SwimmeR
library(SwimmeR)
library(dplyr)
SwimmeR was developed to work with results from swimming competitions. Results are often shared as web pages (.html) or PDF documents, which are nice to read, but make data difficult to access.
SwimmeR solves this problem by importing & cleaning .html and .pdf files containing swimming results, and returns a tidy data frame.
Importing is performed by read_results which takes as an argument a file path as file and a node (for .html only, defaults to ’"pre").
In addition to this vignette I do a lot of demos on how to use SwimmeR at my blog Swimming + Data Science.
Reading PDF Results
ISL results are handled differently, see ISL section below
SwimmeR includes Texas-Florida-Indiana.pdf, results from a tri-meet between the three schools. It can be read in as such:
TX_FL_IN_path <- system.file("extdata", "Texas-Florida-Indiana.pdf", package = "SwimmeR")
TX_FL_IN_text <- read_results(file = TX_FL_IN_path)
TX_FL_IN_text[294:303]
#> [1] "\n --- John Shebat 21 Texas, University of NT SCR"
#> [2] "\nEvent 7 Women 100 Yard Breaststroke"
#> [3] "\n 58.79 A"
#> [4] "\n 1:01.84 B"
#> [5] "\n Name Age School Seed Time Finals Time Points"
#> [6] "\n 1 Lilly King 21 Indiana University NT 59.46 B"
#> [7] "\n r:+0.70 27.62 59.46 (31.84)"
#> [8] "\n 2 Olivia Anderson 21 Texas, University of NT 1:01.88"
#> [9] "\n r:+0.74 29.13 1:01.88 (32.75)"
#> [10] "\n 3 Noelle Peplowski 18 Indiana University NT 1:02.02"
Here we see a subsection of the meet - the top three finishers in the Women’s 100 Yard Breaststroke featuring Olympic gold medalist, and the SwimmeR package’s favorite swimmer, Lilly King.
The next step is to convert this data to a data frame using swim_parse. Because swim_parse works on text strings it is very sensitive to typos and/or nonstandard naming conventions. “Texas-Florida-Indiana.pdf” has two examples of these potential problems.
The first is that Indiana University is sometimes entered as Indiana University, with two spaces between Indiana and University. This is a problem in versions of Swimmer < 0.7.0 because swim_parse will interpret two spaces as a column separator, and will not properly capture Indiana University (two spaces) as a team name. In versions of Swimmer >= 0.7.0 the extra space won’t cause a problem at all.
The second issue is that Texas and Florida are styled as Texas, University of and Florida, University of which I personally disapprove of. It won’t cause any issues in SwimmeR versions >= 0.7.0, but will in earlier versions.
Both of these issues can be fixed with the typo and replacement arguments to swim_parse. Elements of typo will be replaced by the element of replacement with which they share an index, so all instances of the first element of typo will be replaced by the first element of replacement etc. etc. Not specifying typo or replacement will not produce an error, but might negatively impact the results. If your results look strange, or are missing values, look for typos related to those swims.
There is a another argument to swim_parse, called avoid, which will be addressed in the section on reading in html results below.
TX_FL_IN_df <-
swim_parse(
file = TX_FL_IN_text,
typo = c("Indiana University", ", University of"), # not required in versions >= 0.7.0
replacement = c("Indiana University", "") # not required in versions >= 0.7.0
)
Here are those same Women’s 100 Breaststroke results, as a data frame in tidy format:
TX_FL_IN_df[102:104,]
#> Place Name Age Team Finals_Time DQ Exhibition
#> 102 1 Lilly King 21 Indiana University 59.46 0 0
#> 103 2 Olivia Anderson 21 Texas 1:01.88 0 0
#> 104 3 Noelle Peplowski 18 Indiana University 1:02.02 0 0
#> Event Reaction_Time
#> 102 Women 100 Yard Breaststroke 0.70
#> 103 Women 100 Yard Breaststroke 0.74
#> 104 Women 100 Yard Breaststroke 0.71
Reading HTML Results
Reading .html results is very similar to reading pdf results, but a value must be specified to node, containing which CSS node the read_results should look in for results. Here results from the New York State 2003 Girls Championship meet will be read in, from the “pre” node.
