I haven’t really seen much about Workout Levels V2 (or the classification of unscheduled / unstructured activities) in a month or more. I believe Nate mentioned it in the unlisted Youtube video (about Legacy Pricing Update around 3.5 minutes in) as possibly being available for private beta soon. I understand the availability of these types of things are hard to predict, but was just curious if anyone in the know might be able to give us an update on the status of this much anticipated feature 
2 Likes
We have a validation issue 
. We’re working on that now. We’re like 95% of the way there, but that extra 5% is hard!!
39 Likes
What does “validation issue” mean for those of us not in the tech world?
1 Like
An issue was found. Could be a bug or could be they identified scenarios not covered in testing done so far.
Hopefully, they can reproduce it if it’s a bug (the worst ones are spurious and tough to troubleshoot) then find its cause and issue a fix.
1 Like
A few computer scientist friends of mine told me that there is a saying amongst programmers: the first 90 % take 90 % of time, the remaining 10 % take the second 90 % of time …
No worries, get it right and then release it. I’m pretty sure you have enough willing beta testers out there.
37 Likes
validation is the final step in releasing software. it’s where the entire system is tested for reliability and high confidence in all systems working well together. It can take a while to do, especially if issues are found that are inconsistent, but cause major issues, like system crashes, data loss, etc.
Just because there are issues does not mean that there is a lot of rework, but it sounds like they are still trying to root-cause the problems, which means an indeterminate amount of development time and then re-validation…
tldr: It’s in debugging, but that could mean a month to 6 months or more before it’s ready for release.
2 Likes
42 Likes
I think it’s more like “because the only way to get the initial funding approved was to make it sound easy”
6 Likes
The amount of times I’ve said sure thing boss, it’ll be an easy job and then perhaps unsurprisingly it really wasn’t an easy job!
2 Likes
Whenever I ask the developers on my team to break down and estimate some piece of work I always tell them to multiply their estimate by some factor depending on their level of confidence, and the factor for “high confidence” is 1.25, because no matter how confident you are, it’s almost certainly going to take you longer than you think.
3 Likes
Multiply by 2 and raise to the next higher units. 1 week “I can knock that down” becomes 2 months minimum 
4 Likes
But seriously, software estimation is hard and messy, and although I want to see workout levels V2, red/green light and all the other stuff TR is working on as much as anybody, I also know I’d be upset if they released something buggy that messed up my data or my training, so I’d rather they take their time to get it right and work out the kinks instead of trying to hit some arbitrary deadline and burning out their team in the process.
3 Likes
My mental image of this is bridging a gap and making it to the last 15 meters where, somehow against all physics and aerodynamics, the wind resistance gets so much harder to overcome.
When I saw your gif, one particular TNG episode popped into my mind, the one where they free Scotty from the transporter buffer, something which he had rigged up to “keep him in stasis.” He tries to help around the ship. He stands next to Jordi and Captain Picard when Picard asks Jordi how long a repair would take. Jordi answered X. Afterward Picard left, Scotty asks Jordi how long it’d really take. Jordi, confused, answers X. 
5 Likes
And Scotty’s response is “how are you going to get a reputation as a miracle worker?!” Such a good episode.
4 Likes
Yup! By far my favorite “nostalgia” Star Trek episode (i. e. an episode that had some intersection with TOS). 
1 Like
and the last 0.5% is even harder …
Zeno’s paradox applies to sw projects…
2 Likes
Model Validation means feeding a model an input and checking how well the model output conforms to “reality”. In this case that could be as simple as feeding WLv2 a lot of custom workouts and seeing how it score those workouts compared to human scoring. Or as complex as having folks do “free form” workouts and having WLv2 score those and have humans “validate” the scoring.
The problem in this case is the the further from a planned workout the test workouts get (think a group ride) the harder it is to come up with the “reality” score that you are comparing WLv2’s score against
1 Like
Like training the last 5 % are the hardest.
