Members: Samara Feldhacker (samara28), Alina Kirkpatrick (alinayjk)
The Iowa Trail Run Series is a series of trail running events that has taken place over the last several years. Each race is held at a different state park across Iowa, ranging from McFarland Park in Ames to Yellow River State Forest in northeastern Iowa. With the rise in popularity of running over the past few years, trail racing has grown with it. Many people look for specifically trail races because they enjoy the challenge of soft yet technically demanding trails compared to flat road races.
In 2025, the Iowa Trail Run Series featured eight different races with a variety of distance options ranging from a 5K to a half marathon. Participants can sign up for a full season pass granting access to all races, or register for individual events. Many of these runners return year after year, building a tight-knit community around the series.
In addition, the series also includes a points-based competition that runs throughout the year. Runners earn points by placing in the top 5 of the Overall Winners category for their gender, or in the top 3 within their age category. Because of this aspect, this project was motivated in part by personal interest. I (Samara Feldhacker) have signed up for the 2026 Iowa trail Run Series and wanted to better understand the trends among past participants. By analyzing the 2025 race results, we hope to identify patterns in participation and performance that might offer useful insights.
In order to analyse trail race participation and performance, we will also explore the following questions:
- Which cities/states are most represented, and do runners from closer locations perform differently than those who travel farther?
- How does age group affect pace, and does that relationship change across different race distances?
- How does the performance gap between male and female athletes (measured by fastest race times) vary across age categories?
- Is there significant a relationship between race distance and the number of participants in running events?
There are eight datasets that need to be combined, one for each race event offered by the Iowa Trail Run Series in 2025. Race results are published on the True Timer Racing results portal (results.truetimeracing.com), which hosts finisher data for each event as an HTML table. Because the majority of these pages do not offer a downloadable file, the data was collected via web scraping.
The eight race websites scraped were:
- Center Trails
- Sugar Bottom Trails
- Ledges Trails
- Summer Trails
- Yellow River Trails
- MacBride Trails
- Annett Nature Center
- Jester Park Trails
Each event website contains multiple pages of participants across multiple race distances, including 5K (3.1 mi), 4-mile, 10K (6.2 mi), 8-mile, and half marathon (13.1 mi) options.
The necessary variables that are included within the scraped data are: Pos (position), Name, Time, Category (Age Category), Cat Pos (Category Position), Gender, Gen Pos (Gender Position), City, State, Finish (time), Start (time). Distance and Race Name will also need to be extracted from the website, which allows all of the data to be combined into one large dataset. Additionally, many of these datasets included extra variables or were excluding some needed variables, so these would have to be handled accordingly.
Two functions were created to help the scraping process. The first, get_distance() takes the race name (e.g. “Half Marathon” or “5k Run”) and returns the distance in miles, since not all event pages included a distance column within the data. The second function, information_scrapper() takes a race URL that is the base page of the race results. Then it will collect all of the race distance links on the page, iterate through each distance, iterate through each of the pages of information, extract a results dataset, and add on the distance column when it is needed. Then it finally returns a single dataframe for that specific event.
base_address <- "https://results.truetimeracing.com/"
get_distance <- function(race_name) {
if(race_name == "4 Mile Run") {
return(4.0)
}
if(race_name == "8 Mile Run") {
return(8.0)
}
if (race_name == "Half Marathon") {
return(13.1)
}
if (race_name == "10k Run") {
return(6.2)
}
if (race_name == "5k Run") {
return(3.1)
}
return(0.0)
}
information_scrapper <- function(url) {
#First open the base url
site <- read_html(url)
#First gather all of the races
race_links <- site %>% html_elements("#ctl00_Content_Main_divEvents .nav-link") %>% html_attr(name="href")
race_names <- site %>% html_elements("#ctl00_Content_Main_divEvents .nav-link") %>% html_text()
#Get the other page links for that race
page_links <- site %>% html_elements("#ctl00_Content_Main_grdTopPager a") %>% html_attr(name="href")
#Get the first table
tables <- site %>% html_table(fill=TRUE)
data <- tables[[2]]
colnames(data) <- data[1,]
data <- data[-1,]
#If there is not distance in col names
if(!("Dist" %in% names(data))) {
#Create a new column called distance
data$Dist <- get_distance(race_names[[1]])
}
first_iteration = 1
i <- 1
#Iterate through all of the races
for(race in race_links) {
if(!first_iteration) {
#Read that specific race page
url <- paste(base_address, race, sep="")
html <- read_html(url)
#Get the other page links for that race
page_links <- html %>% html_elements("#ctl00_Content_Main_grdTopPager a") %>% html_attr(name="href")
#Get the first table
tables <- html %>% html_table(fill=TRUE)
data1 <- tryCatch(tables[[2]], error = function(e) NULL)
if(is.null(data1)) {
data <- data[, !duplicated(colnames(data))]
data <- data |>
filter(Pos != "") |>
select(-any_of(c("Fav", "Share", "Cat Pos(Gen Pos)")))
return(data)
}
colnames(data1) <- data1[1,]
data1 <- data1[-1,]
#If there is not distance in col names
if(!("Dist" %in% names(data1))) {
#Create a new column called distance
data1$Dist <- get_distance(race_names[[i]])
}
#Only rbind with common columns
common <- intersect(colnames(data), colnames(data1))
data <- rbind(data[common], data1[common])
}
#Iterate through the links in the race
for(link in page_links) {
if(!is.na(link)) {
url2 <- paste(base_address, link, sep="")
html2 <- read_html(url2)
tables2 <- html2 %>% html_table(fill=TRUE)
data2 <- tables2[[2]]
colnames(data2) <- data2[1,]
data2 <- data2[-1,]
#If there is not distance in col names
if(!("Dist" %in% names(data2))) {
#Create a new column called distance
data2$Dist <- get_distance(race_names[[i]])
}
#Only rbind with common columns
common <- intersect(colnames(data), colnames(data2))
data <- rbind(data[common], data2[common])
}
}
first_iteration <- 0
i <- i+1
}
#Filter not needed columns
data <- data[, !duplicated(colnames(data))]
# Only include all ones where the person finished
data <- data |>
filter(Pos != "") |>
select(!c("Fav", "Share", "Cat Pos(Gen Pos)"))
return(data)
}Using these functions, the eight race events were scraped individually:
center_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1441")
sugar_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1446")
ledges_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1478")
summer_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1492")
yellow_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1575")
macbride_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1593&EId=2")
annett_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1546&EId=1")
jester_trails <- information_scrapper("https://results.truetimeracing.com/results.aspx?CId=16535&RId=1582")These eight dataframes will be the base dataframes that will be further cleaned and combined within the next section.