NYS_link <- "http://www.nyhsswim.com/Results/Girls/2003/NYS/Single.htm"
NYS_text <- read_results(file = NYS_link, node = "pre")
NYS_text[587:598]
#> [1] "\nEvent 6 Girls 100 Yard Butterfly"
#> [2] "\n==============================================================================="
#> [3] "\nNY State Rcd: S 54.35 1990 Richelle Depold, Scotia"
#> [4] "\n Name Year School Prelims Finals"
#> [5] "\n==============================================================================="
#> [6] "\nNYSPHSAA 2003 Federation Championship"
#> [7] "\nA - Final"
#> [8] "\n 1 Bridget O'Connor 12 1-Scarsdale 56.16 55.42"
#> [9] "\n 26.12 29.30"
#> [10] "\n 2 Lauren Bonfe 12 5-Alfred-Almond 56.37 56.93"
#> [11] "\n 26.18 30.75"
#> [12] "\n 3 Christa Narus 11 11-Ward Melville 58.67 57.94"
Looking at the raw results above one will see that line 2 is a header and contains NY State Rcd:, showing the New York State record. Lines of this type are a common feature in swimming results, but because they contain a recognizable swimming time, without being a result per say, they can cause problems for swim_parse. Like typos these will not cause an error, but might produce nonsense rows in the resulting data frame. swim_parse deals with strings that should not be included in results with the avoid argument. By default avoid contains a lot of common formulations of these header items under avoid_default. You can create your own list of strings as pass it to avoid, or add to avoid_default via avoid_new <- c(avoid_default, "your string here"). Avoid should also include "r\\:" if your results have reaction times (avoid_default already includes "r\\:").
NYS_df <- swim_parse(file = NYS_text, avoid = c("NY State Rcd:"))
NYS_df[358:360,]
#> Place Name Age Team Prelims_Time Finals_Time DQ
#> 358 35 Amanda Acomb 12 5-Wayland <NA> 75.75 0
#> 359 36 April Bresette 11 7-Ausable <NA> 71.45 0
#> 360 1 Bridget O'Connor 12 1-Scarsdale 56.16 55.42 0
#> Exhibition Event
#> 358 0 Girls 1 mtr Diving
#> 359 0 Girls 1 mtr Diving
#> 360 0 Girls 100 Yard Butterfly
Splits
By setting splits = TRUE inside swim_parse one can read in split times. Splits will then be read in as either 50 splits (the default), or 25 splits, depending on the value provided to split_length. Let’s look at those same Texas/Florida/Indiana Results again.
TX_FL_IN_df_splits <-
swim_parse(
read_results(TX_FL_IN_path),
# typo = c("Indiana University", ", University of"), # not required in versions >= 0.7.0
# replacement = c("Indiana University", ""), # not required in versions >= 0.7.0
splits = TRUE,
split_length = 50
)
TX_FL_IN_df_splits[100:102,]
#> Place Name Age Team Finals_Time DQ
#> 100 NA Alexander Margherio 18 Texas, University of 52.68 0
#> 101 NA John Shebat 21 Texas, University of <NA> 1
#> 102 1 Lilly King 21 Indiana University 59.46 0
#> Exhibition Event Reaction_Time Split_50 Split_100
#> 100 1 Men 100 Yard Backstroke <NA> 25.00 27.68
#> 101 0 Men 100 Yard Backstroke <NA> <NA> <NA>
#> 102 0 Women 100 Yard Breaststroke 0.70 27.62 31.84
#> Split_150 Split_200 Split_250 Split_300 Split_350 Split_400 Split_450
#> 100 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> 101 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> 102 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> Split_500 Split_550 Split_600 Split_650 Split_700 Split_750 Split_800
#> 100 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> 101 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> 102 <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> Split_850 Split_900 Split_950 Split_1000
#> 100 <NA> <NA> <NA> <NA>
#> 101 <NA> <NA> <NA> <NA>
#> 102 <NA> <NA> <NA> <NA>
We can now see split times for the 50 and 100 walls, plus more split columns that are filled in for the longer races.
Care is needed however, because split times are handled inconsistently in source data. For example in these results, from a meet between Indiana and Louisville splits are sometimes by 25:
and sometimes by 50 - within the same meet.
Another example, in these 2017 Junior National results from Singapore, the 1500m splits are by 25 for the first 800m, and then the last split is for the final 700m of the race.
Relays are also traditionally handled differently, with splits summing for individual athletes. In the 2018 Big Ten championship results Lilly King does not split a 25.84 on the second 25 of the breaststroke leg of Indiana’s 200 medley relay, rather 25.84 was her time for the entire 50 yard breaststroke leg.
