used (recommended). Otherwise, provide a data.frame with the same format as that returned by ov_aev_reference_props()

calculate_by

character: names variables in x to group by when doing the calculations. Note that "lineup" is always used as a calculation grouping variable (it is populated inside the function). See Details

report_by

character: names of variables in x to group by for the final results. Note that results are always effectively grouped by team. Any report_by variables will be used in addition to team. Note that:

  • report_by can include "lineup" and/or "setter_id" (these variables are used internally in calculations)

  • report_by variables must also be present in calculate_by, with the exception of "lineup" and "setter_id"

min_N_attacks

integer: minimum number of attacks that must be made in order to be included in the calculations. Attacks are counted by lineup and calculate_by variables (if any). If calculate_by is empty and min_N_attacks is 10, then only lineups that made 10 or more attacks (in total) will be included in the calculations. If min_N_attacks is 10 and calculate_by is "match_id", then calculations will be done match by match, using only lineups in a match that made 10 or more attacks in that match

detail

logical: if TRUE, the returned data frame will include additional columns: "aev_sd" (standard deviation of aev across the calculate_by groups), "rally_win_rate", "kill_rate", "rec_eff" (reception efficiency), "N_rallies" (number of rallies played). Note that including these details makes the calculation noticeably slower

Value

A tibble with at least the columns "team", "aev", and "N_attacks". If detail was TRUE, additional columns will also be present (see the detail parameter)

Details

Note that calculation by group and averaging will not generally give the same results as calculating the average result in one step (e.g. calculating for several matches and averaging those results will probably not give the same answer as calculating for all matches pooled). This is expected: say that my team under-utilizes a particular hitter in one match, and over-utilizes her in another. In both of those matches my team's attack was uneven, and so the average of those two results should return a low evenness value indicating that on average my team's attack was uneven. But if the two matches are pooled and the data analyzed all together, the under-utilization in the first match might be balanced by the over-utilization in the second match, giving a higher attack evenness.

References

https://untan.gl/attack-evenness.html

Examples


px <- plays(dv_read(ovdata_example()))

## for a single team
ov_aev(px, rotation = "SHM", team = "GKS Katowice")
#> # A tibble: 1 × 4
#>   team         aev_sd   aev N_attacks
#>   <chr>         <dbl> <dbl>     <int>
#> 1 GKS Katowice 0.0330 0.861        99

## for all teams in px, and with extra detail
ov_aev(px, rotation = "SHM", detail = TRUE)
#> # A tibble: 2 × 8
#>   team         aev_sd   aev rally_win_rate kill_rate rec_eff N_attacks N_rallies
#>   <chr>         <dbl> <dbl>          <dbl>     <dbl>   <dbl>     <int>     <int>
#> 1 GKS Katowice 0.0330 0.861          0.711     0.455   0.338        99        76
#> 2 MKS Będzin   0.0242 0.75           0.578     0.403   0.372       124        90

## for a single team, calculated by set number but aggregate results when reporting
ov_aev(px, team = "GKS Katowice", rotation = "SHM", calculate_by = "set_number")
#> # A tibble: 1 × 4
#>   team         aev_sd   aev N_attacks
#>   <chr>         <dbl> <dbl>     <int>
#> 1 GKS Katowice 0.0721 0.813        99

## for a single team, calculated and reported by match and set number
ov_aev(px, team = "GKS Katowice", rotation = "SHM", calculate_by = c("match_id", "set_number"),
       report_by = c("match_id", "set_number"))
#> # A tibble: 5 × 6
#>   match_id                         set_number team        aev_sd   aev N_attacks
#>   <chr>                                 <int> <chr>        <dbl> <dbl>     <int>
#> 1 9d209c4d499f8eb78e4e9992fc1c00d0          1 GKS Katowi… 0      0.891        23
#> 2 9d209c4d499f8eb78e4e9992fc1c00d0          2 GKS Katowi… 0      0.8          20
#> 3 9d209c4d499f8eb78e4e9992fc1c00d0          3 GKS Katowi… 0      0.838        17
#> 4 9d209c4d499f8eb78e4e9992fc1c00d0          4 GKS Katowi… 0.0958 0.74         25
#> 5 9d209c4d499f8eb78e4e9992fc1c00d0          5 GKS Katowi… 0      0.804        14

## for a single team, calculated by match and setter position, and reported by setter position
px <- ov_augment_plays(px, to_add = "setters", use_existing = FALSE)
ov_aev(px, team = "GKS Katowice", rotation = "SHM",
       calculate_by = c("match_id", "setter_position"), report_by = "setter_position")
#> # A tibble: 5 × 5
#>   setter_position team           aev_sd   aev N_attacks
#>             <int> <chr>           <dbl> <dbl>     <int>
#> 1               1 GKS Katowice 0        0.8          10
#> 2               3 GKS Katowice 0        0.667        12
#> 3               4 GKS Katowice 0        0.8          10
#> 4               5 GKS Katowice 1.15e-16 0.767        15
#> 5               6 GKS Katowice 0        0.7          15