Animating 2021 Cycling Goals with rStrava

The Strava API

As with any other project in R, we will start by loading free open source libraries available from CRAN and Marcus Beck’s github to pull from Strava’s application programming interface (API). I also set my own graphics theme.

source(here::here("_common.R"))

suppressPackageStartupMessages({
 
library(tidyverse) # data manipulation
library(rStrava) # the Strava API
library(ggmap) # one of the mapping APIs
library(gganimate) # animation tools
library(patchwork) # combine charts and graphs
library(lubridate) # handle dates and times
extrafont::loadfonts(quiet = TRUE)

})

theme_set(theme_jim(base_size = 12))

Strava requires authentication to track the api utilization and compliance with their terms of service. This code is taken directly from Marcus Beck’s rStrava help pages.

if (file.exists(here::here(".httr-oauth"))) {
  stoken <- httr::config(token = readRDS(here::here(".httr-oauth"))[[1]])
} else {
  app_name <- "Rapp" # chosen at
  app_client_id <- "58858" # Client ID
  app_secret <- Sys.getenv("rStravaSecret") # Keep client secrets hidden
  stoken <-
    httr::config(
      token = strava_oauth(
        app_name,
        app_client_id,
        app_secret,
        app_scope = "activity:read_all",
        cache = TRUE # creates a hidden authentication file and masks it in .gitignore
      )
    )
}

mykey <- Sys.getenv("GGMAP_GOOGLE_API_KEY")

register_google(mykey)
# Enable the Google Maps Platform Elevations API at https://console.cloud.google.com/google/maps-apis/

The API retrieval functions are called with the token established above. From there we can build a heat map of my activities. The first call is to download the complete list with get_activity_list() and the second is to convert the list into an easy to use dataframe with compile_activities().

my_acts <- get_activity_list(stoken)
act_data <- compile_activities(my_acts,
  units = "imperial"
) %>%
  mutate(id = factor(id))

all_activity_heatmap <-
  get_heat_map(
    act_data,
    key = mykey,
    col = "darkgreen",
    size = 2,
    distlab = F,
    f = 0.1
  ) +
  labs(
    title = "Cumulative Cycling Activity in Strava",
    subtitle = paste0(format(as.Date(
      min(act_data$start_date)
    ), "%B %e, %Y"), " to ", format(today(), "%B %e, %Y")),
    fill = "",
    x = NULL,
    y = NULL,
    caption = "map: Google"
  ) +
  theme(
    axis.text.x = element_blank(),
    axis.text.y = element_blank(),
    plot.margin = margin(0, 0, 0, 0, "cm")
  )

Let’s pair the map with a graph showing the accumulation of miles over time.

Accessing data, exploring it, and understanding what it means as an analyst is still only half of the work. Connecting with an audience to tell the story in a meaningful way is always the more challenging task. Regardless of whether the topic is supply chain risk or health and fitness decisions, the data alone is not enough.

Business school programs cover data viz and leadership communications at some level, often for a c-suite audience. Most every consultant’s pitch deck has the same MBA style building blocks (for Powerpoint).

The emerging field of data journalism offers hard won process frameworks for conveying insights through media to the broader world.

What makes data journalism different? New possibilities open up when combining the ‘nose for news’ and ability to tell a compelling story with the scale of digital information now available. Those possibilities can come at any stage: using programming to automate data retrieval, or using software to identify connections between thousands of documents.

Data journalism can tell complex stories through infographics. Consider Hans Rosling’s spectacular talks on visualizing world poverty with Gapminder and David McCandless’s work showing the importance of clear design at Information is Beautiful.

Often stories begin with resources in open source software, open access publishing and open data. Even within the venerable New York Times there are repos of open-source and even “inner-source” (widely distributed internal package) digital tools. This leads me to a thought on the roles that fill these communications needs in organizations. This short video montage from a spring 2021 conference conveys a broader perspective of what sorts of people do modern analytical work.

Drilling down deeper, the rStrava package includes functions for pulling elevation or % gradient for individual rides, or all of them. Heart rates, travel speeds, and really anything recorded can be plotted over maps from Google, OpenStreetMap, or Stamen (each with their own APIs).

No purchased software was required to access the personal biometric data here. No purchased software was required to parse it or to visualize it. Generally speaking, less cost, less friction and less gate keeping are all ways of creating value. Oddly, legacy corporate and government bureaucracies may still not see it that way, as there is no longer a need for analysts to beg for licenses from an IT department.

When you think about it, the way humans tend to operate mirrors open source principles. Arguably we are naturally inclined to use, share, learn, copy and improve on things together so that other people can be part of a community. It’s not surprising then that it’s been so widely adopted, and because it’s so widespread, the barriers to entry are relatively low if you want to join the club.

Some companies are learning to tap into the open source community for talent and to upskill themselves. They go beyond reluctantly accepting the use of open source software to encouraging participation by their employees in projects.

No company is more emblematic of the shift than Microsoft, which initially waged a legal battle against it. The digital giant now uses open source software extensively. Most of Microsoft Azure runs on Linux. They acquired GitHub for $7.5 billion in 2018 – then the largest enterprise software acquisition ever – and its employees are heavily engaged, with over 5,000 of them contributing to open source projects in 2020.

Other leading companies have established open communities of excellence. They identify the open source software each department in the organization uses, and foster collaboration as well as best practice sharing. Catalyzing exchanges and getting various functions to share success stories help companies realize the full potential of their people.

On a personal interest level, open source lowers the learning barriers and allows for ad-hoc communities to spring up. It gives people a sense of purpose and helps to increase creativity and problem solving skills which favorably influence productivity.

Setting Goals for 2021

Strava offers their own paid subscription feature that includes graphs that let members “visualize their progress and trajectory” as they strive to hit their targets. As always with this kind of functionality, the quality and usefulness of results depends on the inputs. Set reasonable goals and you’ll have a nice visual reminder that you’re on track. Excessive optimism, on the other hand, yields graphs that serve only to remind you that you really need to try harder.

Another approach is to build our own goal tracking tool in R. One approach is illustrated here:

A closing thought:

Comparing ourselves to the success, or even the failure, of others is only natural. Who should we be comparing ourselves to, anyway? Some thoughts from Leonard Lee:

An acknowledgement of inspirations for building this post:

Author and rStrava Creator Marcus W. Beck

Sascha Wolfer published an very nice blog post at rCrastinate on running and is active in the RStats Strava Club

Jack Rozran’s blog rStrava post on visualizing cumulative mileage with ridgeplots.

Chris Woods’s Cycle Eye Visual of total climb and distance in a radial plot.

John Peters has crafted a Shiny app that leverages Strava and other fitness data at EnDuRA

Daniel Padfield with concepts for animations.

Marcus Volz’s creative data vizualizations

Robin Lovelace’s cyclestreets package

References

rStrava R package web page

Strava API

D. Kahle and H. Wickham. ggmap: Spatial Visualization with ggplot2. The R Journal, 5(1), 144-161.