Durations, algorithm aversion and getting to 0.4.0
- 3 minutes read - 621 words
In last week’s update I talked in a bit of detail about how we were trying to approach durations. This week we’ve properly gotten started on enhancing those recommendations using a genetic algorithm that modifies the roulette wheel selection algorithm. Which is all a bit of a fancy way to say that if we can see you always run for 30 minutes on a Monday Byrd’s going to use that as the starting point for future recommendations as it recommends runs to get you to your goals and objectives. We’re aiming to do a soft-rollout of the feature this time next week and will let you know when we do.
To the bug report
As always a full list of the bug fixes is on the releases page but here are some highlights
Reducing peak paces and removing progressions
Progression runs are hard. Like, really hard. They’re excellent if you’re targeting a fast marathon time less good if you’re not. And they were showing up too frequently in people’s plans. We’re not sure why but given their limited utility we’ve taken them out for the moment. Likewise the peak paces we were suggesting were ever so slightly faster than they should have been. Our aim with Byrd has always been to err on the side of caution so we’ve pulled them back a little. We’ve reduced the intensity about 5%, which means if your peak pace was previously 05:00 mins / km it’ll now be 05:15 mins / km
Avoiding infinite regeneration loops
We’re pretty happy that Byrd regenerates frequently because it means we’re always recommending a run that works for the used based on the latest known information. But earlier this week a bug got introduced that for a couple of users caused the recommended runs to just generate over-and-over again. We got it fixed almost immediately and have added a new item to our release checklist to make sure we don’t have it happen again.
And the features
Durations as a range
A small step in the right direction with improving how we recommend the length of a run. No-one needs to run for exactly 41 minutes. We now frame the length of the run as a range like we do for pace and elevation.
Visible warm up and cool downs
If you sync a structured run from Byrd to your watch it’ll give you a warm up and cool down (until you hit the lap button on your watch) but on the UI in Byrd that wasn’t showing up. We’ve fixed that so that you can it’s probably good idea to warm up and cool down before and after a session.
Everyday goals now persist
It has personally been driving me mad since launch - and lots of other people too - that weekly everyday goals don’t persist from one week to the next. It’s a faff add them every week. There’ll now persist. A future enhancement will be for them to change based on the runs Byrd is recommending but this is a step in the right direction.
This week marked the second point release since launch. We pushed 0.4.0 today, which we expect to be available to most users on Monday. 1.0.0 is obviously our target since that’s the point we’ll feel there’s a fully stable product. 0.4.0 brings the ability to track how you feel after your run. It’s the first step in capturing more qualitative information and we’re super excited about the possibilities that’ll bring up.
And that’s your lot…
As always, get in touch on email@example.com if you’ve got any thoughts, comments or suggestions. We’re only going to be able to make this thing fly if we get input from every angle!