Not all of the race websites included the same variables, some missing some variables and others containing extra variables. Because of this, before combining the dataframes, each one would need to be formatted to match the others. To do this, columns would have to either be added or removed.
#Center and sugar are good
#Ledges is missing City, State, Pace, Start, Finish
ledges_trails <- ledges_trails |>
mutate(City = "", State = "", Pace = "", Start = "", Finish = "")
#Summer is missing City, State, Pace
summer_trails <- summer_trails |>
mutate(City = "", State = "", Pace = "")
#Yellow trails is missing city, State
yellow_trails <- yellow_trails |>
mutate(City = "", State = "")
#Macbride has extra Age variable
macbride_trails <- macbride_trails |>
select(!c("Age"))
#Annett is missing city, State, Pace
annett_trails <- annett_trails |>
mutate(City = "", State = "", Pace = "")
#Jester is missing city, State
jester_trails <- jester_trails |>
mutate(City = "", State = "")Now, since all of the dataframes share the same columns, a data_cleaner() function was written to ensure consistent variable types between each of the events. This function takes in the current state of the data and the trailName. Then, the function removes the Race No, Pace, and Finish columns (Since Pace is recalculate when needed in a more useful format, and Finish is a duplicate of Time), and then it ensures that the variables are all of correct types (Either numeric, time, or category), and adds on a Trail label corresponding to the TrailName. The function then returns this updated dataframe.
data_cleaner <- function(data, trailName) {
#Remove Race No, Pace, Finish
dataNew <- data |>
select(!c("Race No", "Pace", "Finish"))
#Turn gender into a factor
dataNew$Gender <- factor(dataNew$Gender)
#Turn Pos, Cat Pos, Gen pos, and Dist into a numerical variable
dataNew$Pos <- as.numeric(dataNew$Pos)
dataNew$`Cat Pos` <- as.numeric(dataNew$`Cat Pos`)
dataNew$`Gen Pos` <- as.numeric(dataNew$`Gen Pos`)
dataNew$Dist <- as.numeric(dataNew$Dist)
#convert time, and start time into the time variable type
dataNew$Start <- hms(dataNew$Start)
dataNew$Time <- hms(dataNew$Time)
#Turn category into a factor
dataNew$Category <- factor(dataNew$Category)
#Turn city and State into a factor - note may need to change later but whatever
dataNew$City <- factor(dataNew$City)
dataNew$State <- factor(dataNew$State)
#Add on column for the trailName, and set as factor
dataNew$Trail <- trailName
dataNew$Trail <- factor(dataNew$Trail)
return(dataNew)
}center_trails <- data_cleaner(center_trails, "Center")
sugar_trails <- data_cleaner(sugar_trails, "Sugar")
ledges_trails <- data_cleaner(ledges_trails, "Ledges")## Warning in .parse_hms(..., order = "HMS", quiet = quiet): Some strings failed
## to parse
summer_trails <- data_cleaner(summer_trails, "Summer")
yellow_trails <- data_cleaner(yellow_trails, "Yellow")
macbride_trails <- data_cleaner(macbride_trails, "MacBride")## Warning in .parse_hms(..., order = "HMS", quiet = quiet): Some strings failed
## to parse
annett_trails <- data_cleaner(annett_trails, "Annett")
jester_trails <- data_cleaner(jester_trails, "Jester")## Warning in .parse_hms(..., order = "HMS", quiet = quiet): Some strings failed
## to parse
Now since all of the dataframes are consistent, and are identifiable with the TrailName, they can be combined into a single dataset with rbind(). Additionally, after combing the data, one more cleaning step will need to be completed. This includes making the Category labels consist with eachother and ordering them by age.
trail_races <- rbind(center_trails, sugar_trails, ledges_trails, summer_trails, yellow_trails, macbride_trails, annett_trails, jester_trails)
#Check the levels of Category, and manually rename to make consistent
levels(trail_races$Category)## [1] "14 & under" "Ages 20-24" "Ages 25-29" "Ages 30-34"
## [5] "Ages 35-39" "Ages 40-44" "Ages 45-49" "Ages 50-54"
## [9] "Ages 55-59" "Ages 60-64" "Ages 65-69" "Overall Winners"
## [13] "Ages 15-19" "Unknown" "15-19" "20-24"
## [17] "25-29" "30-34" "35-39" "40-44"
## [21] "45-49" "50-54" "55-59" "60-64"
## [25] "65-69" "70-74" "70+"
levels(trail_races$Category) <- c("0-14", "20-24", "25-29", "30-34", "35-39", "40-44", "45-49", "50-54", "55-59", "60-64", "65-69", "Overall Winners", "15-19", "Unknown", "15-19", "20-24", "25-29", "30-34", "35-39", "40-44", "45-49", "50-54", "55-59", "60-64", "65-69", "70+", "70+")
#reorder of age groups
levels(trail_races$Category)## [1] "0-14" "20-24" "25-29" "30-34"
## [5] "35-39" "40-44" "45-49" "50-54"
## [9] "55-59" "60-64" "65-69" "Overall Winners"
## [13] "15-19" "Unknown" "70+"
trail_races$Category <- factor(trail_races$Category, levels = c("0-14", "15-19", "20-24", "25-29", "30-34", "35-39", "40-44", "45-49", "50-54", "55-59", "60-64", "65-69", "70+", "Overall Winners", "Unknown"))
levels(trail_races$Category)## [1] "0-14" "15-19" "20-24" "25-29"
## [5] "30-34" "35-39" "40-44" "45-49"
## [9] "50-54" "55-59" "60-64" "65-69"
## [13] "70+" "Overall Winners" "Unknown"
Finally, city names require manual correction. Several names appear in all-caps, like AMES, or contained duplication, like Waukeeukee. Because these errors were inconsistent and not easily detectable programmatically, they were corrected by hand.
