Web Scrapping

The scrapping process takes considerable time. For this reason, I have commented out the last two lines of code that do the scrapping.

NOTE: To repeat the scrapping process, remove the # before the two lines of code.

pages <- 1:225

url <- "https://www.worldathletics.org/records/all-time-toplists/sprints/100-metres/outdoor/men/senior?regionType=world&timing=electronic&windReading=regular&page="

url_2 <- "&bestResultsOnly=false&firstDay=1900-01-01&lastDay=2021-09-20"

################################################################################
## Scrapping function 

scrapper <- function(x){
        
        Sys.sleep(2)
        
        read_html(paste0(url, x, url_2)) %>% 
                
                html_nodes("table") %>% 
                
                html_table()
        
}

################################################################################
## Below is the code for web scrapping. 
## I have commented it out as it takes time to run.
## I have saved the data in a .csv file.: my_100_dash_data.csv
## You can uncomment to rerun the harvesting of data from the web.
################################################################################

# my_100_dash_data <- pages %>% map_dfr(~ scrapper(.x))

#write_csv(my_100_dash_data, "my_100_dash_data.csv")

################################################################################

Feature Engineering

The resultant dataset has 22,400 rows and 14 columns. The data cleaning process involves converting the date of birth (dob) and date to the date/ time format. After this, I did feature engineering, adding the following variables.

• Age of athletes at the time of the race in days.

• Age of athletes at the time of the race in years. I divided the age in days by 365.25 to get years.

• Venue country code: The code of the country where the race happened.

• Venue country name: I used the countrycode package in R to convert the country codes into country names. Where missing, I used information available on https://www.olympiandatabase.com/index.php?id=1670&L=1 to fill in the country names.

my_100_dash_data <- read_csv("data/my_100_dash_data.csv") %>% 
        
        clean_names() %>% 
        
        select(-x8) %>% 
        
        mutate(dob = lubridate::dmy(dob), 
               
               date = lubridate::dmy(date), 
               
               age_days = (date - dob),
               
               age_years = as.numeric(age_days / 365.25),
               
               venue_country_code = str_extract_all(venue, "\\([A-Z]*\\)"), 
               
               venue_country_code = str_remove_all(venue_country_code, "\\(|\\)")) %>% 
        
        mutate(venue_country_name = countrycode(venue_country_code, 
                                                
                                                origin = "ioc", 
                                                
                                                destination = "country.name")) %>% 
        
        mutate(venue_country_name = case_when(venue_country_code == "AHO" ~ "Netherlands Antilles",
                                              
                                              venue_country_code == "FRG" ~ "Germany",
                                              
                                              venue_country_code == "GDR" ~ "Germany",
                                              
                                              venue_country_code == "MAC" ~ "Macau",
                                              
                                              venue_country_code == "TCH" ~ "Czechia",
                                              
                                              venue_country_code == "TKS" ~ "Turks and Caicos Islands",
                                              
                                              venue_country_code == "URS" ~ "Russia",
                                              
                                              TRUE ~ venue_country_name
                                              
                                              ))

################################################################################

Missing Data

Only four variables have missing data- wind, dob, age_days, age_years, and pos. However, the extent of missingness is not great, as the proportion of missing data shows in the table below.

Amelia::missmap(my_100_dash_data)

Visualisation of Missing Data

sapply(my_100_dash_data, is.na) %>% 
        
        colSums() %>% 
        
        tibble(variables = names(my_100_dash_data), missing = .) %>% 
        
        arrange(desc(missing)) %>% 
        
        mutate(prop_percent = missing / nrow(my_100_dash_data) * 100) %>% 
        
        head(10) %>% 
        
        kbl(., booktabs = TRUE, caption = "Missing Data") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Missing Data

################################################################################

Distribution of 100 Meters Male Sprint Best Times

The graph below shows the distribution of the times posted by male athletes in the 100 metres sprint. It shows the razor-thin margins that separate the world beaters like Usain Bolt and other athletes. The maximum difference between the world record holder and the worst time recorded in the dataset is 0.72.

my_100_dash_data %>% 
        
        ggplot(mapping = aes(x = mark)) + 
        
        geom_histogram(col = "black", fill = "skyblue", binwidth = 0.01) + 
        
        ggthemes::theme_economist() + 
        
        labs(x = "Mark", y = "Count", 
             
             title = "Histogram of Best Times of Top 100 Metres Male Sprinters")

################################################################################

I also examine the trend in world-leading times. The figure below shows a gradual decline in the best times posted by male sprinters over 100 meters. There is no telling whether this trend will continue.

## Improvements over time- best times per year

my_100_dash_data %>% 
        
        group_by(year = year(date)) %>% 
        
        summarise(year_record = min(mark)) %>% 
        
        ggplot(mapping = aes(x = year, y = year_record)) +
        
        geom_line(col = "blue") +
        
        labs(x = "Year", y = "Mark/Best Time", 
             
             title = "Trend in 100 Meters Men Best Times") + 
        
        ggthemes::theme_economist()

Most Successful Countries

The table below shows the countries with the highest number of athletes among the top sprinters. The United States leads, as the table below shows. However, as noted earlier, the athletes are repeated, given that one athlete may have posted multiple times. In the next table, I remove the duplicates to get the country with the most athletes.

my_100_dash_data %>% 
        
        count(nat, sort = TRUE) %>% 
        
        head(10) %>% 
        
        kbl(., booktabs = TRUE, caption = "NUmber of Athletes by Nationality") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

