High Burden FSRS w high card difficulty

Hi! I am in a situation with a high card burden where I am getting 800-1000 reviews a day when I know my peers are around 300-500 because we are doing essentially the same boards/card unsuspension. I originally had my learning steps at 15 mins but a friend suggested I go to 5m and 15min. Originally I was hitting again quite a bit when I was learning new cards which I have found has extremely increased many of my cards difficulty even if I dont think they really are. I also dont think I over use the hard button.

Here are my parameters:

0.2334, 1.5401, 16.7664, 100.0000, 6.9682, 0.5001, 2.4553, 0.0020, 1.3719, 0.1386, 0.9972, 1.7914, 0.1382, 0.2522, 2.3911, 0.1302, 2.9556, 0.5404, 1.2944



What do you need help with?

As a general thought:

  1. Your true retention is around 85–90%. If you’re okay with reducing your retention (and getting a lower workload as a result), you can decrease your desired retention in the deck options.
  2. You should use the answer buttons as described in Studying - Anki Manual. If you do that but your workload is too high, then stop doing new cards for a few days or a week.
  3. It doesn’t really matter how may cards your peers do. a) They might use anki wrong (e.g. say they knew the card even though they didn’t), b) they might just learn faster than you do, c) they might have less difficulties with the topics they are learning compared to you (i.e. it’s easy for them).
2 Likes

I think my main problem is why my card difficulty is so high. I don’t think it is representative. some cards that show up all the time and now I know and hit good, will not send the card out very far in advance. but this is the case for like thousands of cards

  • What is your Desired Retention set to?
  • Have you Evaluated your parameters? Click Evaluate and make a note of your RMSE.
  • When was the last time you Optimized?

The initial Difficulty (D) is set on a card based on the first grade you give it. Then that is adjusted up/down based on future grades you give it. From the looks of your Answer Buttons graph, you use Again quite a bit in Learn, and you don’t use Easy at all. Your Avg. D is probably high because it starts high, and you never grade your easy cards Easy.

That’s not necessarily a problem. Your D values are concentrated in a particular range, but there is some variation, and they aren’t imbalanced within that range. D is one of the values FSRS uses to schedule your cards accurately, but it’s not the only thing.

From: A technical explanation of FSRS | Expertium’s Blog

Your parameters also show that “reversion” of Difficulty happens very slowly for you: w[7] = 0.0020. It’s not the lowest I’ve seen, but it’s quite low.

Do I think you should start using Easy? It’s an option, but if you do, you’ll probably need to switch to the default parameters for a bit, because your initial S for Easy is currently 100d (the maximum, which also suggests there’s no/not much data there).

The questions I asked up top might help come up with other ideas.

Hi!

My retention is set to 91% and I would like to stay at least above 90.

Eval: Log loss: 0.3745, RMSE(bins): 2.33%.

I optimized today, and at least once a week.

Thank you for all the insight, yes I almost never hit easy on the deck I primarily use so that is a good option I did not realize hitting good was equal to nothing difficulty wise.

Your actual retention results are coming in a bit below that, so FSRS is going to keep scheduling your with more cards at shorter intervals to help you get your retention to your goal. That is a significant reason for your high workload.

Has your DR been at 91% for a while or have you recently raised it to that? I’m a little surprised that your RMSE is so good/low when your retention is ending up that far from your DR. [But wait – are those the Stats for exactly the same decks as are in that preset? If not, we might be comparing apples to oranges.]

The ordinary recommendation for lower workload is decreasing your DR. So you might have to decide which you want more – DR above 90% or a lower workload.

That’s more often than you need to optimize. Once a month is plenty. You’ve got a good amount of review history – are you getting the parameters-already-optimal response a lot?

I don’t want to overblow the text I quoted (because that’s how rumors get started). That “nothing” is only a baseline. There is still the mean reversion factor to consider – but exactly how much Good affects the next calculated D is going to vary from user to user, based on their parameters and the formulas explained in that post.

My DR has been never been below 91, so for at least 5 months. Maybe it is not right then because the RMSE was for ankihub that is used by 20 decks where as the stats are specifically for the AnKing Step Deck? But that is confusing because I only have three decks total so maybe that is the RMSE for my collection. Is there a way to get the RMSE for specifically that deck?

Yes, i get already optimal a lot

Haha okay yes that makes sense thank you.

Here is the stats from my whole collection that might make the RMSE make more sense?

There isn’t. RMSE and parameters are per preset, but you can get the Stats for that preset if you want a better comparison. Just search for it in the box at the top of the Stats window.

image