# 2023 in Exercise

Posted on Dec 31, 2023

Since another year has come to an end, I can’t help but be curious: What does the exercise data have to tell me? I did very similar things in 2022, 2021, 2020 and 2019.

## Executive summary

The following radar-plot compares the past three years in terms of endurance, variety, strength and consistency:

To my surprise, the plot indicates that the polygons were almost subsets of each other. I was rather suspecting a bunch of trade-offs to appear. In other words: the years have mostly improved in these four metrics. :)

Endurance measures how many endurance sessions (e.g. running, cycling, swimming, ..) I performed per day.

Variety reflects how broadly my exercise activities were spread out over various disciplines. Concretely, this is measured by the entropy of the empirical distribution of exercise activities over disciplines. The maximal value is $\log_2(\#disciplines)$ – in this case approximately 4.32 for $\# disciplines = 20$ – and the minimal value is $1 \cdot \log_2(1) = 0$.

Strength measures how many strength sessions (e.g. weight lifting or climbing) I performed per day.

Consistency measures the standard deviation of the durations between two consecutive exercise activities. Note that this truly only reflects the regularity - not the actual value of the duration between two activities.

More background on variety and consistency can be found in these two plots:

The left-hand side plot shows the empirical distribution of exercise activities over disciplines. We see that overall, they were fairly similar in 2022 and 2023. Yet, 2022 had three disciplines which were stronger contendors: core strength training, swimming and the good old elliptical. 2023, on the other hand, only had one remarkable additional contendor: climbing. Very vaguely speaking, this leads to a flatter distribution for 2022 and thereby a higher entropy, i.e. variety.

The right-hand side plot shows a histogram of the durations between two consecutive exercise activities for each year. We see that all three years have their mode around the 24 hours mark. Yet, while 2023 is more narrowly concentrated around that, 2021 lashes out to the right – i.e. there was often just one day between exercise activities but also often way more – and 2022 lashes out to the left – i.e. there was often one day between exercise activities but also often less. Therefore, the standard deviation of these durations is smallest for 2023, mirrored by a high consistency score in the radar plot above.

## Not so executive summary

disciplinemetric2020202120222023
runningdistance [km]1295140311481312
#activities123127127139
share of running activities among all activities [%]58.5754.2738.8939.60
cyclingdistance [km]104466013401264
#activities183641
gym#activities6382113115
climbing#activities01729
all#activities210234324356
#discipline kinds111716

Given that many aspects of my running - long distances vs. short distances, lots of elevation vs. no elevation, injury-free vs injury-ridden, fore-foot vs mid-foot strike - have differed quite a bit between years, I was very surprised to see that the yearly distances in running have been as similar as they are. In particular, this year’s running volume accrued fairly smoothly. It’s blatantly obvious that until my injury in mid 2022, 2022 has been by far my strongest year in running.