# Before
levels(trail_races$City)## [1] "" "Adel" "Alburnett"
## [4] "Altoona" "Ames" "AMES"
## [7] "Ankeny" "Atlanta" "Audubon"
## [10] "Belmond" "Bondurant" "Boone"
## [13] "Clive" "Colfax" "De Soto"
## [16] "Des Moines" "Earlham" "Elberon"
## [19] "Eldora" "Ellisville" "Exira"
## [22] "Fort Worth" "Grimes" "Hamlin"
## [25] "Huxley" "Indianola" "Iowa City"
## [28] "Iowa Falls" "Janesville" "JeffersonJefferson"
## [31] "Johnston" "La Crosse" "Lamoni"
## [34] "Linden" "Marshalltown" "Maxwell"
## [37] "Melbourne" "Minneapolis" "Mitchellville"
## [40] "Muscatine" "Nevada" "Newton"
## [43] "North English" "Norwalk" "Ogden"
## [46] "Omaha" "Osage" "Oskaloosa"
## [49] "Otley" "Parkersburg" "Pella"
## [52] "Perry" "Pleasant Hill" "POLK CITY"
## [55] "Prescott" "Riverside" "Runnells"
## [58] "Seattle" "Slater" "Solon"
## [61] "State Center" "Sully" "Tama"
## [64] "Urbandale" "Waukee" "Waukeeukee"
## [67] "West Des Moines" "Windsor Heights" "Albia"
## [70] "Amana" "Aurora" "Basco"
## [73] "Belleville" "Bennett" "Bettendorf"
## [76] "Cambridge" "Carlisle" "Cary"
## [79] "Cedar Falls" "Cedar Rapids" "CEDAR RAPIDS"
## [82] "Center Point" "Central City" "Charles City"
## [85] "Cheyenne" "Clarence" "Clinton"
## [88] "Columbus Junction" "Coralville" "Council Bluffs"
## [91] "Crawfordsville" "Davenport" "Decorah"
## [94] "Dickeyville" "Dubuque" "Dyersville"
## [97] "Dysart" "Ely" "Fairbank"
## [100] "Fairfax" "Fairfield" "Farley"
## [103] "Fort Dodge" "Fort Knox" "Geneseo"
## [106] "Gilbertville" "Grinnell" "Grundy Center"
## [109] "Hedrick" "Hiawatha" "Hudson"
## [112] "IOWA CITY" "Iowa CityI" "Jefferson"
## [115] "Jesup" "Kalona" "Keota"
## [118] "Knoxville" "Liberty" "Lisbon"
## [121] "Marion" "Mason City" "Mechanicsville"
## [124] "Moline" "Monticello" "Monticelloo"
## [127] "Mount Pleasant" "Mount Vernon" "New London"
## [130] "North Liberty" "Oak Forest" "Oelwein"
## [133] "Orange City" "Ottumwa" "OttumwaOttumwa"
## [136] "Oxford" "Packwood" "Palo"
## [139] "Parnell" "Platteville" "Port Byron"
## [142] "Punta Gorda" "Radcliffe" "Redfield"
## [145] "Robins" "Saint Paul" "Silvis"
## [148] "South Amana" "Stanwood" "Strawberry Point"
## [151] "Swisher" "Tiffin" "Tipton"
## [154] "Unionville" "Vinton" "Waterloo"
## [157] "WATERLOO" "Waverly" "Webster City"
## [160] "Wellman" "Wellsburg" "West Branch"
## [163] "West Liberty" "Weston" "Williamsburg"
## [166] "Winterset" "Blue Grass" "Decatur"
## [169] "Des Plaines" "Durant" "Eugene"
## [172] "Galveston" "Independence" "Le Claire"
## [175] "Lone Tree" "Madison" "Northbrook"
## [178] "Peosta" "Pittsburgh" "Rochester"
## [181] "Shreveport" "Sigourney" "Vernal"
# Manual change the levels where needed
levels(trail_races$City)[levels(trail_races$City) == "AMES"] <- "Ames"
levels(trail_races$City)[levels(trail_races$City) == "JeffersonJefferson"] <- "Jefferson"
levels(trail_races$City)[levels(trail_races$City) == "POLK CITY"] <- "Polk City"
levels(trail_races$City)[levels(trail_races$City) == "Waukeeukee"] <- "Waukee"
levels(trail_races$City)[levels(trail_races$City) == "CEDAR RAPIDS"] <- "Cedar Rapids"
levels(trail_races$City)[levels(trail_races$City) == "IOWA CITY"] <- "Iowa City"
levels(trail_races$City)[levels(trail_races$City) == "Iowa CityI"] <- "Iowa City"
levels(trail_races$City)[levels(trail_races$City) == "OttumwaOttumwa"] <- "Ottumwa"
levels(trail_races$City)[levels(trail_races$City) == "WATERLOO"] <- "Waterloo"
# After
levels(trail_races$City)## [1] "" "Adel" "Alburnett"
## [4] "Altoona" "Ames" "Ankeny"
## [7] "Atlanta" "Audubon" "Belmond"
## [10] "Bondurant" "Boone" "Clive"
## [13] "Colfax" "De Soto" "Des Moines"
## [16] "Earlham" "Elberon" "Eldora"
## [19] "Ellisville" "Exira" "Fort Worth"
## [22] "Grimes" "Hamlin" "Huxley"
## [25] "Indianola" "Iowa City" "Iowa Falls"
## [28] "Janesville" "Jefferson" "Johnston"
## [31] "La Crosse" "Lamoni" "Linden"
## [34] "Marshalltown" "Maxwell" "Melbourne"
## [37] "Minneapolis" "Mitchellville" "Muscatine"
## [40] "Nevada" "Newton" "North English"
## [43] "Norwalk" "Ogden" "Omaha"
## [46] "Osage" "Oskaloosa" "Otley"
## [49] "Parkersburg" "Pella" "Perry"
## [52] "Pleasant Hill" "Polk City" "Prescott"
## [55] "Riverside" "Runnells" "Seattle"
## [58] "Slater" "Solon" "State Center"
## [61] "Sully" "Tama" "Urbandale"
## [64] "Waukee" "West Des Moines" "Windsor Heights"
## [67] "Albia" "Amana" "Aurora"
## [70] "Basco" "Belleville" "Bennett"
## [73] "Bettendorf" "Cambridge" "Carlisle"
## [76] "Cary" "Cedar