3 min read

How good are fitness watches (pt 2)

Samsung watch did OK at cycling, but cycling is pretty steady - get up to working heart rate and stay roughly there. I wondered how it does at barbell training, so I took it to Thursday’s deadlift session 1. Deadlifts are a few seconds of high effort followed by a few minutes catching your breath.

The Wahoo chest strap wasn’t overly pleased with this concept - it is marketed for runners/cyclists. I got these data out of it by saying I was running on a treadmill. This disabled the GPS, so when I ran the Wahoo data through GPS babel it discarded all the data because babel likes working with nice GPS traces. I ended up deep in the documentation and got a spreadsheet of the heart data.

wahoo   <- read_csv(here::here("static/data/deadlifts/wahoo.csv")) %>% 
  select(timestamp=Timestamp, heart_rate=`heart rate`) %>% 
samsung <- read_csv(here::here("static/data/deadlifts/samsung.csv")) %>% 
  rename(timestamp=start_time) %>% 

data <- bind_rows(wahoo, samsung)

data %>% 
  ggplot(aes(x=timestamp, y=heart_rate, colour=sensor)) +
    geom_line() + 
      title = "Heart rate over a training session",
      y = "heart rate (bpm)"

Looks like the watch was consistently below the chest-strap. Apart from about 20:10 when I went out of range of the phone to grab a drink and the sensor started logging 0s.

errors <- left_join(samsung, wahoo, by="timestamp") %>% 
  filter(heart_rate.y > 0) %>% 
    mutate(error = heart_rate.y-heart_rate.x) %>% 
  select(timestamp, error)

errors %>% 
  ggplot(aes(x=timestamp, y=error)) + 
    geom_line() + 
      title="Difference between chest strap and fitness watch",
      y = "heart rate difference (bpm)"

Last time they were within about 10 bpm of each other, but this is almost never that close.

Recovery time

Heart rate is a reasonable proxy for recovery between working sets. Immediately after a set of deadlifts our subject is too tired to do more deadlifts, and their heart rate shoots upwards, then starts to go back down. If it settled down to a lower level and stayed there then we might say they were resting too long.

There’s all sorts of strategies for “are you resting enough/too much between sets?”. For this test I rested as much as I thought I needed, and eyeing the graphs it looks like I was about right.

wahoo %>% 
  mutate(heart_rate = 
           if_else(heart_rate==0, NA_real_, heart_rate)) %>% 
  #The zeros really should be NA. 
  ggplot(aes(x=timestamp, y=heart_rate)) +
    scale_y_continuous(breaks = seq(100, 180,by=10)) +

The main other option is some form of timer. I am outraged at how much some of these cost and I’m designing one for myself with an arduino. Watch this space.

As I’ve been falling down the rabbit-hole of evidence-based fitness2, I’ve discovered that people are generally OK at self-regulating recovery time. Given that the heart strap is awkward to deal with and the watch was garbage, I think I’ll play with my timer and occasionally check with the chest strap that the timer is set appropriately.

There’s also a possible complication - it was about 3 degrees C in the garage. This makes my skin redder than normal, which affects how easily the camera on the watch can detect a pulse as more red.

  1. before anyone points out, yes England is in Lockdown Mk 3. I have weights in my garage.↩︎

  2. Instead of blogging, obviously. I generated these data 24 days ago↩︎