Just be forewarned - splits, even within the same meet, will often require some after-import attention and swimming-specific knowledge to clean.
Formatting Splits
The preferred format for splits is “lap” format, where each split is the duration of a single lap (or length) of the pool. Splits are sometimes also presented in cumulative format, where each split is the total time elapsed at a particular point in the race. For example consider this data frame, containing two swimmers swimming the exact same times and splits, but with one in lap format and the other in cumulative format.
df <- data.frame(
Place = 1,
Name = c("Lenore Lap", "Casey Cumulative"),
Team = rep("KVAC", 2),
Event = rep("Womens 200 Freestyle", 2),
Finals_Time = rep("1:58.00", 2),
Split_50 = rep("28.00", 2),
Split_100 = c("31.00", "59.00"),
Split_150 = c("30.00", "1:29.00"),
Split_200 = c("29.00", "1:58.00")
)
df
#> Place Name Team Event Finals_Time Split_50
#> 1 1 Lenore Lap KVAC Womens 200 Freestyle 1:58.00 28.00
#> 2 1 Casey Cumulative KVAC Womens 200 Freestyle 1:58.00 28.00
#> Split_100 Split_150 Split_200
#> 1 31.00 30.00 29.00
#> 2 59.00 1:29.00 1:58.00
Cumulative splits can be converted to lap splits with the split_to_lap function.
df %>%
filter(Name == "Casey Cumulative") %>%
splits_to_lap()
#> Place Name Team Event Finals_Time Split_50
#> 1 1 Casey Cumulative KVAC Womens 200 Freestyle 1:58.00 28.00
#> Split_100 Split_150 Split_200
#> 1 31.00 30.00 29.00
Splits that are already in lap format can be avoided using the threshold parameter in splits_to_lap. The value is threshold is effectively a maximum lap split value. If no swimmer in the data frame will swim a split slower (i.e. greater than) 35.00 then 35.00 makes a good threshold value. Failing to set threshold in data frames containing both lap and cumulative split times will result in nonsensical splits and warnings from SwimmeR.
df %>%
splits_to_lap(threshold = 35)
#> Place Name Team Event Finals_Time Split_50
#> 1 1 Lenore Lap KVAC Womens 200 Freestyle 1:58.00 28.00
#> 2 1 Casey Cumulative KVAC Womens 200 Freestyle 1:58.00 28.00
#> Split_100 Split_150 Split_200
#> 1 31.00 30.00 29.00
#> 2 31.00 30.00 29.00
Converting to cumulative, although not the preferred format, is possible as well with splits_to_cumulative.
df %>%
filter(Name == "Lenore Lap") %>%
splits_to_cumulative()
#> Place Name Team Event Finals_Time Split_50 Split_100
#> 1 1 Lenore Lap KVAC Womens 200 Freestyle 1:58.00 28.00 59.00
#> Split_150 Split_200
#> 1 1:29.00 1:58.00
Similarly, setting threshold allows the exclusion of splits that are already in cumulative format. Here threshold is a minimum split value.
df %>%
splits_to_cumulative(threshold = 20)
#> Place Name Team Event Finals_Time Split_50
#> 1 1 Lenore Lap KVAC Womens 200 Freestyle 1:58.00 28.00
#> 2 1 Casey Cumulative KVAC Womens 200 Freestyle 1:58.00 28.00
#> Split_100 Split_150 Split_200
#> 1 59.00 1:29.00 1:58.00
#> 2 59.00 1:29.00 1:58.00
Relay Swimmers
The final argument to swim_parse is relay_swimmers, which defaults to FALSE. Setting relay_swimmers = TRUE will cause swim_parse to read in the names of relay swimmers for each relay, and add them to the normal swim_parse output as columns. I don’t love this, because the result is having individual swimmers as rows, and relay swimmers as columns (because relay swimmers are associated with their particular relay). This is not very tidy, and SwimmeR strives to be tidy. Still, the functionality does exist.