Falls" "Cedar Rapids"
## [79] "Center Point" "Central City" "Charles City"
## [82] "Cheyenne" "Clarence" "Clinton"
## [85] "Columbus Junction" "Coralville" "Council Bluffs"
## [88] "Crawfordsville" "Davenport" "Decorah"
## [91] "Dickeyville" "Dubuque" "Dyersville"
## [94] "Dysart" "Ely" "Fairbank"
## [97] "Fairfax" "Fairfield" "Farley"
## [100] "Fort Dodge" "Fort Knox" "Geneseo"
## [103] "Gilbertville" "Grinnell" "Grundy Center"
## [106] "Hedrick" "Hiawatha" "Hudson"
## [109] "Jesup" "Kalona" "Keota"
## [112] "Knoxville" "Liberty" "Lisbon"
## [115] "Marion" "Mason City" "Mechanicsville"
## [118] "Moline" "Monticello" "Monticelloo"
## [121] "Mount Pleasant" "Mount Vernon" "New London"
## [124] "North Liberty" "Oak Forest" "Oelwein"
## [127] "Orange City" "Ottumwa" "Oxford"
## [130] "Packwood" "Palo" "Parnell"
## [133] "Platteville" "Port Byron" "Punta Gorda"
## [136] "Radcliffe" "Redfield" "Robins"
## [139] "Saint Paul" "Silvis" "South Amana"
## [142] "Stanwood" "Strawberry Point" "Swisher"
## [145] "Tiffin" "Tipton" "Unionville"
## [148] "Vinton" "Waterloo" "Waverly"
## [151] "Webster City" "Wellman" "Wellsburg"
## [154] "West Branch" "West Liberty" "Weston"
## [157] "Williamsburg" "Winterset" "Blue Grass"
## [160] "Decatur" "Des Plaines" "Durant"
## [163] "Eugene" "Galveston" "Independence"
## [166] "Le Claire" "Lone Tree" "Madison"
## [169] "Northbrook" "Peosta" "Pittsburgh"
## [172] "Rochester" "Shreveport" "Sigourney"
## [175] "Vernal"
After cleaning and combining all events, the final dataset (trail_races) contains:
- Pos: Overall finishing position
- Name: Runner’s First and Last Name
- Time: Official finish time (In hours:minutes:seconds)
- Category: Age Category
- Cat Pos: Finishing position within age category
- Gender: The Runner’s Gender
- Gen Pos: Finish position within gender
- City: Runner’s home city
- State: Runner’s home state
- Start: The start Time of the runner
- Dist: Race distance in miles
- Trail: Which race event the data came from
A summary of the dataset is shown below:
summary(trail_races)## Pos Name Time
## Min. : 1.00 Length:1669 Min. :17M 29S
## 1st Qu.: 24.00 Class :character 1st Qu.:37M 49.3200000000002S
## Median : 49.00 Mode :character Median :53M 41.1100000000001S
## Mean : 57.11 Mean :1H 7M 43.7139544637507S
## 3rd Qu.: 82.00 3rd Qu.:1H 21M 43.6999999999998S
## Max. :199.00 Max. :4H 26M 25S
##
## Category Cat Pos Gender Gen Pos
## 35-39 :282 Min. : 1.000 Female:779 Min. : 1.00
## 40-44 :263 1st Qu.: 2.000 Male :890 1st Qu.: 12.00
## 30-34 :234 Median : 3.000 Median : 25.00
## 25-29 :162 Mean : 4.189 Mean : 29.44
## 45-49 :154 3rd Qu.: 6.000 3rd Qu.: 42.00
## Overall Winners:101 Max. :20.000 Max. :114.00
## (Other) :473
## City State Start
## :938 :931 Min. :1S
## Iowa City : 68 Iowa :689 1st Qu.:8H 7M 19.2999999999993S
## Cedar Rapids : 58 Illinois : 17 Median :9H 4M 5.88000000000102S
## Des Moines : 53 Wisconsin: 7 Mean :9H 40M 1.86328641801811S
## North Liberty: 30 Nebraska : 5 3rd Qu.:9H 58M 50.9000000000015S
## Ankeny : 26 Minnesota: 4 Max. :19H 46M 40S
## (Other) :496 (Other) : 16 NA's :248
## Dist Trail
## Min. : 3.100 Sugar :367
## 1st Qu.: 3.100 Ledges :246
## Median : 4.000 Jester :228
## Mean : 5.828 Summer :199
## 3rd Qu.: 6.800 MacBride:188
## Max. :13.100 Center :184
## (Other) :257
Which cities/states are most represented, and do runners from closer locations perform differently than those who travel farther?
When answering the first part of the question, which cities/states are most represented, a bar chart would show the most popular home cities/state from the participants. The top 20 cities are plotted below:
# Making a data set a copy of the original for this question, and remove not needed variables
trail_races_q1 <- trail_races |>
select(!c("Name", "Pos", "Cat Pos", "Gen Pos", "Start"))
trail_races_q1 |>
filter(!is.na(City), City != "") |>
count(City, sort = TRUE) |>
slice_head(n = 20) |>
mutate(City = reorder(City, n)) |>
ggplot(aes(x = n, y = City, fill = factor(n))) +
geom_col(show.legend = FALSE) +
geom_text(aes(label = n), hjust = -0.2, size = 3.2) +
labs(title = "Top 20 Cities by Trail Race Participation",
x = "Number of participants",
y = NULL) +
scale_colour_discrete() +
theme_bw()
As expected, participation is dominated by larger Iowa population centers, with Iowa City, Cedar Rapids, and Des Moines appearing near the top. This makes sense given their population size and proximity to many of the race locations.
