Low predicted running load

I’m not sure if this a bug or simply an issue of how I’ve set up ICU. I’ve been recreating my planned running workouts in ICU using power values (I use a Stryd footpod and manually added in their workouts) to forecast training load. What I’m finding is that ICU seems to considerably underestimate the load. This seems most apparent when I run a steady/easy/long run - I end up with a load at 130-140% of predicted. When I plan a workout I give a power range so this could impact the prediction (I usually end up with an average power at the higher end of the range). However if I create a workout with the actual power used it still gives me a predicted load score lower than I actually achieved. I’ve set my Training Load priority to Power/HR/Pace.

Appreciate predicting a load has lots of variables but unsure if I’m doing something wrong with my setup. Wanted to make sure I’m balancing my running stress but currently this isn’t really possible. Any ideas?

I have seen the same thing using HR zone ranges: ICU seems indeed to considerably underestimate the load of a workout. Most of my activities end up with a compliance of 140%-170% when using zone ranges.

When using a range, 100% will be the midpoint of the range. Here’s my example of a (cycling) workout in the VO2 range.

This was completed as 100%


The previous week was a struggle, but showed 95% completion because I was in the lower half of the range.

Perhaps share your plan, and the workout, and let’s see if you have any other problems worth noting.

1 Like

Thanks, that actually makes lots of sense.

This is a typical example:

Workout:
image

Activity:
image

The plan shows 31 load, and the actual was 51. This is the 164%, which is 64% more than the plan. So that’s where the % compliance is derived from.

You are using TRIMP, and it might be better to use HRSS. See the explanation below on the settings page.

1 Like

Hi @Gerald my issue is slightly different as am using running power. See an example screenshot below.

I appreciate that I’ve run an average power that is at the higher range of the interval but unsure if this is the sole reason that load is off (in this example its 126%). The time that is automatically created for the workout is too short. Aware that ICU doesn’t have enough data to predict how fast I run (interestingly the stryd system has algorithms that predict this really well) but there is no way when I set a workout to change this time to something more appropriate.

I had thought about just adding the expected load to future workouts (I can estimate what this likely to be) but I then miss out on ICU showing me future mileage and what zones I plan to run in.

Not sure I’m the right person to ask about running with power, as I have zero experience.

However, power is in watts and normalized power is used to calculate load. Those spikes in the power graph will add a little more than what average power would, thanks to the formula going into normalised power. I don’t mean spikes as in 1000W compared to average, but the slight surges.

The screenshot is low resolution, so it’s difficult to see the data clearly, but it looks like you were above average for most of the run plus the bumps in the beginning and end. These all add up, especially over the duration of your run.

If this is a standard run, that you do often, I would simply look at the average load on these types of runs, the pace and duration, then add this into the workout, but not create the actual workout. Or, option 2 would be to drag this workout into your library, and then use it as a workout for the future.

Here’s my library with past races and last workouts. If I want to do a workout similar to that race/ride then I use that load.

Example, for tomorrow, it’s a 4.5h ride with some mixed-intensity. Nothing structured, but with a plan to add race-pace efforts and some time above/below threshold (not quite Tabata, but similar). I know the planned load is about 240, so I look for previous rides of a similar distance and duration.

Thanks @Gerald for the explanation. Happy to do what you suggested (it’s what I used to do). However adding an activity this way doesn’t seem to give an option of distance. With running knowing upcoming mileage is helpful to visualise. Is there any way of recording this without creating a workout?

You can add the distance in the Add Entry process.
Click on the Add Step, then add your predicted duration, and the distance.
You can add an planned load, based on history.

The summary information on the calendar will show the distance.
In my case, it automatically created 2h38m, which I think comes from the Run eFTP of 300W (for me it’s a made-up number only.
The load can also be added in manually, based on previous activities of similar duration/effort.
image

You can also add the weekly summary in the total’s popup
Click on the weekly totals, and then select Run.
You can enter total Load, Duration and Distance, plus notes.

Just created a workout from a completed workout. Used the same average wattage and time. The load came out the same.

Screenshot 2022-09-24 at 7.04.52 AM

The same run on Stryd PowerCenter… (Note: 100w in Stryd = Average, ICU shows 112w Normalised, 100w Average too. Training Peaks shows 55TSS which is isn’t too far off from intervals.icu)
Screenshot 2022-09-24 at 7.08.18 AM

Quite possibly the planned workout in intervals is using your MIDDLE boundary of power. your chart shows 231 - 267w, but your actual workout is at the HIGH end at 262/260w hence inflating the Load. You can try what I did and do a planned workout at 260w and see the resultant.

Note : For Intervals.icu, the load is calculated based on the AVERAGE watt from your plan eg: (267 - 231) / 2. This is my guess that what it is differnet.

ps: not really a runner, per say, (prefer trails) so, don’t have that much data on stryd or power or paces

Thanks @app4g. I’ve tried comparing a planned workout with accurate power with what I actually achieve and its still quite different. Below is a screen shot of my run this morning which gave me a load of 100 with an average power of 252. When I create a workout using this power I get a load of 86.

The issue seems to relate to time running. ICU calculates that running this distance at this power should take 1h42m when it actually takes around 2h. If I create a workout based on the right time (2 hours) then load is correct - 100.

As mentioned above I wanted to add distance (rather than time) as I like to keep track of upcoming mileage. @Gerald suggestion of simply adding the appropriate load seems the best solution for me (I didn’t realise I could do that). It doesn’t allow me to alter the duration though which seems fixed based on distance and power.

1 Like

As you allude to, adding the load based on a previous effort of similar effort/duration works for me. Here was the screenshot from the plan which I shared yesterday, and the actual ride from this morning.

Once you find what works, go with it.

You have specified Z1-Z2 for the planned workout. Intervals.icu will calculate the load based on the HR in the middle of that range which is 135. You averaged 147 for the Run so it makes sense that the load is higher. You could try specify the workout as Z2 (not Z1-Z2) and it will likely much more closely match what you did.

1 Like

There is. You can specify power and pace targets for the workout steps. You can still do the workout with power but Intervals.icu will use the pace info to better estimate the duration and the load will be more accurate.

Main set
- 13 miles at 65-75% 69% Pace

That come out at 1h59 and load of 98 for you which closely matches your actual run (1h59 and 100).

That’s perfect David. Didn’t realise this was an option and solves my issue. Will add a sensible pace percentage to my future steady state workouts.