Simple issue - when I add “Retrievability” field to the card browser, I’ll get one value, but very often, if I actually inspect the cards and go to the Card Details (with date added, first review, etc.), I’ll get a completely different value.
The one on the card seems correct intuitively (on my examples), but how do I get THAT one in my Browser?
Running Anki 25.02.1 on macOS, tried in Safe Mode. It happens only with cards that were shifted around decks, and which were subject to using of the “Postpone Cards” Add-on (not the FSRS one). That Add-on is something I definitely still intend to use now and then (I know about the drawbacks, but it makes sense for me in some particular cases).
tl;dr I can check a card in the card browser, it says (wrongly) that the retrievability is e.g. 80%, but when I open card info, it will be like 60%.
Might be the same, as I’ve heavily used “Set Due Date” when going through Backlog. What I noticed was that intervals were very broken - e.g. I’d see a card with the interval 24 days, but 3 months after having actually seen it, and when I know it, the new interval is 1.5 months, which makes absolutely no sense (as it should be like 4.5-6 months).
If you’re talking about an add-on that pushes all of your due dates by a certain number of days – aside from any of the general drawbacks, some of those are known to not work very well. (I suppose it’s not too surprising that a short-sighted, ill-considered, and outdated add-on would also have design issues.) Since you’re now seeing this inconsistency on cards it has affected, that’s all the more reason to avoid using it.
Perhaps it’s a failure of my imagination, but I can’t think of any “particular cases” that wouldn’t be better handled in some other way – like a catch-up deck, FSRS Postpone, or even Set Due Date. And I don’t think any of those would create this R inconsistency you’re seeing.
Yeah, that’s the one. I mean, I might be missing something, but my situation is as follows: I had a fairly long hiatus with an obscene backlog, and after coming back and going through it, my daily workload is… obscene as well. On 90+% of the days I manage it, but some days life just gets in the way and I want to skip the day.
Yes, I know all about the Anki logic and how my recall will suffer, but in practice this approach has worked perfectly for me over many years and I’m ready to pay for that tradeoff.
Using FSRS Postpone, from what I gathered, will just increase my workload in the following days. What I usually do is postpone all cards due in the next 30 days by one day. That way, all cards that are 31+ days due will still be due on the same day, and the ones in the 30 days after the day I postpone them will be due a day later.
Is there a more reasonable way to handle that? And no, I am not adding any new material right now, but it will take a month or two more for the daily workload to be a bit more normal, and even then I won’t be able to make all my reviews a 100% of the time.
And, either way, my problem still persists. Also, most of the time the intervals make sense, like, the one on my Card Info actually seems realistic, but I’d really like for a way to take the current interval into account when evaluating card’s memory state because this discrepancy is killing me.
If you’re not willing to use a catch-up deck, and you want to have more control about when those cards are handled than you would be able to get with Postpone – How about: take that one days’ cards and use Set Due Date to scatter them among the next however-many days? It will create an increased workload on those following days, but it will be smaller, and it harms fewer cards. Your method also creates an increased workload day – it just kicks it down the road for a month.
If you’re struggling with your workload overall – have you considered lower your Desired Retention to get some actual relief?
But FSRS doesn’t need to consider the current interval when figuring the card’s memory state. I know it’s a bother to see different R numbers in different places, but do you have any reason to think that the wrong R is being used in scheduling your next due date?
Okay, but how does it make sense that when I open a card that was last reviewed 6 months ago, and I know it, the new proposed interval is e.g. 2 months? I can’t wrap my head around the math in that.
Like, I can live with the discrepancies between browser and card info if it makes no practical difference, but I’d like to understand it.
And also, no, I’m not kicking the problem down the road for a month, as the number of cards due in 30 days is 10x smaller than in 1 day, so the “extra” day is negligible.
Okay, but how does it make sense that when I open a card that was last reviewed 6 months ago, and I know it, the new proposed interval is e.g. 2 months? I can’t wrap my head around the math in that.
That’s actually entirely possible. In FSRS, the “overdue bonus” or whatever you want to call it, isn’t linearly proportional to interval length. The “bonus” saturates.
What this means in practice is that if your interval is 2 times longer than it was supposed to be, your “bonus” (how much the next interval increases) will be, say, x1.5. If your interval is 100 times longer than it was supposed to be (idk, you went on vacation or something), your bonus will be, say, x2.5. The numbers are arbitrary, just to illustrate the point.
If you recalled a card after 50 years, it doesn’t necessarily mean that the next interval will be >50 years. It can, of course, but it’s also possible that you will have a shorter interval.
The best we can tell you is – that’s how your collection looks. Part of the magic of personalized scheduling is that I can’t explain to you how FSRS gets to that number. You can look for yourself though! Try the FSRS Visualizer to see what a set of parameters would do with a set of grades. You can also look at that card’s path on your memory curve in the Card Info.
If FSRS sees that you’re consistently correct with the next review after a 6mo gap like that, your optimized parameters will start to skew towards allowing that gap. If FSRS sees that you’re wildly inconsistent with that next review, your parameters will make that gap less likely. There are flukes and outliers, and FSRS is designed to quiet some of that noise.
Also, most of the time the intervals make sense, like, the one on my Card Info actually seems realistic, but I’d really like for a way to take the current interval into account when evaluating card’s memory state because this discrepancy is killing me.