The same approach was applied to states:
trail_races_q1 |>
filter(!is.na(State), State != "") |>
count(State, sort = TRUE) |>
mutate(State = reorder(State, n)) |>
ggplot(aes(x = n, y = State, fill = factor(n))) +
geom_col(show.legend = FALSE) +
geom_text(aes(label = n), hjust = -0.2, size = 3.2) +
labs(title = "States by Trail Race Participation",
x = "Number of participants",
y = NULL) +
scale_colour_discrete() +
theme_bw()
As shown above, nearly all participants are from Iowa, with small representation from neighboring Midwest states. Since the Race Series is located within Iowa this makes sense and would be expected.
Now when answering the second part of the question, whether runners from
farther away perform differently, a performance factor would have to be
selected. In this case, time would not be a good factor because of the
different race distances, and instead deriving pace in min/mile would be
an easy way to compare performance between groups. Pace was calculated
from the existing Time and Dist variables:
trail_races_q1 <- trail_races_q1 |>
mutate(time_sec = as.numeric(Time),
pace_min_mile = (time_sec / 60) / Dist)In order to calculate the distance between the race location and the runner’s home location, geocoder was used. Geocoder allows for an address to be entered, and the lat/long coordinates would be returned. In this case, the address had to be City, State_Abbr. These were derived for each of the runners. Then, the race location lat/long were manually derived. Finally, the straight line distance from each runner’s home to the specific race location was calculate using the Haversine function.
#First remove any null city/state values
trail_races_q1 <- trail_races_q1 |>
filter(!is.na(State), State != "", !is.na(City), City != "")
# get the state abbreviation for each State
trail_races_q1 <- trail_races_q1 |>
mutate(State_abbr = state.abb[match(State, state.name)])
# Combine city and state into 1 address
trail_races_q1 <- trail_races_q1 |>
mutate(full_address = paste(City, State_abbr, sep = ", "))
# Get the lat and long from the address
trail_races_q1 <- trail_races_q1 |>
geocode(address = full_address, method = "osm", lat = latitude, long = longitude)## Passing 174 addresses to the Nominatim single address geocoder
## Query completed in: 176.6 seconds
# Now, I'm going to get the location of the races, note: I am manually doing this by looking up the venue locations and the corresponding lat/long. I could also make this a function, but doing it by hand would be easier in this case.
race_locations <- data.frame(
Trail = c("Center", "Sugar", "Ledges", "Summer", "Yellow", "MacBride", "Annett", "Jester"),
race_latitude = c(41.58687, 41.80723, 41.87665, 42.02676, 43.20060, 41.80723, 41.3605, 41.76110),
race_longitude = c(-93.62495, -91.49406, -93.82328, -93.61704, -91.15254, -91.49406, -93.56133, -93.82439)
)
# Calculate distance from the race location
# Combine race_locations data and the trail_races data
trail_races_q1 <- trail_races_q1 |>
left_join(race_locations, by = "Trail") |>
rowwise() |>
mutate(dist_to_race_miles = round(distHaversine(c(race_longitude,race_latitude), c(longitude, latitude)) * 0.00062137), 2) |>
ungroup()The plots below show travel distance vs. pace across all runners, first with no distance filter and then filtered to runners within 500 miles to remove some outliers:
# Everyone
trail_races_q1 |>
ggplot(aes(x = dist_to_race_miles, y = pace_min_mile)) +
geom_point(alpha = 0.3, size = 1.5) +
geom_smooth() +
scale_colour_brewer(type = "qual", palette = "Set1") +
labs(title = "Travel Distance vs. Pace",
x = "Distance Travelled to Race (miles)",
y = "Pace (min/mile)") +
theme_bw()## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
## Warning: Removed 1 row containing non-finite outside the scale range
## (`stat_smooth()`).
## Warning: Removed 1 row containing missing values or values outside the scale range
## (`geom_point()`).
# Within 500 miles
trail_races_q1 |>
filter(dist_to_race_miles <= 500) |>
ggplot(aes(x = dist_to_race_miles, y = pace_min_mile)) +
geom_point(alpha = 0.3, size = 1.5) +
geom_smooth() +
scale_colour_brewer(type = "qual", palette = "Set1") +
labs(title = "Travel Distance vs. Pace",
x = "Distance Travelled to Race (miles)",
y = "Pace (min/mile)") +
theme_bw()## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
Within the plots above, there is no strong linear relationship between the travel distance and pace overall. The only thing notable is that the pace starts as faster for those within the city the race is located in, but as the participants get farther away, the pace goes up slightly, but then remains stable for all distances.
Now, breaking the data down by gender reveals something interesting:
trail_races_q1 |>
filter(Dist <= 4, dist_to_race_miles <= 500) |>
ggplot(aes(x = dist_to_race_miles, y = pace_min_mile, color = Gender)) +
geom_point(alpha = 0.3, size = 1.5) +
geom_smooth() +
scale_colour_brewer(type = "qual", palette = "Set1") +
labs(title = "Travel Distance vs. Pace for races 4 miles or less",
x = "Distance Travelled to Race (miles)",
y = "Pace (min/mile)", color = "Gender") +
theme_bw()## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
trail_races_q1 |>
filter(Dist > 4, dist_to_race_miles <= 500) |>
ggplot(aes(x = dist_to_race_miles, y = pace_min_mile, color = Gender)) +
geom_point(alpha = 0.3, size = 1.5) +
geom_smooth() +
scale_colour_brewer(type = "qual", palette = "Set1") +
labs(title = "Travel Distance vs. Pace for races over 4 miles",
x = "Distance Travelled to Race (miles)",
y = "Pace (min/mile)", color = "Gender") +
theme_bw()## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
The slight trend where the pace starts faster for those within the city where the race is at shows up again, but there are some differences between gender. For Females who race less than 4 miles, there is a significant decrease in pace when someone had to travel about 50 miles to get to the race location. The pace then trends up again after 50 miles traveled. While Males of the same distance have the same trend as all genders shown above. Now when it comes to races over 4 miles, it is more obvious that the further someone has traveled there is an increase in pace. Males also have a further dip in pace at around 125 miles, though not as durastic as the Females for distances less than 4 miles.
