Time Waits for No One:
How Olympic Athletes’ Ages Correlate with Performance

Yuexi (Tracy) Chen, Hana Hailu, Joshua McCurry, Junran Yang

Email: {ychen151, hhailu, jmccurry, jryang}@umd.edu

Click to begin!

Introduction: Time waits for no one

The 2020 Tokyo Olympics was postponed => All Olympic athletes will be one-year older next year. The simple observation inspires us to explore the correlation between athletes’ ages and performance thoroughly.

We used a dataset of 260k+ Olympic athletes' information (name, age, gender, nationality, medal, etc) from 1896 to 2016. We created following interactive data visualizations:

  1. Overview: frequency plots (and box plots) for athletes of different ages, sports, Olympic years, and countries.[1]
  2. Sports vs. age: a stacked bar chart and a dot-line chart
  3. Country/gender vs. age: based on athletes’ nationalities, we visualize athletes’ performance of different income-level country groups according to the World Bank’s report.
  4. A prediction tool: we trained a machine learning model to support real-time performance prediction based on the user input.

[1] To reduce visual clutter and support smooth interactions, we sampled 5% of all datasets in part 1.

Part 1: Overview

How to interact:

  1. Sort by: sort the frequency plot by age, number of athletesor the alphabetical order
  2. Filter: filter data points by season or gender
  3. Change y-axis: the data could be aggregated by sports, country, year and gender
  4. Change mode: toggle to show the density plot or the box plot
  5. Animations: scroll down and the y-axis will be updated automatically!

Try it yourself!

Sports & Age

Now the y-axis shows all the Olympic sports

The relative importance of the physical, technical, tactical and psychological components is not the same for a marathon runner as for a gymnast or a basketball player. These differences likely contribute substantially to the broad range of ages of peak performance of the top Olympic athletes.

Too small? Scroll inside the plot to zoom and pan! Or try the filter to see sports of winter and summer Olympics

Background knowledge

Some sports appear in every summer Olympic Games, e.g. athletics, swimming, fencing, etc., while some sports appear only recently, e.g. badminton (since 1992), Triathlon (since 2000). Similary, sports like Figure skating, Ice hockey and Ski jumping appear in every winter Olympics, while some sports appear only recently, e.g. short track speed skating (since 1992), snowboarding (since 1998).

Fun facts

Art competitions formed part of the modern Olympic Games during its early years, from 1912 to 1948. Medals were awarded for works of art inspired by sport, divided into five categories: architecture, literature, music, painting, and sculpture. The juried art competitions were abandoned in 1954 because artists were considered to be professionals, while Olympic athletes were required to be amateurs.

Country

Now the y-axis shows all the nations and regions have participated in Olympics

Too small? Scroll inside the plot to zoom and pan!

Year

Now the y-axis shows all the Olympics years

Try to toggle the mode button to see the distribution more clearly

For Olympic Athletes, Is 30 the New 20?

The ages of Olympic athletes are getting steadily older in some sports. The age of swimmers is up 13%. The age of gymnasts is up 12%. Even the age of athletes in track and field is up 5%. Competitors are also older than ever in canoeing, handball, fencing, judo and table tennis. Advances in sports science are helping preserve skilled but aging athletes. New ideas about training have lengthened their careers.

Background knowledge

Winter and summer Olympics used to be held in the same year. Since 1994, summer Olympics were held in leap years, and winter Olympics were held between two leap years (still every 4 years). In 1967, the committee banned performance enhancing drugs, in the 4 olympiad following that the average of athletes' age dropped from 24 to 22.After 1984, professional athletes became allowed to compete.

Gender

Now the y-axis shows male and female athletes

Background knowledge

In the first modern Olympic games (1896), there were no female athletes. In 2016, 45% athletes are female.



Sort by:




Color by:

Filter by:



Filter by:


Choose Y-axis Attribute:
dot-plot box-and-whisker-plot
Mode


Part 2: Sports vs Age in details

The world has an aging population, are Olympic athletes also older than their peers in the past? Does one-year older influence medals a lot? In part 2, we provided two visualizations to facilitate insight discovery: on the left is a stacked bar chart of medals of all time, and on the right is a dot-line chart across years, of which the black line represents the age range of non-medalists, and gold/silver/bronze represents the median age of corresponding medalists.

How to interact:

  1. Drop-down menu:
    2. Users can select a season, sport and event.
  2. Tooltips:
    3. Users can hover on the bars or dots to see the detailed information
  3. Year slider:
    4. Users can drag and move the year slider to see how the two plots change accordingly. (The green bars indicate active years of that sport/event)

Part 3: Country vs Age in details

For this visualization we focused on giving insights about the total number of medals won by athletes and medalists with their corresponding age and gender. Research shows that higher income per capita hints higher medal totals. Here we allow users to observe additional insights by looking at both the income level groups and each countries.

How to interact:

  1. The map:
    2. Users can hover on the map to see country names, and also zoom and pan.
    Users can click a country on the map, and the tornado chart will change to reflect the selected country in darker colors.
  2. The slider:
    3. Users can drag the year slider to see how both the map and bar charts changes.
  3. The bar charts ("tornado charts"):
    4. Users can hover on each bar to see details.
    Users can choose to show "all-athletes" or "all-medalists" by the radio button.

World Income Groups

Note: only for summer Olympics as much fewer countries participate in winter Olympics

all-athletes all-medalists

Finally: the prediction

How to interact:

Users can select a season, sports, gender, income level of the home country from the dropdown menu, and drag the age slider to see probabilities of different medal types.

How did we train the model

Our prediction tool works by generating a random forest of several hundred decision trees based on seperate training data sets for each sport. The training data sets of athlete-events include age, gender and national income level as features, and the type of medal awarded (including no medal) as labels. After fitting the random forest to the training set, and calculating test scores, predictions are generated for hypothetical athletes of different income levels, gender, and ages. By calculating the relative ratio of trees that predict each medal type we can provide a rough estimate of medal probabilities.

For every 100 such athletes, our model predicts