New beta SRS algorithm -- does it actually work?

If I understand correctly, with the new beta, if I get an item wrong, I won’t see it again until the next day.

If the SRS algorithm says that I can only remember something for 5 minutes or two hours, you make the interval “next day” anyway.

Does that actually work? How can I learn new items this way, without reviewing them extra myself? Isn’t that the job of the app to make sure I learn and review things on time?

1 Like

I would be interested in this as well.

Also, when one chooses a certain number of cards to test from a deck section, is there SRS applied to that section, or only when one selects the whole deck to review?

1 Like

Test mode! The idea is not to remove power away from the system but to make it more organized and focused. The proposed study flow for a day would be: review any items due for review, if you’d like to add more words, use the Learn mode on a deck, and then use the Test mode as many times as you’d like until you’re comfortable. (When this cycle repeats the next day, those items will be due for review for the first time).

The test mode doesn’t have any impact on the SRS, only the Learn mode and Review mode have an impact. (The learn mode is how to add new items to the SRS).

Ok, yet so far I cannot find a button to restrict my review to a certain deck. In the landing page I just have the review button, a message “1776 items to review” and two of my decks shown. I assume the review count is across all my decks, while I can only choose the ones which are unfinished.
In the old Skritter I could choose which deck to “study”, and then consequently have fewer cards to review.

Is there anyway to just review one specific deck?
For instance I am currently learning HSK3, and want to review only that one. Yet sometimes I want to review it together with say HSK2, but this one is neither displayed nor can I choose two decks together. Thanks.

OK, I think I understand. However, it is still not clear to me how items are selected for test mode. Is it going to choose items I am learning at random from the deck? In that case, how can I make an effective use of my time if I am continually being retested on things I already know, simply to learn the 5% of the deck I have trouble with?

This hasn’t been implemented yet, however I agree it’s important to be able to review words from one list at a time (or as a step further a subset of lists of your selection).

This is an extremely good point. Right now it does grab the words at random, however we’re discussing ways of selecting a single list section or a selection of certain words you choose to test on.

Thats good to hear. It would also be very helpful if the app could highlight the words that I’m still learning / having trouble getting right so I could focus on them.

1 Like

Yes that would be great. Have you thought about also integrating the review button/controls into the deck sections? So that you could select specific sections of a deck (or the whole deck) either for testing, or also for review?

What is the rationale for keeping review and testing in separate parts of the app? I understand that testing is for checking your own knowledge for just learnt sections, but the deck interface could be actually a very nice location for intuitively handle multiple deck sections for review selection (especially if your retention score / test score would be somehow indicated for each of the sections)

1 Like

I get completely different reviews on the website vs on the app, in general I get way less reviews than I should given I have 95% retention set. I need to manually use the “test” feature a lot in order to get reviews. If I had to guess the cause it might be because the “learn mode” sets the SRS to “known” for new characters so they get buried. In my reviews I am regularly at 50% retention day after day so it’s pretty clear the SRS isn’t reviewing for me often enough. With the old old non-beta app I would get a lot more reviews and regularly be at 90%+ in my reviews, and I haven’t got that much older/stupid over the last year.

1 Like

I’ve also noticed a lot of reviews appear to come in at midnight…perhaps the SRS daily clock limit. I can finish my reviews at 3pm and keep checking through the day and only have a couple more appear. At 12:05am, there will suddenly be 300 reviews.

And yes, there are words I learned three weeks ago that I’ve now forgotten, that I haven’t seen in weeks. The SRS currently seems to schedule things too far apart for my memory…and I’m using the 97% retention rate.

Hi everyone,

Thanks for posting and keeping a discussion going about the SRS algorithm in the beta app. Yesterday we uncovered a bug in the system which has been causing scheduling issues with multi-character words while studying on the beta. Essentially, multi-character words were getting pushed out way too far into the future when graded incorrectly, which is not expected behavior at all!

Sorry that it took so long to uncover and debug, but we’ve included a patch in the latest update (3.2.2. 300249) and we’ll be rolling that update out today on both Android and iOS. This update will not alter any of the previous scheduling that was handled on the client, but moving forward the queue and retention goals should be aligned with expected grading behavior.

Thanks again for keeping the discussion going, and for helping us uncover this issue!


1 Like

Hi, I feel like my retention using Skritter Beta is a little low/slow, even with the highest retention rate setting. I rarely hit over 90%, even 80%. I have to be honest, I don’t use test mode because it doesn’t make sense to me in a daily study flow. It’s basically review, but you spend time on things you already know, which is exactly what SRS is there to prevent. Any plans to make the SRS intervals smaller than a day?