Tx! If you want to analyse cycling data and do machine learning and so on you are on the right track with Python. Its easy to get started. If you want to do more web stuff Javascript and Nodejs for server side work are great but I wouldn’t pick JS for data analysis or ML.
The Intervals.icu back-end is Java which is a awesome language for production applications (there is a well documented and solid lib for everything). I can support a ton of users on very modest hardware and do all the data analytics and so on all in one simple app. But its not as nimble as Python for exploring data and new ML stuff tends to be in Python first, Java libraries come later.
Hi @davidtinker, could you please look into load calculation for two of my activities Intervals.icu and Intervals.icu. One is 1h28m @ 137HRavg, another 1h33min @ 150HRavg, but load is calculated as 100 and 95, respectively.
Hmm. The “time in HR zones” model for training load isn’t working very well for you. I suggest changing to the “Average HR” model in /settings it is likely to work better. Its a very simple linear model.
I suspect it is because you have being doing a lot of base so the model doesn’t have a lot of Z5/6/7 data to work with. The under estimated run had a lot of time in those zones.
I am going to add CSV download of activity data soon for people who want to try build models to predict performance and so on. Of course it will be only your own data.
My Garmin watch automaticly uploads my walks to Strava, these are then synced to Intervals.icu and the “TSS” from these walks are included into my calendar. I don’t want that. Is there anyway to only sync bike rides, or ignore walks?
Thank you. Will it adjust after time with more work at higher intensities? Also is there a description somewhere how different training load options work?
No but you can edit the training load on the walks to 1 or zero to take them out of fitness. I have a todo list item to make the contribution of training load to fitness and fatigue configurable per activity type. So gym sessions only contribute to fatigue as they don’t help fitness etc.
Another question on Load calcs (but from power this time, not HR): something I’ve just noticed, highlighted by the Training Plan Compliance value displayed for Activities, is that the Load calcs for my recent TR workouks are materially (~10%) lower than the TSS number TR calculates.
eg. this week:
36(Intervals) vs. 39(TR);
73 vs. 80;
100 vs. 109
Shouldn’t these be the same or almost the same? TIA.
They should be very close if your FTP is the same in both systems. Intervals.icu calculates those in the standard (Coggan) way. Please msg me a link to one that isn’t right. Tx.
I’d manually lowered my TR FTP number a couple of times (inactivity + sickness), but not reflected that drop in intervals.icu.
In my defence, I had remembered to enter a lowered eFTP number (but this is used for a different purpose), and as I’ve mentioned before, I do find the entering of FTP number in intervals.icu a bit unintuitive and “hidden”, it being recorded as an Activity attribute (which then applies to future Activities also) vs. it being recorded as an attribute of me the punter…
David – how do I reset preferences so that all rides default to the estimated FTP at that time?
When I reset duration to 600sec I got FTPs that lined up with Xert – but, that skewed everything with the Fitness/Freshness charts.
Is there a way to undo any manual FTP yutzing and have the whole thing just work off of estimated FTP and let the model do its thing without user interference?
You can’t at the moment unless you use the “Edit” on the calendar page in chunks. I will add an option on that dialog to use eFTP. Should get that done tonight.