NUmber of Athletes by Nationality

############
my_100_dash_data %>% 
        
        select(competitor, nat) %>% 
        
        filter(!duplicated(.)) %>% 
        
        count(nat) %>% 
        
        arrange(desc(n)) %>% 
        
        head(10) %>% 
        
        kbl(., booktabs = TRUE, caption = "Number of Distict Atletes by Nationality") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Number of Distict Atletes by Nationality

################################################################################

World Record Holders

The table below shows the ten athletes who have posted the best times in the 100 metres dash. Note that Usain Bolt appears in this list four times.

my_100_dash_data %>% 
        
        select(competitor, mark) %>% 
        
        arrange(mark) %>% 
        
        head(10) %>% 
        
        kbl(., booktabs = TRUE, 
            
            caption = "Top 10 Best Times in 100 Meters Dash") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Top 10 Best Times in 100 Meters Dash

###########################################

Athletes With the Most appearances in the Fastest Athletes List

The issue here is to examine the athlete that has appeared in the list of elite athletes the most times. In other words, which athlete has had the most races that appear in the dataset of the fastest 100 meters athletes? Here, Michael Rodgers from the USA leads the way, having participated in 267 races and posting some of the best times in the World that appear in this dataset.

my_100_dash_data %>% 
        
        count(competitor, sort = TRUE) %>% 
        
        head(10) %>% 
        
        kbl(., booktabs = TRUE, caption = "Most Appearances (Races) in the Top Sprinters List") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Most Appearances (Races) in the Top Sprinters List

################################################################################

The Age Structure of Athletes

This section examines the ages of the athletes featured in the data. The average age is 25.119 years, while the median is 24.65, with a standard deviation of 3.835. The Nigerian sprinter Chinedu ORIALA is the youngest athlete in the dataset, while Kim Collins is the oldest at around 41 years. We refer to the athlete’s age when they participated in a race and not how old the athlete is currently.

my_100_dash_data %>% 
        
        skim_without_charts(age_years, mark) %>% 
        
        kbl(., booktabs = TRUE, caption = "Summary Statistics for Athletes Age and Mark") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Summary Statistics for Athletes Age and Mark

################################################################################
## Youngest athlete 

my_100_dash_data[which.min(my_100_dash_data$age_years), ] %>% 
        
        kbl(., booktabs = TRUE, caption = "Youngest Athlete in the Dataset") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Youngest Athlete in the Dataset

################################################################################
## Oldest athlete 

my_100_dash_data[which.max(my_100_dash_data$age_years), ] %>% 
        
        kbl(., booktabs = TRUE, caption = "Oldest Athlete in the Dataset") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Oldest Athlete in the Dataset

The figures below show the age structure of the athletes in this dataset. Note that the age determination is the length between the time a race occurs and the date of birth of a given athlete. Here, we see a slight increase in both the mean and median ages of athletes after 1980. Afterwards, the mean and median ages stabilize between 25 and 26 years.

my_100_dash_data %>% 
        
        group_by(year(date)) %>% 
        
        summarise(mean_age = mean(age_years, na.rm = TRUE),
                  
                  median_age = median(age_years, na.rm = TRUE)) %>% 
        
        rename(year = `year(date)`) %>% 
        
        pivot_longer(-year, names_to = "metric", values_to = "age") %>% 
        
        ggplot(mapping = aes(x = year, y = age, col = metric)) + 
        
        geom_line() + 
        
        ggthemes::theme_economist() +
        
        scale_colour_manual(values = c("red", "blue"))

################################################################################

Next, I examine the overall age structure of the athletes using a histogram. The plot shows that most athletes tend to be in their early 20s when they post leading times. After 30 years of age, the performance of athletes declines markedly with minor exceptions.

my_100_dash_data %>% 
        
        ggplot(mapping = aes(x = age_years)) + 
        
        geom_histogram(col = "black", fill = "skyblue", binwidth = 1) + 
        
        ggthemes::theme_economist() + 
        
        labs(x = "Age in Years", y = "Count", 
             
             title = "Histogram of Age of Top 100 Metres Male Sprinters")

###############################################

Slight Detour: A Focus on Usain Bolt

In this section, I delve deeper into the performance by Usain Bolt, the most successful 100 meters athlete of all time. First, I examine the times Usain posted for each year. The figure below shows that the years that Usain recorded the best times also exhibit higher a standard deviation. The table below confirms this observation.

my_100_dash_data %>% 
        
        filter(competitor == "Usain BOLT") %>% 
        
        ggplot(mapping = aes(x = factor(year(date)), y = mark)) + 
        
        geom_boxplot(mapping = aes(fill = factor(year(date)))) + 
        
        geom_point() + 
        
        ggthemes::theme_economist() +
        
        theme(legend.position = "none") +
        
        labs(x = "Year", y = "Mark/Time in Seconds", 
             
             title = "Usain  Bolt 100 Meters Races History 2007-2017")

my_100_dash_data %>%
        
        filter(competitor == "Usain BOLT") %>% 
        
        group_by(year = lubridate::year(date)) %>% 
        
        summarise(median = median(mark, na.rm = TRUE), 
                  
                  sd = sd(mark, na.rm = TRUE)) %>% 
        
        arrange(desc(sd)) %>% 
        
        kbl(., booktabs = TRUE, 
            
            caption = "Median and Standard Deviation for USAIN BOLT") %>% 
        
        kable_classic(full_width = FALSE, latex_option = "hold_position")

Median and Standard Deviation for USAIN BOLT

################################################################################