How has the Tour de France changed in the last 100 years?

Cycling's premier event has become shorter, hillier and faster paced since it began more than a century ago.

Data

Data analysis can be found in analysis.ipynb

Filename	Description
k-stage_data.csv	Data on riders for each stage
k-tdf_stages.csv	Data on the stages
k-tdf_winners.csv	Data on the winners

k-stage_data.csv (1903-2019)

Name	Description
edition	Edition of race
year	Year of race
stage_results_id	Stage ID
rank	Rider rank for the stage
time	Time of rider
rider	Rider name
age	Age of rider
team	Team name
points	Points for the stage
elapsed	Time elapsed
bib_number	Bib number

k-tdf_stages.csv (1903-2017)

Name	Description
Stage	Number of stage
Date	Date of stage
Distance	Distance in kilometers
Origin	Stage start city
Destination	Stage end city
Type	Type of stage
Winner	Winner of the stage
Winner_Country	Winner nationality

k-tdf_winners.csv (1903-2019)

Name	Description
edition	Edition of the tour de france
start_date	Start date
winner_name	Name of the winner
winner_team	Team of the winner
distance Type	Distance in kilometers
time_overall	Time in hours
time_margin	Difference between race winner and runner up
stage_wins	Number of stage wins
stages_led	Number of stages spent as the race leader wearing the yellow jersey by the eventual winner
height	Height in meters
weight	Weight in kilograms
age	Age of winner
born	Year of birth
died	Year of death
full_name	Full name
nickname	Nickname
birth_town	Town of birth
birth_country	Country of birth
nationality	Nationality

Relevant figures

Length of Tour de France 1903 v. 2019

• Converted kilometers into miles by multipling by 0.621371

Percent change in hilly or mountainous stages

Percent change in flat or plain stages

• Simplified the naming of the categories
• Added a decade column
• Grouped years into decades, 1900-1909, 1910-1919 etc.
• Summed each type of terrain by decade
• Did a percent change calculation compare 1990s to 2010s using ((v2-v1) / v1) * 100
• Dropped the "Other" category as it was often 0 or less than 1 percent

Percent of type of stage by decade

• Added decade column
• Grouped years into decades, 1900-1909, 1910-1919 etc.
• Number of times the stage appears / total number of stages that year

Average speed of winning rider in 1903 v. 2019

• Calculate percent change in average speed by using ((v2-v1) / v1) * 100
• Convert the kilometers per hour into miles per hour by multipling by 0.621371

Average speed of winning rider by decade

• Added a decade column
• Grouped years into decades, 1900-1909, 1910-1919 etc.
• Converted kilometers into miles by multipling by 0.621371

Name, age, country of youngest modern-era Tour de France winner

• "Modern-era" is 1990-present
• Extracted from Tour de France website directly, would have liked to scrape 2018-present data into a smaller CSV and merge it into the larger dataset

BMI

• Weight / Height ^2

Charts

Uphill climb for Tour de France riders

https://app.datawrapper.de/archive/team/theledeprogram/https://www.datawrapper.de/_/qLZ5i/

Pace of the peloton picks up:

Line chart: https://app.datawrapper.de/archive/team/theledeprogram/https://www.datawrapper.de/_/aRsTN/

Horizontal bar graph: https://app.datawrapper.de/archive/team/theledeprogram/https://www.datawrapper.de/_/cq8HD/

To create the graphic in Illustrator

• Created the horizontal bar chart and line chart in Datawrapper
• Exported the charts as SVGs into Illustrator
• Imported an image of a cyclist from The Noun Project into Illustraor
• Appended the image the ends of each of the 12 bars
• Grouped the images and aligned them to the bottom to make one single-file row
• Removed the labels from the line chart and turn it into a sparkline with labels for both the starting and endpoint year and speeds displayed as an annotation
• Removed the labels from the bar chart as there were two different scales
• Created two guidelines for the start and end of the pictogram and increased the size of the grouped SVGs to align with the start and endpoint of the sparkline.
• Changed the fill of all images the middle points on the pictogram to grey

What I learned / Challenges

• The Tour de France website does not have an easy way to download all historical data as a CSV: https://www.letour.fr/en/history
• The map of the full course with all the segments combined together is only available in PDF format.
• Generic information about the route for this year's Tour can be found in a table on: https://www.letour.fr/en/overall-route
• Each stage of the Tour has its own page, i.e. https://www.letour.fr/en/stage-1
• Each segment is presented in a small map using ESRI, but can only be downloaded as an image not as a SHP or KML file
• Reuters, The Guardian were among the few newsrooms to have the individual segments.
• The Strava segments do not match the actual course presented on ESRI.
• Others have attempted to scrap the entire, but many repositories were in R: https://github.com/alastairrushworth/tdf
• There are purportedly some errors in the data presented on the official Tour de France website
• Did manage to find a scraped dataset as a CSV that I imported into Jupyter Notebook from Kaggle: https://www.kaggle.com/datasets/pablomonleon/tour-de-france-historic-stages-data
• Elevation data for each of the stages is not available and requires a third-party tool such as GPS Visualizer
• The types of stages and how they are named changed over the years, which needed to be simplified for the line chart
• Was missing data for 2018-present in some of the CSV files

Other notes

I'm still unclear in Pandas how to do the following:

• Groupby and then filter by a specific column
• When to use dropna in order to remove NaN values
• A more efficient way to group by decade from timestring start point
• Knowing what I know now, I would have liked to try scraping the Tour website myself with BeautifulSoup