Hello there! I suppose it’s good to spam my sporadic study logs here instead of in the other thread I created about questions about Recommended Learning material along side Hanzihero.
Today I’m coming back after not having done reviews for 4-7 days… and it seems painful starting with 400+ reviews 
Yeah, that’s the tough part about doing SRS 
I remember once doing ~2.2k reviews after a long break when doing reviews within Anki for Japanese.
I’ve not given much thought as to what we can do to make returning easier, but I would suggest to not feel pressured to complete all 444 and maybe do 100 a day as a way to ease back into it. If you have any interesting ideas on how to make the return easier, I’m all ears!
Also my current learning scheme seems to be something like this
I’ve not given much thought as to what we can do to make returning easier, but I would suggest to not feel pressured to complete all 444 and maybe do 100 a day as a way to ease back into it. If you have any interesting ideas on how to make the return easier, I’m all ears!
I’ve been playing around with ideas about it. Maybe something like being able to start review sessions with only X amount of items. Like doing sessions with X=100, and after doing those 100 the session is complete, and you still have Y-X items at that point, where Y was the total number of reviews. But being able to just do 100 in a “boxed” session would maybe make the brain think it’s more manageable. Even though you might do 2-5 of those X=100 sessions in a row, you might feel more of a progress. I’m not a psychologist, so I might not explain it in the best way - but I do think this is how most of our brains work, where it feels much easier to “eat parts of the cake” and then take one more bit, etc - until you suddenly have eaten the whole cake.
Splitting up the work into chunks makes sense. It sounds nice if it were supported at the application level.
Maybe after one completes 100 questions there can be a pop up suggesting to wrap up that session and take a 10 minute break 
–or it can just be an actual dialog that has a countdown before continuing, but that may be a bit invasive for some. I suppose it can be thrown into the application settings.
I am surprised to hear about your experience with HelloChinese! It seems to be recommended quite frequently. I’ve not personally tried it because it’s on the phone. I suppose the main advantage is that it is in fact on the phone, making the barrier of entry lower(?).
I am surprised to hear about your experience with HelloChinese! It seems to be recommended quite frequently. I’ve not personally tried it because it’s on the phone. I suppose the main advantage is that it is in fact on the phone, making the barrier of entry lower(?).
I did learn from HelloChinese, but it was more that it didn’t give me any motivation and felt repetitive. It’s the same feeling Duolingo has given me for other languages. I might go back to HelloChinese for the grammar after a while though, and see if it seems to stick better. I also think my brain goes into another mode while on the phone and I get bored really fast, while on the computer I have much more patience - it seems to me. I did reach the first checkpoint on HelloChinese, which is supposed to be all of HSK1-ish I think. (I think @kevin mentioned a similar fatigue or lack of motivation with HelloChinse and Duolingo in another thread where I mentioned it previously).
I felt much more motivation doing Hanzihero than HelloChinese, but they focus on different things obviously.
Splitting up the work into chunks makes sense. It sounds nice if it were supported at the application level.
Maybe after one completes 100 questions there can be a pop up suggesting to wrap up that session and take a 10 minute break
I was thinking more of letting the user select the number of items before starting the session (or at any time during the session) - with potentially easily selectable numbers like 20-50-100.
@phil Another unrelated suggestion. To me it seems to make sense, but it might have it’s downsides that I do not see right now.
Anyways here are my current upcoming reviews (I haven’t started on the 444 for today yet )
Previously I’ve had something like
Sunday 0-10
Monday 20
Tuesday 100-200
Wednesday 0
etc
I find it interesting to spread the load as to maybe move some of the Monday items over to Sunday, for example if there are some items I’ve learned recently they should probably have priority to stay on Monday to give my brain a little more time “to forget”. But for items I learnt a 30+ days ago it doesn’t matter if I have the review tomorrow or the next day. So could you implement a algorithm that maybe tries to spread items out over an interval of X days, where X might vary depending on the SRS level of the item.
Suggestion of an algorithm:
accepted_smoothing_radius = {"iron": 0, "bronze": 0, "silver": 2, "gold": 3, "platinum": 4}
for day in next_10_days:
if day has significantly less days than the other 10 days:
using the accepted_smoothing_radius "steal" items from other days that are close enough as too smooth out the review distribution
# Maybe for simplicity/mvp add a rule that you should prefer stealing items from the future rather than the "past" relative to the day your "stealing" for
I think smoothing out the review distribution like this might make it easier to keep up the good work, as I personally find it harder to start a review when I suddenly hit 100-200+ items (and I have been doing my reviews regularly unlike right now 
) - and I assume this is representative for most people.
We have a ticket in the backlog for adding some sort of smoothing to the SRS, at the minimum a fuzzing interval where the next review period is +/- X days. This is similar to what Anki does. We’ll look into this some more, I can see us implementing this within the next month.,
Yeah, sounds smart! Distributing the reviews in a truncated normal/uniform/other distribution with a width +/- x is probably much simpler and might even work better - simple logic is always nice to keep the code base tidy.
Following up–
We’ve added some fuzzing to the review scheduling. Generally working within a range of ±2 days–so you should see a healthier spread of your reviews from here on out.
If we find that the max spread of ±2 days isn’t enough then we can see about increasing it 