# Training Load Error % curve - what can I learn from it?

I’m fairly new to Intervals (meaning in no way an advanced user!) and took some time to look at the “Training load error %” curve after a ride today (found under my specific activity today, then Activity HR).

Does anyone have more information on interpreting this curve (there is some [brief] discussion about this on this forum, but it didn’t seem to be ‘introductory’ which I suppose is what I’m looking for!).

The graph depicts Training load error % (Y axis) vs. date (X axis) where I have ~80 rides in the past ~18 months, with a Y range from -30% to +50.

‘Helper text’ next to the graph is (note, slight formatting changes) …

"This activity has power data so its training load (79) has been calculated using the standard Coggan formula. If power was not available, the training load from HR data would be 80 (based on the model).

Training load for this activity has been estimated using moving time in each HR zone. Your previous activities with power and heart rate are used to build a linear regression model to estimate training load.

R-squared: 89%

Each dot on the chart is from an activity used to build the model. The y-axis shows the difference between the estimated and actual training load for each activity as a percentage of the actual load."

… so what other interpretations can I draw from the graph?

That is, am I correct to assume that my R-squared (0.89) is fairly high, indicating that my HR data is closely indicative of my training load? I do have a power meter (so don’t necessarily need to rely on HR) but it’d be nice to understand a bit more about this.

Or is Intervals learning the correlation between my HR and power to come up with a relationship? (If that’s the case, are there further results displayed?)

What else can I learn from the graph?

Thanks!

R squared is a mathematical number that calculates the reliability of a theoretical model. It´s not specific to Intervals or endurance training, it´s a widely used mathematical indication.
The higher the number, the more the data is copliant to the used model.
Wikipedia will certainly give you more info on what it exactly is.

R-squared was actually the easy part of the question (thank you for the review - matches my understanding) … I suppose I’m curious as to what the 0.89 (R-squared) correlation means (presumably, that my HR load closely matches power data … and I suppose that if I’m cycling somewhere with HR and without power, my HR load data will be pretty accurate?)

It does indeed mean that the HR load calculation with the model built by Intervals matches quit well with the load calculated from power data.
A similar ride without power meter will yield a HR load that is very close to what it would have been with power data.
Intervals continuously adapts the model from activity data with both power and hr, so that it follows performance improvements.
If you do something completely different, the match may be less accurate. Rides with lots of power bursts can still be off quite a bit. But that´s related to HR lag and can´t be solved with a power/hr model. My experience is that it works very well for anything endurance and longer tempo, even threshold. That´s the scope where power/hr relation is linear. Just make sure to have enough input data for the model by regularly using both power and HR.
Can be very usefull in case you have multiple bikes but only one with power meter.

2 Likes