Athlete Demographics and Performance Trends

The visualizations below explore the performance of athletes based on their demographic information in the Summer and Winter Olympic Games from 1896 to 2016. The data is visualized using heatmaps, faceted bar charts, scatter plots, and area density charts to analyze the performance of male and female athletes of varying heights, weights, and ages in different events over time. The visualizations are interactive and allow users to explore the data by selecting specific countries, sports, and seasons. Following are the research questions supported by the visualizations:

• How does the importance of height and weight for winning a medal differ between men's and women's categories across different disciplines/events?
• What is the top-performing age (age at which athletes are winning the most medals) for men and women in different sports?
• Are there countries in which women consistently win Olympic medals in one discipline while men do not, and vice versa in another sport?

The first view allows users to view sports with different peak heights and weight (height and weights at which athletes win the most medals relative to the number of athletes participating) for all sports in the scatterplot, colored by the type of medal won in each sport. The distribution plots show a few specific sports that might be interesting to look at in terms of height and weight. When a type of medal is selected from the legend, the distribution of this type is emphasised and only the points for that medal type are shown in the scatter plot

The second view consists of a faceted histogram and a faceted line plot over time, both faceted by sex. The histogram presents the distribution and performance of male and female athletes of varying ages per sport selected, while the line plot demonstrates how the success of the age groups have changed over time

This third view is a heatmap that displays the difference in performance by female and male athletes in the same sport over time, given the user has selected a specific country and season for which to view this data. It aims to help the user identify in which countries and for those countries in which sports there is a large, consistent discrepancy between the number of medals won by the female versus male athletes.

Tasks Supported

The visualizations above support the following tasks:

• Compare between the performance of male and female athletes in different sports, and the distributions of various of their physical attributes such as height, weight and age.
• Identify potential correlations between athletes heights and weights and their performance.
• Explore gender differences over time for different countries.
• Compare the difference in success of female and male athletes between various sports for the same country.
• Identify sports in which there has been a consistent discrepancy between the success of the male and female teams for the same country.

Visualization Design Justification

Linked Scatterplot and Density Plot

• Mark: The mark for the scatterplot being used is mark_circle() as it allows the plot to point out a specific combination of peak height and peak weight, for different sports. This results in each sport on the scatter plot having three points to represent it: one for each type of medal. For the density chart the area mark was used as we are attempting to represent the number of athletes over the height and weight ranges of athletes in each sport without using bins.


• Channels: In the scatter plot the horizontal x-channel is being used to indicate the peak heights, while the vertical y-channel is being used to indicate the peak weights of athletes. The density plot is faceted vertically by sport, and horizontally by sex, and the two density plots for weight and height have been horizontally concatenated to one another. The horizontal x-encoding depicts weight or height, depending on which chart, and all the plots for weight/height for the different sports use a common x-axis which makes the distributions easier to compare between sexes and sports. The vertical y-encoding is used to show the density of the different medal groups over the height and weight ranges.


• Characteristics of Channels: The color in this view is encoding the type of medal won (Gold, Silver, Bronze, No Medal). This color scheme is kept uniform across the two plots so they are easy to compare, so the orange points on the scatter plot correspond to the orange distributions on the distribution chart, and they both represent athletes who won Gold medals. Tooltips are used in the scatterplot to denote which sport each point represents while in the density plot tooltips show the type of medal won as well as the height or weight on the x-axis that the location of the cursor corresponds to. This way it is easier to accurately identify the peak height and weight per gender of the different sports plotted.


• Interactivity: Opacity was used for interactivity encoding, which allows the user to select and view only one of the medal types. For example, one can select gold medals in the legend, and the distributions of athletes who have won gold medals will be emphasised by the reduction of opacity for the non-selected medal types. Upon legend selection, the scatter plot only shows the points belonging to that medal type for each sport (so now each sport will only have one point in the scatterplot), which we have encoded by using a transform filter on the selection.


• Critique: This view successfully allows us to answer the research question because it lets us visualize the heights and weights at which athletes perform best for all sports, while also providing us more detailed information of the distributions of athletes’ heights and weights for some pre-selected sports. However, the drawback of this view is that it doesn’t allow us to visualize the more detailed distribution plots for any sport - only for the ones displayed there. An improvement on the plot would be to allow a brush selection on the scatterplot, which would be linked to the distribution plots so that they only show the top 5 sports in the brush-selected sports in the scatterplot.

Faceted Histogram and Line Plot

• Mark: The mark being used in the histogram is mark_bar() as this allows for effective comparison of the length of bars to determine the difference in counts of records. In the line graph the mark being used is mark_line() although we have specified point=True so that the ‘resolution’ of the data is also shown. The line mark was used for this part of the chart because it shows connections between points, which are intuitively connecting the points in the user's mind, making it clear they belong to the same age group. This makes the graph more easily interpretable.


• Channels: In the histogram the horizontal axis is being used to represent age groups (binned every 2 years), while the vertical axis represents the count of athletes in each age group, as is customary for a histogram. The histograms are faceted by sex, in order to allow for comparison between female and male athletes, as is pertinent to the research question. This faceting highlights the similarities and differences between both the distribution of ages of athletes per gender and their success in the sport selected. The horizontal channel of the line graph is the year, while the y-axis shows the ratio of medals to no medals. Although the line graph is also faceted by sex, they still use the same x-axis, as is true for the histogram. The line graph allows the user to determine which age groups have had the most success over time.