Lastly, splitting by age category shows how pace varies with travel distance within each age group:
#display the data
trail_races_q1 |>
filter(dist_to_race_miles <= 200) |>
ggplot(aes(x = dist_to_race_miles, y = pace_min_mile, color = Category)) +
geom_point(alpha = 0.2, size = 1.5) +
geom_smooth(se = FALSE, method = "loess", span = 0.75) +
labs(title = "Travel Distance vs. Pace",
x = "Distance Travelled to Race (miles)",
y = "Pace (min/mile)", color = "Age Category") +
theme_bw()## `geom_smooth()` using formula = 'y ~ x'
Older age groups show slower paces regardless of travel distance. This means that age is a likely a stronger predictor of performance rather than distance traveled.
First, a new data set was created for this question by recalculating pace and removing variables not needed. Runners categorized as “Overall Winners” or “Unknown” were excluded, since their age is not listed.
trail_races_q2 <- trail_races |>
mutate(time_sec = as.numeric(Time),
pace_min_mile = (time_sec / 60) / Dist) |>
select(!c("Name", "Pos", "Cat Pos", "Gen Pos", "Start", "City", "State"))trail_races_q2 <- trail_races_q2 |>
filter(!is.na(Category), !Category %in% c("Overall Winners","Unknown"))A boxplot of pace by age group gives an initial overview of the relationship:
trail_races_q2 |>
ggplot(aes(x = Category, y = pace_min_mile, fill = Category)) +
geom_boxplot(alpha = 0.65, outlier.alpha = 0.25, outlier.size = 1, show.legend = FALSE) +
labs(title = "Pace by age group",
x = "Age group",
y = "Pace (min/mile)") +
theme_bw()
It is generally shown that younger runners within 15-44 range tend to have slightly faster median paces when compared to the older age groups. There is a general slight trend upwards within all of the age groups.
To see if this pattern holds across different race distances, including 3.1, 3.4, 3.65, 4, 6.2, 6.8, 8, 13.1 miles, these will be group together into 3 distinctive groups. Short (3-4 miles), Mid (6-8 miles), and Long (13.1 miles). This will be done to allow for cleaner comparisons between race types. Here a new distance_category label will be created to signifying which group the race belongs to:
levels(factor(trail_races_q2$Dist))## [1] "3.1" "3.4" "3.65" "4" "6.2" "6.8" "8" "13.1"
trail_races_q2$Dist_category <- factor(trail_races_q2$Dist)
levels(trail_races_q2$Dist_category) <- c("Short (3-4 mi)", "Short (3-4 mi)", "Short (3-4 mi)", "Short (3-4 mi)", "Mid (6-8 mi)", "Mid (6-8 mi)", "Mid (6-8 mi)", "Long (13.1 mi)")
levels(trail_races_q2$Dist_category)## [1] "Short (3-4 mi)" "Mid (6-8 mi)" "Long (13.1 mi)"
The boxplots below show pace by age group, faceted by distance category:
trail_races_q2 |>
ggplot(aes(x = Category, y = pace_min_mile, fill = Category)) +
geom_boxplot(alpha = 0.65,outlier.alpha = 0.25,outlier.size = 1, show.legend = FALSE) +
facet_wrap(~ Dist_category, scales = "free_y", ncol = 1) +
labs(title = "Pace by age group across race distances",
x = "Age group",
y = "Pace (min/mile)") +
theme_bw()
The age-pace relationships still shows up within this boxplot between all three distance categories. But the spread in pace does become longer within the longer distances. An interesting outlier also appears, which is that within the 60-64 age group, there are some fast people who are shown within the half marathon distance. To summarize this data more clearly, a heat map of median pace by age group and distance category is shown below:
heatmap <- trail_races_q2 |>
group_by(Category, Dist_category) |>
summarise(median_pace = median(pace_min_mile, na.rm = TRUE), n = n(), .groups = "drop") |>
filter(!is.na(median_pace))
mid_pace <- median(heatmap$median_pace)
heatmap |>
ggplot(aes(x = Dist_category, y = Category, fill = median_pace)) +
geom_tile(colour= "white") +
geom_text(aes(label = sprintf("%d:%02d", floor(median_pace),round((median_pace %% 1) * 60))), size = 3) +
scale_fill_gradient2(low = "green",mid = "white",high = "red", midpoint = median(mid_pace), name = "Median pace\n(min/mile)") +
labs(title = "Median pace heat map — age group x race distance",
subtitle = "Green = faster, red = slower",
x = "Race distance",
y = "Age group") +
theme_bw()
The heat map makes the relationship between age, pace, and distance much easier to read. The fastest paces are located within the 35-39 age group at the mid distances, and the 15-19 age group at the short distances. The slowest paces are in the older age groups in general and both at the long and short distance categories. An interesting trend that is shown overall is that most of the age groups get faster between the short vs mid distances, and the paces for the long distances are about the same as the short distances. This is likely due to the fact that newer and slower runners are more likely to run those shorter distances, and more experienced runners run at the longer distances. The exception to this trend are those people 0-19, where they become slower as the distance is increased. This is likely due to the fact that younger people mostly train speed in school sports, and they do not focus on endurance runs.
How does the performance gap between male and female athletes (measured by fastest race times) vary across age categories?
