Garmin + Data + R = Combining Hobbies

-

 


I'm considering trying to go for the Fastest Known Time at a 60 mile route in South Florida near me. Here is a link to the route I've been looking at. The idea is simple: no official race course, no guides, water stations, aid stations, race fees or any structure - just go run as fast as you can! The current FKT is just under 10 hours...so we'll see.

From the site: 


"A near perfect 100 kilometer loop along an Everglades boundary levee in South Florida...Big challenges due to heat, humidity, sun exposure, and effectively no access to drinkable water. Gator sightings are near guaranteed."

I've decided to keep a closer eye on my eye running data to make sure I'm training effectively to run the 60 miles and to identify any potential health concerns before they become serious. Fortunately, by wearing a Garmin watch during all of my training, this data is cleaned and aggregated already - with a significant amount of interesting data points (cadence, distance, calories, time, vertical ratio, etc..). 

On Garmin connect it is pretty easy to download all activities as a .csv and then move into R. 

Data Aggregation in R

First, I read in the data and have to perform some simple cleaning
  • Remove "," from calories burned
  • Remove "," from distance
  • Convert date from a timestamp to date
  • Aggregate activities to the day level
After cleaning the data, I create some simple calculated fields to be used in the analysis. 
  • Day of Year (number day)
  • Miles Per Hour
  • Days since New Year
  • Cumulative Distance by Year
  • Pace

 df = read_csv('Activities.csv', col_names = TRUE,guess_max = 1000)
	
 running = df %>%
      filter(`Activity Type` %in% c('Running','Treadmill Running')) %>%
      mutate(cleancal = gsub("[^0-9]","",Calories)) %>%
      ungroup %>% 
      mutate(fdate = round_date(Date, unit="day")) %>%
      mutate(cleancal = gsub("[^0-9]","",Calories)) %>%
      mutate(cleandist = gsub("[,]","",Distance)) %>%
      group_by(fdate) %>%
      summarize(distance = sum(as.numeric(cleandist)),
                cals = sum(as.numeric(cleancal)),
                runtime = sum(Time)) %>%
      mutate(DayOfYear = strftime(fdate,format="%j")) %>%
      mutate(secondtime = as.numeric(runtime)) %>%
      mutate(MPH = (distance/secondtime)*60*60) %>% #convert to MPH
      mutate(YR = as.factor(year(fdate))) %>%
      mutate(NYD = paste(year(fdate),"-01-01",sep="")) %>%
      mutate(Days =  ymd(fdate) - ymd(NYD)) %>%
      arrange(fdate) %>%
      group_by(YR) %>%
      mutate(cumsumd = cumsum(distance)) %>%
      ungroup %>%
      mutate(PACE = (secondtime/60)/distance) %>%
      mutate(PACET = paste(floor(PACE), round((PACE-floor(PACE))*60),sep=":")) %>%
      mutate(PACET = lubridate::ms(PACET)) %>%
      mutate(PACETM = paste(PACET@minute,ifelse(PACET@.Data<10 ata="" paste="" sep=":">%
      mutate(PACETM2 = as.POSIXct(strptime(PACETM,format = "%M:%S")))

Plots!

All of the plots were created using ggplot and the code is at the bottom.

Some of the more interesting takeaways: 
  • Speed looks to have decreased over the years, and peaked during Ironman Chattanooga training
    • I think this is thrown off because I've been much more attentive to using slow long runs for training more recently. The top end speed looks to still be there based on the late 2020 runs! 
  • 2018 and 2020 are eerily similar in terms of running distance and distance by day
  • 2019 was the biggest running year - makes sense given the Ironman training
  • Regardless of the distance, pace is relatively flat. This seems to suggest I should be able to run faster on those shorter runs! 