• Characteristics of Channels: In the histogram color is used to depict the success of athletes in different age groups. The success was calculated by computing the ratio of medals (of all types) won to null medals per sport, per age group. Originally a continuous color scale was used, stretching from light green to dark blue, which resulted in all the colors above a ratio 1 looking like the same shade of dark blue, and many of the frequent ratios near zero all appeared as the same light green shade. Therefore a manual color scale was created, wherein color changes happen fast at small ratios close to zero, and then they happen slower at higher ratios because there are fewer higher ratios. I also specified that above 1 the color scale should change to another color, as 1 is sort of a neutral/central point to the data where the number of athletes who won medals in a specific age group of a sport is equal to the number of athletes who did not win medals. This color encoding makes the plot more effective because it allows the user to see the data at ‘high resolution’ of color. The user also knows that if they see purple there are more athletes who have won medals in a group than athletes who have not won medals, making that age group more ‘successful’. In general the more blue the color the higher the ratio of athletes who have won medals to those who have not won, although still below 1. The color encoding was not kept the same for the line graph; the colors are representing age groups instead of the medal ratio and therefore it would be misleading to have a similar/the same color encoding. Additionally, a categorical color scheme was used for the age groups as we are treating them as a categorical variable because our intention with this plot is to investigate the success of specific age groups over time.


• Interactivity: Initially the user must select a sport that they would like to view the athlete’s age information for via the dropdown widget. This plot also employs bi-directional linking between the bars in the histogram and the lines on the line graph. When a range of age groups is selected on the histogram using the brush, only those age groups are plotted on the line charts. This was executed in code via a transform filter using the brush on the line plot. Conversely, if the user clicks on one of the lines presented in the line graph, this line is made more opaque compared to the other lines, and it also makes the selected age group appear more opaque than the non-selected groups in the histogram.


• Critique: This visualization is effective for finding the top performing ages of male and female athletes for different sports because it displays information expressively and it reduces the amount of time needed to be spent figuring out the visualization by being intuitive to the user. A bar chart is a widely recognized graph, understood by most people so it takes little time to fathom the information it is displaying. Additionally, faceting the graphs vertically to display the male and female statistics separately allows for effective comparison between the sexes against a common axis. The use of the ratio successfully encodes the medal data as one variable represented by color, which is easy for users to process. The use of interactivity in this graph makes it particularly intuitive because it connects the charts within the view to one another. The line graph on the right is thus clearly linked to the information on the left while also providing more detailed information about the plot on the left. This plot allows us to go beyond simply answering the research question by adding the element of time.

Heatmap

• Mark: Since this is a heatmap, the mark used is mark_rect() as this allows us to keep the areas the same, thus not encoding any information within area, but allowing us to change the color of each rectangle to depict the difference between medals won by female and male athletes, normalized by the sum of medals won by all athletes in that sport. This way we can show this difference per sport and as it changes over time, as the research question demands.


• Channels: The x-axis on this chart shows the year in which the olympics took place, while the y-axis reflects the sport the athletes competed in. This is most effective because it allows us to see all the sports at once, over time, with information about medals won and sex of athletes who won the medals - so there’s a lot of data being displayed. Therefore the user can determine in which sports the female or male team of a country have been dominating, and for which sports the success of both genders is roughly the same. The dropdowns allow the user to choose only one country to view, and also the option to choose which season (winter or summer) they would like to view the sports for. We have provided the option to select the season as it allows users who know roughly which data they want to see to visualize less data at a time which makes it less overwhelming and easier to process.


• Characteristics of Channels: The color encoding is being used to show the difference in medals won by female athletes minus male athletes, normalized by the total number of medals won in that sport in that year in the selected country. Since the color scale is depicting a difference there is a clear central point to the data where the difference is zero and the number of medals won by both genders is equal. Consequently we have employed a diverging color scale where the color is gray at zero. Gray at zero was chosen because there are years and sports for each country where none of their athletes (female or male) won a medal, so there is ‘no data’ for that rectangle on the graph, which appears white. Thus the nan values are differentiated from the points/rectangles where the number of medals won by men and women was the same. This color encoding, along with how we have placed the channels, is successful because it allows users to easily identify rows (which are sports) where the color is a dark hue over a longer horizontal range, answering our research question. Since the heatmap’s color itself doesn’t show any information about the total number of medals won by that country (since we have normalized by this), and it might be useful to know, we have included this information in the tooltip.


• Interactivity: The interactivity of this map is mainly in the dropdown menu which allows the user to select which country for which they would like to view the data, as well as which season of sports they would like to see. This makes the graph accessible to people who are just wanting to explore the data without a real purpose, as well as those who want to investigate the details of a season, for example.


• Critique: This map is highly effective for answering the research question as it allows us to see all the sports at once, and we can evaluate the ‘consistency’ of male or female dominance in a sport due to the time component. Additionally, the color encoding is extremely effective because it allows users to easily visually pick out rows where there is a long horizontal line where the color is staying a consistently dark due. This chart’s pitfall, however, is that the user is unable to compare between countries as the country must be selected beforehand. We also cannot infer much information about the performance of the country overall in each sport as we cannot simply assume each white rectangle is where they haven’t won any medals because we have no assurance that they even participated in that year. This information, however, is not required by our research question.