For easier time comparisons, I made a new column where the time is mutated to be in seconds:
Q3_df <- trail_races %>%
mutate(time_seconds = as.numeric(Time))
Q3_df## # A tibble: 1,669 × 13
## Pos Name Time Category `Cat Pos` Gender `Gen Pos` City State
## <dbl> <chr> <Period> <fct> <dbl> <fct> <dbl> <fct> <fct>
## 1 1 Caleb Soren… 26M 33.82S Overall… 1 Male 1 Pres… Iowa
## 2 2 Joshua Burb… 28M 46.57S Overall… 2 Male 2 Des … Iowa
## 3 3 Dillon Ver … 29M 38.18S Overall… 3 Male 3 Runn… Iowa
## 4 4 Anthony All… 31M 26.23S 30-34 1 Male 4 Colf… Iowa
## 5 5 Nathan Holm… 32M 4.91S 30-34 2 Male 5 Des … Iowa
## 6 6 Ryan Klumpe… 33M 23.17S 30-34 3 Male 6 Des … Iowa
## 7 7 Jeremy Muel… 33M 37.84S 45-49 1 Male 7 Wauk… Iowa
## 8 8 Michael Mey… 33M 43.37S 35-39 1 Male 8 West… Iowa
## 9 9 Chad Mokles… 34M 2.33S 30-34 4 Male 9 Wind… Iowa
## 10 10 Brian Stewa… 34M 24S 50-54 1 Male 10 Wauk… Iowa
## # ℹ 1,659 more rows
## # ℹ 4 more variables: Start <Period>, Dist <dbl>, Trail <fct>,
## # time_seconds <dbl>
Now, I will remove the 0.0 distance:
Q3_df <- Q3_df %>%
filter(!is.na(Dist), Dist != 0.0)
unique(Q3_df$Dist) #Checking it was removed (it was)## [1] 4.00 8.00 13.10 6.80 3.65 6.20 3.10 3.40
I will also create a new column that removes the category variable “Overall Winners” for simplicity:
Q3_df <- Q3_df %>%
filter(Category != "Overall Winners")
Q3_df # checking it worked## # A tibble: 1,568 × 13
## Pos Name Time Category `Cat Pos` Gender `Gen Pos` City State
## <dbl> <chr> <Period> <fct> <dbl> <fct> <dbl> <fct> <fct>
## 1 4 Anthony All… 31M 26.23S 30-34 1 Male 4 Colf… Iowa
## 2 5 Nathan Holm… 32M 4.91S 30-34 2 Male 5 Des … Iowa
## 3 6 Ryan Klumpe… 33M 23.17S 30-34 3 Male 6 Des … Iowa
## 4 7 Jeremy Muel… 33M 37.84S 45-49 1 Male 7 Wauk… Iowa
## 5 8 Michael Mey… 33M 43.37S 35-39 1 Male 8 West… Iowa
## 6 9 Chad Mokles… 34M 2.33S 30-34 4 Male 9 Wind… Iowa
## 7 10 Brian Stewa… 34M 24S 50-54 1 Male 10 Wauk… Iowa
## 8 11 Luis Gomez 34M 42.08S 35-39 2 Male 11 Des … Iowa
## 9 12 Michael Van… 35M 7.37S 35-39 3 Male 12 Anke… Iowa
## 10 14 Jason Pugh 35M 20.28S 30-34 5 Male 13 West… Iowa
## # ℹ 1,558 more rows
## # ℹ 4 more variables: Start <Period>, Dist <dbl>, Trail <fct>,
## # time_seconds <dbl>
Because there are inconsistencies in some of the age catergories across race distances, we will only be removing the following age categories: 0-14, 15-19, 65-69, 70+
Q3_df <- Q3_df %>%
filter(!Category %in% c("0-14", "15-19", "65-69", "70+"))
unique(Q3_df$Category)## [1] 30-34 45-49 35-39 50-54 40-44 20-24 55-59 60-64 25-29
## [10] Unknown
## 15 Levels: 0-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 ... Unknown
I will now find the top 3 runners for each trail, for each distance on the trail, based on gender, and age:
top3_runners <- Q3_df %>%
group_by(Trail, Dist, Category, Gender) %>%
arrange(time_seconds, .by_group = TRUE) %>%
mutate(rank_in_group = row_number()) %>%
filter(rank_in_group <= 3)
top3_runners <- top3_runners %>%
select(
Trail,
Dist,
Category,
Gender,
Name,
time_seconds,
rank_in_group
)
top3_runners## # A tibble: 732 × 7
## # Groups: Trail, Dist, Category, Gender [297]
## Trail Dist Category Gender Name time_seconds rank_in_group
## <fct> <dbl> <fct> <fct> <chr> <dbl> <int>
## 1 Center 4 20-24 Female Haileigh Steffen 3122. 1
## 2 Center 4 20-24 Male Ashwin Sinha 2208. 1
## 3 Center 4 25-29 Female Shay Vandersluis 2525. 1
## 4 Center 4 25-29 Female Molly Luzbetak 2629. 2
## 5 Center 4 25-29 Female Devyn Atzen 4720. 3
## 6 Center 4 25-29 Male Calvin Knuth 2639. 1
## 7 Center 4 25-29 Male AJ Stills 2782. 2
## 8 Center 4 25-29 Male Rian Simpson 2951. 3
## 9 Center 4 30-34 Female Danielle Curtis 2491. 1
## 10 Center 4 30-34 Female Olivia Meyer 2491. 2
## # ℹ 722 more rows
# Making sure it made a new df instead of altering the original
trail_races## # A tibble: 1,669 × 12
## Pos Name Time Category `Cat Pos` Gender `Gen Pos` City State
## <dbl> <chr> <Period> <fct> <dbl> <fct> <dbl> <fct> <fct>
## 1 1 Caleb Soren… 26M 33.82S Overall… 1 Male 1 Pres… Iowa
## 2 2 Joshua Burb… 28M 46.57S Overall… 2 Male 2 Des … Iowa
## 3 3 Dillon Ver … 29M 38.18S Overall… 3 Male 3 Runn… Iowa
## 4 4 Anthony All… 31M 26.23S 30-34 1 Male 4 Colf… Iowa
## 5 5 Nathan Holm… 32M 4.91S 30-34 2 Male 5 Des … Iowa
## 6 6 Ryan Klumpe… 33M 23.17S 30-34 3 Male 6 Des … Iowa
## 7 7 Jeremy Muel… 33M 37.84S 45-49 1 Male 7 Wauk… Iowa
## 8 8 Michael Mey… 33M 43.37S 35-39 1 Male 8 West… Iowa
## 9 9 Chad Mokles… 34M 2.33S 30-34 4 Male 9 Wind… Iowa
## 10 10 Brian Stewa… 34M 24S 50-54 1 Male 10 Wauk… Iowa
## # ℹ 1,659 more rows
## # ℹ 3 more variables: Start <Period>, Dist <dbl>, Trail <fct>
I am now going to summarize the average times for each age group and gender per race and distance:
gap_run_time <- Q3_df %>%
group_by(Trail, Dist, Category, Gender) %>%
summarise(
avg_time = mean(time_seconds, na.rm = TRUE),
.groups = "drop"
)
