Add option to weight students during shuffling

[wpovell]

Talked to CS15 TAs who mentioned this was one reason why they didn't use SMU.

[gregcarlin]

Do the CS15 TAs currently do this weighting? How?

[athyuttamre]

++Will for asking the TAs.

To echo Greg, could you expand on how they actually weight students? What's their actual end-goal with the process?

Usually TAs request a very specific feature, whereas the root problem can be solved with a more broad solution. Digging in a little will help. Happy to hop on a call with the TAs too if we want to coordinate.

[wpovell]

In their current re-implementation of signmeup using google sheets, I believe they prevent students from signing up for 1 hour after they were last scene. I'm not sure what their ideal implementation would be though, something to ask about.

[athyuttamre]

That’s interesting. We have the signup-gap feature which does what they do now. I’m not sure I understand what we want to weight students on and how that would affect their place in the line?

…

On Sun, Oct 8, 2017 at 7:41 PM William Povell ***@***.***> wrote: In their current re-implementation of signmeup using google sheets, I believe they prevent students from signing up for 1 hour after they were last scene. I'm not sure what their ideal implementation would be though, something to ask about. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#207 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABaqJq0UyptNX4mUuTo2WPCT4NgyieU8ks5sqYfpgaJpZM4PhrHw> .

[gregcarlin]

Hijacking this with more detail, per a request from twd for CS33. He's looking for the ability to weigh students who haven't been to hours as much so they appear higher (possibly in expectation) after shuffling. Two thoughts on possible implementations:

• Shuffle randomly, but expected position is somehow proportional to amount of times they've been to hours (for that course) over the semester. This could be done without too much effort by representing a student multiple times in the preshuffled list if they haven't been to hours much and then choosing the highest position in the final list, but there may be a more efficient solution.
• Deterministically break students into strata, where a student who has been to hours less is guaranteed to be placed above another student who has been more. Only shuffle randomly within these strata.