TX_FL_IN_df_relay_swimmers <-
swim_parse(
read_results(TX_FL_IN_path),
# typo = c("Indiana University", ", University of"), # not required in versions >= 0.7.0
# replacement = c("Indiana University", ""), # not required in versions >= 0.7.0
relay_swimmers = TRUE
)
TX_FL_IN_df_relay_swimmers[1:3,]
#> Place Name Age Team Finals_Time DQ Exhibition
#> 1 1 <NA> <NA> Indiana University 3:36.59 0 0
#> 2 2 <NA> <NA> Texas, University of 3:36.84 0 0
#> 3 3 <NA> <NA> Texas, University of 3:38.75 0 0
#> Event Reaction_Time Relay_Swimmer_1 Relay_Swimmer_2
#> 1 Women 400 Yard Medley Relay <NA> Morgan Scott Lilly King
#> 2 Women 400 Yard Medley Relay <NA> Claire Adams Olivia Anderson
#> 3 Women 400 Yard Medley Relay <NA> Julia Cook Brooke Hansen
#> Relay_Swimmer_3 Relay_Swimmer_4
#> 1 Christie Jensen Shelby Koontz
#> 2 Remedy Rule Anelise Diener
#> 3 Emily Reese Joanna Evans
It is of course also possible to read in both splits and relay swimmers, by setting both of the relevant arguments to TRUE.
Reading ISL Results
International Swimming League results are technically .pdf files, but they’re formatted very differently, so they have their own special function, swim_parse_ISL. Handling of ISL results is otherwise the same, with the file first going to read_results and then to swim_parse_ISL, returning a data frame.
The SwimmeR package’s favorite swimmer, Lilly King, is involved in the ISL. Let’s see what she got up to at this particular meet.
file_url <-
"https://github.com/gpilgrim2670/Pilgrim_Data/raw/master/ISL/Season_1_2019/ISL_16112019_CollegePark_Day_1.pdf"
if (SwimmeR:::is_link_broken(file_url) == TRUE) {
warning("External data unavailable")
} else {
file_read <- read_results(file_url)
df_ISL <- swim_parse_ISL(file = file_read)
df_ISL[which(df_ISL$Name == "KING Lilly"), ]
}
#> # A tibble: 2 x 7
#> # Rowwise:
#> Place Lane Name Team Finals_Time Event DQ
#> <dbl> <dbl> <chr> <chr> <chr> <chr> <dbl>
#> 1 1 8 KING Lilly CAC 29.00 Women's 50m Breaststroke Final 0
#> 2 1 7 KING Lilly CAC 2:17.78 Women's 200m Breaststroke Final 0
Two first place finishes for Ms. King - very nice! Otherwise all the normal information is here, place, time, team, event etc. Beginning in the 2020 season ISL starts reporting points in their results, which swim_parse_ISL will also read. swim_parse_ISL also handles splits and relay swimmers via the arguments splits and relay_swimmers. All ISL meets thus far have splits at the 50 walls, so there is no split_length argument. Otherwise splits and relay swimmers are handled exactly the same way as swim_parse, detailed above.
Formatting Swimming Times
Once results are captured in R as tidy data frames the real fun can begin - but there’s another problem. Times in swimming are recorded as minutes:seconds.hundredth. This is fine when a time is less than a minute, because 59.99 can be of class numeric in R, but times greater than or equal to a minute 1:00.00 are stuck as class character. SwimmeR provides two functions, sec_format and mmss_format to convert between times as seconds (for doing math), and times as minutes:seconds.hundredths, for swimming-specific display.
data(King200Breast)
King200Breast
#> # A tibble: 50 x 4
#> Event Year Time Date
#> <chr> <chr> <chr> <date>
#> 1 200 Breast Junior 2:02.60 2018-03-17
#> 2 200 Breast Senior 2:02.90 2019-03-23
#> 3 200 Breast Sophomore 2:03.18 2017-03-18
#> 4 200 Breast Freshman 2:03.59 2016-03-19
#> 5 200 Breast Senior 2:03.60 2018-11-17
#> 6 200 Breast Sophomore 2:04.03 2017-02-18
#> 7 200 Breast Junior 2:04.68 2018-02-17
#> 8 200 Breast Senior 2:05.14 2019-02-23
#> 9 200 Breast Junior 2:05.49 2018-03-17
#> 10 200 Breast Freshman 2:05.58 2016-02-20
#> # ... with 40 more rows
Included in SwimmeR is King200Breast, containing all Lilly King’s 200 Breaststroke times for her NCAA career. Times recorded as character values, in standard minutes:seconds.hundredth format. We can use sec_format to format them as seconds, and mmss_format to go back to minutes:seconds.hundredth. Both functions work well with the tidyverse packages.