gap_run_time## # A tibble: 297 × 5
## Trail Dist Category Gender avg_time
## <fct> <dbl> <fct> <fct> <dbl>
## 1 Center 4 20-24 Female 3122.
## 2 Center 4 20-24 Male 2208.
## 3 Center 4 25-29 Female 3867.
## 4 Center 4 25-29 Male 2885.
## 5 Center 4 30-34 Female 2919.
## 6 Center 4 30-34 Male 2194.
## 7 Center 4 35-39 Female 3422.
## 8 Center 4 35-39 Male 2565.
## 9 Center 4 40-44 Female 4090.
## 10 Center 4 40-44 Male 2610.
## # ℹ 287 more rows
From here, I will create a grouped bar-chart to visualize the data:
#for better labeling
gap_run_time$Dist_label <- paste0("Dist - ", gap_run_time$Dist)
#the actual graph
ggplot(gap_run_time, aes(x = Category, y = avg_time, fill = Gender)) +
geom_bar(stat = "identity", position = "dodge") +
scale_fill_manual(values = c("Female" = "pink",
"Male" = "lightblue")) +
facet_wrap(~Dist_label) +
labs(
title = "Performance Gap Between Male and Female Runners",
x = "Age Category",
y = "Average Time (seconds)"
) +
theme_minimal() +
theme(
axis.text.x = element_text(angle = 45, hjust = 1),
strip.text = element_text(face = "bold"),
legend.position = "top"
)
Conclusion: from this bar chart we can see that male runners consistently have faster average race distances. The performance gap is present in every subgroup but does vary in magnitude depending age. It tends to be smaller in younger age groups and become varied in older age groups, however female average times increase more notably.
Overall, while gender does have an association with performance differences, the size gap is influenced more by age, and varied across race distances.
Is there a relationship between race distance and the number of participants in running events (does race distance affect number of participants)?
First, I need to find the number of participants at the races, keeping in mind I do not want to count those at 0.0, and keeping them grouped by distance as well:
participants <- trail_races %>%
filter(!is.na(Dist), Dist != 0.0) %>%
group_by(Trail, Dist) %>%
summarise(
n_participants = n(),
.groups = "drop"
)
participants## # A tibble: 18 × 3
## Trail Dist n_participants
## <fct> <dbl> <int>
## 1 Center 4 112
## 2 Center 8 72
## 3 Sugar 3.65 103
## 4 Sugar 6.8 150
## 5 Sugar 13.1 114
## 6 Ledges 3.1 150
## 7 Ledges 6.2 96
## 8 Summer 3.1 199
## 9 Yellow 3.1 24
## 10 Yellow 6.2 34
## 11 Yellow 13.1 64
## 12 MacBride 3.4 68
## 13 MacBride 6.8 62
## 14 MacBride 13.1 58
## 15 Annett 3.1 81
## 16 Annett 6.2 54
## 17 Jester 3.1 129
## 18 Jester 6.2 99
Now that I have all the participant counts, we can visualize the data. I added red man dots to visualize the mean number of runners in each race distance.
ggplot(participants, aes(x = factor(Dist), y = n_participants)) +
geom_jitter(width = 0.15, alpha = 0.6, size = 2) +
stat_summary(fun = mean, geom = "point", size = 4, color = "red") +
labs(
title = "Participants by Race Distance",
x = "Distance",
y = "Number of Participants"
) +
theme_minimal()
From here, I am going to use descriptive statistics in order to find if there is evidence that race distance affects number of participants.
participants_summary <- participants %>%
group_by(Dist) %>%
summarise(
mean_participants = mean(n_participants),
median_participants = median(n_participants),
sd_participants = sd(n_participants),
min_participants = min(n_participants),
max_participants = max(n_participants),
.groups = "drop"
)
cor(participants$Dist, participants$n_participants)## [1] -0.2528598
#-0.3384045
participants_summary## # A tibble: 8 × 6
## Dist mean_participants median_participants sd_participants min_participants
## <dbl> <dbl> <dbl> <dbl> <int>
## 1 3.1 117. 129 66.9 24
## 2 3.4 68 68 NA 68
## 3 3.65 103 103 NA 103
## 4 4 112 112 NA 112
## 5 6.2 70.8 75 32.0 34
## 6 6.8 106 106 62.2 62
## 7 8 72 72 NA 72
## 8 13.1 78.7 64 30.7 58
## # ℹ 1 more variable: max_participants <int>
Conclusion: From my corr analysis, the number we are given is -0.3384045. This means that as there is a weak to moderate negative relationship between race distance and number of participants. This suggests that shorter races tend to attract more runner, while longer races tend to have fewer participants. However, the relationship is not strong, meaning that distance alone does not majorly affect number of participants. The summary statistics and the scatter plot support this, particularly the scatter plat as there is no sense of linearity in it.
Therefore I have come to the conclusion that there is not a strong relationship between race distance and number of participants.
From a personal standpoint, the most important thing to consider would be the median paces within the age groups in each race distance category. This is important because I am currently in the 15-19 age category. Since that age group is faster within the 3-4 mile range but slows down at 6+ miles, it may be to my benefit to run the longer distances if they are offered, since I would be more likely to win my age group at these races. But once I enter the 20-24 age group, they have the opposite trend where most people are slower within the shorter distances and faster within the mid and long distances, so then it would be to my advantage to choose to run the shorter distances instead. As long as these trends continue into the 2026 series, this analysis provides a genuinely useful roadmap for race selection and seeing where I fall within my age group.