Is there a indicator of the "level" of a running session

Hlb

Hi,
I am looking for an indicator of the "level" of a session.
For example, take these two running sessions
Session 1 - 1 hour running at constant pace 4 min/km (15 km/h)
Session 2 - 30 minutes at constant pace 6 min/km (10 km/h) followed by 30 minutes at constant pace 3 min/km (20 km/h)
The two running sessions take 1 hour for 15 km in total (4 min/km average pace for both).
But, the two sessions do not correspond to the same level of running, session 1 beeing far easier than session 2.

Is there a Runalize indicator that is able to order sessions according to their "running level" ?

Laufhannes

Heart rate based: You should see a difference in TRIMP if heart rate data is available. There's also an (estimated) Peak EPOC value which is related to the intensity and should show an obvious difference between those activities.
Pace based: There's xPace which tries to estimate your pace for the same effort at a constant level. For your example that should result in an xPace of 3:30/km and a Variability Index of 1.14 (which is very high).
Power based: For (estimated) power, there's also xPower and the Variability Index.

Hlb

Thank you answering.

Clearly, TRIMP is not what I am looking for, because it does not evaluate the session itself, but rather how ones body reacts to the session. Same session will have different TRIMP for different runner and different TRIMP for same runner at different moments.

I will check xPace, it may be the kind of data I am looking for. Thank you.

Hlb

xPace goes in the right direction but is not perfect.
My understanding is that the two following sessions will have exact same xPace, although the second one beeing harder than the first one :
Session 1 : 30 times (30 seconds at 20 km/h + 30 seconds at 10 km/h)
Session 2 : 15 minutes at 10 km/h + 15 minutes at 20 km/h.
(I may be wrong about xpace, but, this is what I have understood for now)

Laufhannes

Yeah, that's right. But I guess then there's not really an indicator for what you're looking for.

Maybe the "W'" (W prime) model helps, but it's not fully enabled/integrated so far. It's based on the 'Critical Pace' model and requires to have custom CP/W' values for running.

If an athlete has entered those values, there's a 'Work capacity' view for an activity and avg/min W'bal are shown. The model assumes there's a 'Critical Pace' which you can - theoretically - hold for kind of as long as you want and a 'Capacity' which you will 'consume' if running faster than your CP. The value of 'min. W'bal' (expressed in %) will tell you how close you were to complete exhaustion.

Runalyze is not able to automatically estimate CP/W' so far and that's why you probably don't have those values (me neither). You can try to set some (for e.g. 1 week ago) and recalculate all activities since then.

Still, this model and the results depend on an athletes' CP/W'. It does still not give a generic 'level', but I don't think there's a method that does.

Hlb

I checked more in my running history with Runalyse in order to evaluate if xPace can evaluate the "running level" of a session.

Finally, xPace alone is not a good indicator of the "level" of the session (but I think I can mix xPace with session duration to get a better result).

I have two sessions (real ones this time !) :
Session 1 : 8 km at 4 min/km
Session 2 : 3.1 km warm up at 6min15s/km + 8 km at 4 min/km + 2.33 km cool down at 5min54s/km

Session 1 has an xPace of 4 min/km.
Session 2 has an xPace of 4min30s/km.

So, if xPace were to be seen as a "running level" metric, it would imply that session 1 shows a much better running level than session 2.

But it can't be the case since session 2 includes a sequence that is the same as session 1, adding a warm up and a cool down. Someone that has the level to execute the Session 2 will of course be capable of doing Session 1, so the running level of Session 2 can't be lower than the running level of Session 1.

Hlb

Thanks for answering. Our two last messages crossed.
Reading your answer, I make this precision for clarity :
I am looking for a metric that could be computed from the speed recording during the session. My Garmin watch stores one speed value every second of a run.

Laufhannes

And based on pace only, right? Independent of heart rate and any other performance indicators? Do I get you right? Something that’s able to compare sessions based on there pace across multiple athletes?

Of course, for world class athletes, 10k at 4:30/km is very easy, for others it’s a race and for a lot of athletes it’s not even possible. But you want to give it some kind of a level, right?

I will need more time to think about how one could achieve something like that.

Hlb

Yes, you have catched the point.

I want to order distance running sessions by intrinsic difficulty, not by the difficulty for a particular athlete.
I restrict to distance running (let's say a minimum 20 minutes of continuous running) and flat terrain, else things would become more complex.

I understand that "mismatch" cases will exist, where, for example, athlete 1 is able to perform session 1 but is unable to perform session 2, while it is the contrary for athlete 2. But, these cases will only occur when the two sessions have almost identical "running level" value.

A way to see the "running level" of a given session can be:
What is the percentage of all distance runners able to perform this session ?

A session that only 1% of all distance runners can do has a higher level than a session 10% of all distance runners can do.

Hlb

A good starting point could be the World Athletics Scoring Tables where each performance on a particular race correspond to a number of "points". This allows, for example, to compare a performance on a 10 km race with a performance on a marathon. But, I would like to compare real running sessions, that often involve time intervals at various speed.

Laufhannes

If anyone has something in mind, feel free to post it.

I'm not aware of anthing else than W'balance (based on Critical Power/Pace) that at least tries to handle fatigue/capacity. To make this independent of a particular athlete, my approach is to calculate/estimate the minimum required CP/W' to do such a workout, i.e. argmin W' such that min. W'bal >= 0 for the whole activity. I need to fix W' in this scenario to have only one missing parameter. Otherwise there are a lot of combinations possible. I used W' = CP / 20, which results in e.g. W' = 1.0 km for CP = 20 km/h or W' = 0.6 km for CP = 12 km/h. Just a wild guess.

As a reminder from what I've stated/linked above:
Critical Pace is the pace that an athlete can (theoretically) run for ever.
W' is the anaerobic capacity (in [m] for pace) which is consumed when running faster than CP. When running slower, the 'tank' is filled again (slowly).
W'balance is somehow the 'live fuel gauge'. For its calculation, see e.g. https://medium.com/critical-powers/comparison-of-wbalance-algorithms-8838173e2c15
CP(t) = CP + W‘/t

For a given pace stream you can simply use bisection starting with an interval from e.g. 5 to 25 km/h for CP and calculate resulting min. W'bal to find CP such that min. W'bal is nearly 0.

I did the calculations for 3 example activities from myself (1) 3x10x30s, 2) 5k tempo run, 3) 5k race). Time (s) is on the x axis, Pace in (km/h) on the y axis. Pace is shown in blue, W'bal in red. The required CP can be interpreted as 'Only athletes with a CP of at least X can run this workout.'. No surprise that the 5k race in 18:55 results in a CP which equals that time (off by 1s due to rounding somewhere I guess). The tempo run (2nd activity) was near all-out, though that's not exactly right as downhill part is not treated correctly. The first activity was not all-out. For athletes with known CP, this metric could yield another metric 'Relative intensity', similar to current min. W'bal but more powerful for activities below CP (for which min. W'bal is 100%).

All of this can be done with Grade Adjusted Pace to make it usable for non-flat terrains. The second activity shown above has a downhill section in the beginning which is not treated correctly with usual pace.

Hlb

Thanks.
I certainly catch the general idea, but unfortunately, I don't have the skill to implement this approach for my runnings.
I assume you are using specific tooling to make this kind of analysis ?

Laufhannes

An implementation within Runalyze will follow soon.