Threshold Feature Analysis

Last week we introduced our first performance centric feature in the app called Power Factors.  It analyzes the peak thresholds over the last 8 weeks to identify what variables are having the most positive and negative impact on your results.   We present the results in donut graphics depicting the direction (green-for-positive or red-for-negative) and magnitude (size of the segment) of impact each variable (e.g. Cadence) has on your peak power.

Consider my 1 minute power example below.  Apart from more red (things hurting my power) than green, you can see that cadence is having the largest positive impact.  My biggest negative factor comes from altitude which is unique to my 1 minute power for some reason.  It was very interesting to discover my weight has a noticeable negative impact on my one and five minute power.  Given I weigh in once a week and have my Withings scale send it to Strava I’m able to see what impact weight plays in the power thresholds.  Any variable that is constant for the 8 weeks is eliminated.

We create one of these graphics for each of the 1, 5 and 20 minute durations.  For now, Power Factors uses data from every ride over the most recent 8 weeks in the analysis.  We’re testing the impact of slicing off only the rides where you had a Threshold Achievement (1st, 2nd or 3rd) best power to see if that gives a more specific focus to what variables are impacting your results.  The number in the middle is the magnitude of the impact.   The specific number isn’t important per-say other than it represents the degree of the impact while the color conveys the direction.

We include a link to this insight in the Peak Threshold notifications (if they include power) and it’s always available here.

We’re testing the same concept for run pace thresholds.  I’m curious to see what impact the power measured by Stryd is having on pace.  We can also measure what factors are impacting the running power just like we do here with cycling power.  I’m looking for some beta users with Stryd power meters and 8-12 weeks worth of runs we can use to test this out.  Please ping me if you’d like to offer to be our test subject.

Ultimately the vision is to do this analysis with much more contextual information.  We’re researching what is the best source of additional context.  At first blush I like Apple Health as it’s a clearing house of information that all parties can read and write to (with end user permission).  The idea here is to start to understand what factors outside of those we gather from our Garmins are impacting our performance.  For example, how does sleep, blood oxygen levels, hydration level impact our performance?