This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Pacing for Trail/Mountain: PacePro unusable, VirtualPacer keeps resetting

Use Case: Running a mountainous run/race, and I wish to pace myself to get a good time.

Solution 1, VirtualPacer) External map plotting tools such as plotaroute have good (close to realistic) pacing formula to take into account elevation. When I export a route from plotaroute, and copy it into my Forerunner 955, I can use the "VirtualPacer" feature to compare myself against this pacer.

Issues:

- Whenever the watch detects "Off-Route", which can happen due to a short d-tour, or simply just because the watch GPS was a bit confused for short time in tricky terrain, the Virtual Pacer is reset to my position: Basically my pacing strategy is no longer valid.

- If I try to use "Race an activity" using the file from plotaroute instead, this resetting does not happen, but allas, I sacrifice any form of navigation and "Up Ahead" features.

Resetting of Virtual Pacer when going off route has been a (complained about!) issue on 945, Fenix 6/7, Edge 1030 etc. devices since about 2 years. Please fix this long standing pain point!

Solution 2, PacePro) PTThe more modern approach to (race) pacing has been PacePro, sadly it is not usable for mountainous races at all, as it completely underestimates elevation, giving much too fast splits for uphill, and laughably slow splits for the downhill + flats. There is a slider to push uphill effort to "easy", but even pushing that slider all the way to easy, the uphill is still completely unrealistic and the flats still much too slow! So I can't create a PacePro strategy that makes any sense...

Example: https://connect.garmin.com/modern/course/110436333 33km Trail Race with 1600m ascent. First a steep ascent, some meandering on trails, some downhills and finally a few kilometers completely flat in the valley to finish. Winning times in good conditions for men have historically been around 3h30min. Pace Pro strategy for 3h30min suggests running a pace of 5:45min/km on the final 5-6 flat kilometers, if I slide the slider all the way to "easy". The reality (splits from the race) shows top males are running around 4:15min/km splits on these kilometers, see for example https://www.strava.com/activities/1643886941/pace-analysis

Here is the comparison of PacePro (with easiest possible uphill setting!) compares to a real-world 3h37m guy (the above strava link):

We can see the pace is consistently overestimating our champion on the up-hills (suggesting 2-6min less per kilometer at times!). We can also see the same error in the other way for the flat and the downhills. This is not something particular about this course, it is a consistent failure of the pacepro formular when considering elevation gain. Suggesting things like a 9min kilometer that contains 300m of gain is world record pace ( https://en.wikipedia.org/wiki/Vertical_kilometer ), and completely ridiculous to suggest as a feasible race pace for a 3 hour+ race for any human.

  • The race prediction for an ultratrail is also very funny. Even Kipchoge can't run that fast in the flats. Joy

  • Another way to see the scheduled pace using AppBuilder:

    setv(1,'8:38'); setv(2,'10:46'); setv(3,PACE);getv(LapCount+1)


    Although it does require setting Auto Lap to manual mode and pressing the Lap Key upon reaching each waypoint

  • I've taken this seriously Grin

    and I have made a PacePro that works, although it requires some patience... and create a couple of fields in AppBuilder 5

    the result is shown in this image:

    1. Create or upload the route to PloraRoute. Define Pace Plan. Include several waypoints strategically (refreshment stations, summits, finish line). Name the waypoints with the expected time of arrival (including stops if you prefer). Something like this: "0h50 Aid Station"

    2. Enter the distances and times of arrival at each waypoint in an Excel template and calculate the Paces between waypoints in km or miles/h, and time of arival at each waypoint in seconds.

    3. With this same Excel template, the formula for the two AppBuilder 5 fields ((a) and (b)) is easily calculated:

    PACEPRO (format PACE): setv(1.6.953);setv(2.5.489);setv(n,LAPPACE in km or mile/h);getv(LapCount+1)

    ETE AT NEXT (format TIME H:MIN):
    ((distanceToNextPoint/(lapdistance / laptime))+ElapsedTime) /60 + ' | ' + (setv(1.2280);setv(2.5100);setv(n,ARRIVALTIME in sec);ElapsedTime + (distanceToNextPoint/(lapdistance / laptime))-getv(LapCount+1))/60

    In order to work properly, AutoLap must be set to OFF Mode, and the LapKey must be pressed after each waypoint is reached.

    Here you can see a screenshot of my Excel template:

    Cheers!

  • Nice work! Not so easy to use that I would start using that, but maybe Garmin can take some ideas from here and see that if users are willing to go so far to get pacing for trails they could fix their pacing solutions.

  • Course navigation gets thrown off by sharp turns - that's when it often goes off-course. It is a 100% artificial algorithmic issue which has nothing to do with GPS accuracy. I investigated this in one of Fenix 7 forum threads. I was able to reproduce this issue in 100% of cases when approaching a 180 degree turn, but it often happens on generally sharp turns. The same issue applies to tracking Strava live segments.

    If Garmin fixes this issue then PacePro will become more reliable too. 

  • I absolutely agree. 
    I is time for garmin team to incorporate some AI in garmin connect. 

    AI for pacepro, which learn from runs on a similar terrain to estimate the best pacepro strategy.

    AI with adaptative trainning plan (similar to the TrainAsOne work) 

    Actually, even without AI, a good determinisitc approach can be better than what is currently existing and not functionnal.

  • You do not need AI to improve prediction a lot! See for example https://forums.garmin.com/sports-fitness/running-multisport/f/forerunner-965/335446/pacepro-how-broken-is-pace-pro-with-elevation-e-g-trailrunning - here the tool "Plot-a-Route" predicts much better splits, by using a very simple formula that calculates a "flat equivalent" distance for each split.

    Probably you can improve on this simplistic model, but since this simple model already kicks PacePro's ass, maybe implement this first, before going for something fancy.