You are using learning steps longer than one day. That would cause odd problems in FSRS. Please remove them.
Sorry for misunderstanding your data. I figure it out. There are now a direct method to force new cards to be reviewed the next day/2 days. Because FSRS schedules every card based on its stability. The stability relies on your parameters. The parameters are generated from your review logs if you used the optimizer.
I did run the optimizer, but I switched back to the default parameters after a while, because some of the decks I had included in the optimizer were old and near 100% retention at the point that I ran it, and it was giving me some crazy intervals for some cards. The cards info I posted before were using default parameters unless the earlier pasted parameters can cause something else to change.
OK. If you switch to the default parameters, please reschedule all cards with the helper add-on.
Oh I meant that I already was using the default parameters while learning the cards I had posted I even rescheduled beforehand and everything.
Actually, I forgot that I did not change requested retention back as I currently have it on .85 if that would cause this to happen.
The interval relies on stability and requested retention. So you should increase the requested retention to 90%.
So is it not possible to request a lower retention and still have it schedule newly graduated cards the next day? I also should note that all of the new cards that I learned did not always receive the same graduating intervals. It was a spread between some cards being the next day and some cards being anywhere up to 5 days into the future.
I was able to find out a bit more after searching around and feel like it makes more sense now. I’ll just have to try some stuff out and see how it works for me.
Also when I run the optimizer should I change requested retention in the optimizer itself if I want it lower or higher? Or should I just run the optimizer on it’s default .90 retention rate and lower my retention rate within the deck options? Or maybe even both? Thanks for all your help by the way. I hopefully won’t have many more questions.
The requested retention doesn’t affect the optimizer.
Sorry to bother you again, but I wanted to check something just to make sure I am not messing things up. Does fsrs look at time spent per card? I would assume it wouldn’t, but read from someone else that it does (although the source for this is months old by now) Wanted to check because I have a bad habit of staring at cards longer than necessary.
1 Like
The time spent per cards are considered in the search of the best retention. The optimization of parameters don’t use it.
1 Like
I am learning languages in anki and i create flashcards from words i see while watching media / reading books etc.
Sometimes when reading i encounter a word i think i don’t know only to realize that i already have a flashcard with it, so i set the due date to 0 and fail it the next time i see it when reviewing.
Would fsrs take that into account as a flashcard i forgot when optimizing / scheduling? Setting the date manually adds a 0 rating to the history of the flashcard, does it impact anything? The next rating after that is a 1/fail, so i think it should take it as if i was show the flashcard sooner and failed it.
Or am i doing it the wrong way and i should instead use the Forget function?
1 Like
I just started to use fsrs4anki today and I just finished getting everything set up. When I plugged in my parameters and used the helper addon to reschedule all cards, it looks like it gave a very weird output. I just wanted to know if something is wrong here, or if a forecast with a huge backlog like this is expected behavior.
This is before I rescheduled (note that I have no backlog at all yet):
This is after I rescheduled (the backlog is huge):
If I just ignore the backlog though, things look a lot more normal:
Is having a backlog like this after the rescheduling okay?
I searched this thread but it looks like I’m the first person here to have this issue unless there’s somewhere else I should look.
In optimizer, these revlogs are considered as outliers and removed.
1 Like
It is normal if your request retention is higher than your past true retention.
1 Like
Ah, makes sense. Do you know if manually removing Set Due Date / 0 from the revlog database would make it work like i described (algorithm thinking as if i was show the flashcard sooner and failed it)?
你好，老哥
我想问你一个问题。
ankimobile的图片遮挡模版，如下图
一张小的遮挡板，由一行{{…}}描述。其后则是图片。另外2个字短通常是空的。
（相比于 字段有一大段文字的其它模版），请问fsrs对这个官方模版有优化吗
感谢老哥对anki的贡献
算法不关心卡片内容，只关心历史复习和记忆规律。
1 Like