       

#pace (Mph)
p1 = ggplot(data = running, aes(x=fdate, y = MPH, color = distance)) +
  geom_point() +
  geom_smooth(color = "orange") + 
  labs(x = "Date", y = "Miles Per Hour",title = "Running Speed Over the Years") +  
  ylim(6,10) + 
  #geom_vline(xintercept = ymd('2019-08-03'))
p1   #overall speed trending down, due to intro of slower training runs. More overall high speed runs

#cumulative each year
p2= running %>% group_by(YR) %>% mutate(cumsum = cumsum(distance)) %>%
  ggplot(data = ., aes(x=fdate,y=cumsum,group=YR,color=YR)) +
  geom_line(lwd=1.3) + 
  labs(x = "Date", y = "Cumulative Distance (Miles)", title="Cumulative Distance by Year")
p2
#cumulative by each day of year
p3=ggplot(data = running, aes(x=Days,y=cumsumd, group=YR,color=YR)) +
  geom_line(lwd=1.4) + 
  scale_x_continuous() +
  labs(x="Days in", y = "Cumulative Distance (Miles)",title="YoY Running Distance")
p3
#pace vs. distnace
p4=ggplot(data=running, aes(x=distance,y=PACETM2,color = fdate)) +
  geom_point() +
  scale_y_datetime(date_labels = "%M:%S") +
  geom_smooth(color="red")+
  labs(x="Miles", y = "Pace", title = "Running Distance vs. Pace")
p4
p5=running %>% filter(!fdate == ymd('2017-09-30')) %>%
ggplot(data=., aes(x=distance,y=PACETM2,color = fdate)) +
  geom_point() +
  scale_y_datetime(date_labels = "%M:%S", limits = c(floor_date(Sys.time(),"day") + 300,
                                                     floor_date(Sys.time(),"day") + 700)) +
  xlim(2,20) + 
  geom_smooth(color="red") + 
  labs(y = "Pace", x = "Miles",title = "Running Distance vs. Pace (Zoom in)")
p5

#save plots
ggsave("mph.png", plot = p1, width = 8, height = 4, dpi = "print")
ggsave("yearcum.png", plot = p2, width = 8, height = 4, dpi = "print")
ggsave("yoy.png", plot = p3, width = 8, height = 4, dpi = "print")
ggsave("distancepace.png", plot = p4, width = 8, height = 4, dpi = "print")
ggsave("zoom_distancepace.png", plot = p5, width = 8, height = 4, dpi = "print")

Comments

Post a Comment

Popular posts from this blog

Homeruns Are King

-
Image
Summary It’s a Tuesday night, the Cubs game is postponed, and I should either be working or doing homework. Instead, here I am exploring homeruns in baseball because it is way more fun. I’d argue that aside from seeing something historical (no hitter/perfect game), most fans want to see home runs when they go to games. In this analysis, I perform some data cleanup to calculate raw numbers from the %’s that fangraphs provides and shift around some data to build a predictive model to predict homerun totals. The inputs to this model will be the players previous season statistics. Homeruns are becoming increasingly important in baseball, and if you read my analysis on walks, you’d remember that they are being hit at an alarming rate. Teams are building their teams around power hitters, but want to be fiscally conscious. So if a team can identify a future homerun hero and save some cash, it’s a major win. Data Support To begin, let’s take a look at the league leader and league aver...
Read more

Visualize and Understand Your Golf Game

-
Image
Every golfer wants to shoot a better score. Every golfer also thinks they know what they're bad at or need to improve. What if you could challenge those assumptions? Instead of saying "I always slice the ball and miss the fairway right" - you should be saying I miss the fairway right 27% of the time, hit the fairway 40% of the time and miss the fairway left 33% of the time. This data is more actionable. Instead of focusing on "missed fairways", now you are focusing on why you miss left 33% of the time. Instead of saying "I'm not any good at putting" - you should be saying I 3 putt 10% of the time, 2 putt 60% of the time and 1 putt 30% of the time. Now instead of generally being bad at putting, you know that in order to get better at golf, the first focus is on the 10% of holes you 3 putt on. This data can be measured and tracked over time, so you know if you are getting better. Once this is in a comfortable range, you can focus on decreasing ...
Read more