King200Breast <- King200Breast %>%
dplyr::mutate(Time_sec = sec_format(Time),
Time_swim_2 = mmss_format(Time_sec))
King200Breast
#> # A tibble: 50 x 6
#> Event Year Time Date Time_sec Time_swim_2
#> <chr> <chr> <chr> <date> <dbl> <chr>
#> 1 200 Breast Junior 2:02.60 2018-03-17 123. 2:02.60
#> 2 200 Breast Senior 2:02.90 2019-03-23 123. 2:02.90
#> 3 200 Breast Sophomore 2:03.18 2017-03-18 123. 2:03.18
#> 4 200 Breast Freshman 2:03.59 2016-03-19 124. 2:03.59
#> 5 200 Breast Senior 2:03.60 2018-11-17 124. 2:03.60
#> 6 200 Breast Sophomore 2:04.03 2017-02-18 124. 2:04.03
#> 7 200 Breast Junior 2:04.68 2018-02-17 125. 2:04.68
#> 8 200 Breast Senior 2:05.14 2019-02-23 125. 2:05.14
#> 9 200 Breast Junior 2:05.49 2018-03-17 125. 2:05.49
#> 10 200 Breast Freshman 2:05.58 2016-02-20 126. 2:05.58
#> # ... with 40 more rows
This is useful for comparing times, or plotting
plot(King200Breast$Date, King200Breast$Time_sec, axes = FALSE, ann = FALSE)
axis(1, at = c(16800, 17200, 17600, 18000), labels = c(2016, 2017, 2018, 2019))
axis(2, at = c(125, 130, 135, 140), labels = mmss_format(c(125, 130, 135, 140)), las = 1)
par(mar = c(5,7,4,2) + 0.3)
The same thing can be done in ggplot.
King200Breast %>%
ggplot(aes(x = Date, y = Time_sec)) +
geom_point() +
scale_y_continuous(labels = scales::trans_format("identity", mmss_format)) +
theme_classic() +
labs(y= "Time",
title = "Lilly King NCAA 200 Breaststroke")
Using get_mode to clean swimming data
Swim teams often have abbreviations, for example Lilly King swam for Indiana University, and sometimes “Indiana University” was listed as her team name. Other times though the team might be listed as “IU” or “IUWSD”. James (Sulley) Sullivan swam (probably) for Monsters University, or MU Regularizing these names is a useful part of cleaning data.
Name <- c(rep("Lilly King", 5), rep("James Sullivan", 3))
Team <- c(rep("IU", 2), "Indiana", "IUWSD", "Indiana University", rep("Monsters University", 2), "MU")
df <- data.frame(Name, Team, stringsAsFactors = FALSE)
df
#> Name Team
#> 1 Lilly King IU
#> 2 Lilly King IU
#> 3 Lilly King Indiana
#> 4 Lilly King IUWSD
#> 5 Lilly King Indiana University
#> 6 James Sullivan Monsters University
#> 7 James Sullivan Monsters University
#> 8 James Sullivan MU
Lilly has 4 different teams, but all of them are actually the same team. Similarly Sulley has two teams, but actually only one. Using get_mode to return the most frequently occurring team for each swimmer is easier than manually specifying every swimmer’s team.
df <- df %>%
dplyr::group_by(Name) %>%
dplyr::mutate(Team = get_mode(Team))
df
#> # A tibble: 8 x 2
#> # Groups: Name [2]
#> Name Team
#> <chr> <chr>
#> 1 Lilly King IU
#> 2 Lilly King IU
#> 3 Lilly King IU
#> 4 Lilly King IU
#> 5 Lilly King IU
#> 6 James Sullivan Monsters University
#> 7 James Sullivan Monsters University
#> 8 James Sullivan Monsters University
Drawing brackets
To aid in making single elimination brackets for tournaments and shoot-outs SwimmeR has draw_bracket. Any number of teams between 5 and 64 can be used, with byes automatically assigned to higher seeds.
teams <- c("red", "orange", "yellow", "green", "blue", "indigo", "violet")
draw_bracket(teams = teams)
Now add the results of round two:
round_two <- c("red", "yellow", "blue", "indigo")
draw_bracket(teams = teams,
round_two = round_two)
And round three:
round_three <- c("red", "blue")
draw_bracket(teams = teams,
round_two = round_two,
round_three = round_three)
And crown the champion:
champion <- "red"
draw_bracket(teams = teams,
round_two = round_two,
round_three = round_three,
champion = champion)
