Advice on arrangement of decks for FSRS to properly work

Anki FSRS
1

Hello Anki and FSRS developers

So I know that FSRS uses certain parameters specific for each deck.
However, I have a concern given my way of arranging my decks.

So I prefer to study all my cards from all decks (language, medicine, etc) randomly and together in one session. That is why I have arranged all of my decks as subdecks of a Main Deck (I have named it “King Deck”). So whenever I decide to study anki, I only have to choose the main deck and start studying all decks randomly. (I am attaching the screenshot of the arrangement.)

Also, I have always kept the same settings for all decks under my default settings in the SM2 algorithm and it has mostly worked well so far. I get 85-92% recall on mature cards regularly. I have not yet changed the settings to FSRS but I intend to change it to FSRS after getting advice on this.

So, if I keep my arrangement of decks, will FSRS work according to its true intention ie, separate parameters for each subdeck, or will it have the same parameters for all decks? Should I keep or change the arrangement of my decks? If I have to change, is there a way to randomly study all decks together without a Main Deck?

Thank you very much

Basil

image
image1152×996 120 KB

2

Your initial premise is not quite right. FSRS applies Optimization, Parameters and Desired Retention per Deck Options Preset group, not per deck. So it doesn’t matter what your arrangement of decks is, or what deck you click to study – it only matters how you have your decks assigned to Preset groups.

3

Thanks for the reply.

So does that mean, in order to get the benefit of separate parameters for each subdeck, I should create a separate preset group for each of my subdecks?

4

In order to get separate parameters for each subdeck – yes, you’d need separate presets. But I think you’re assuming that would a “benefit” – when it might not be, or it might not be significant enough to be worth the effort.

The general advice about that goes something like –

  • The more review data FSRS has, the better it can optimize for you. So dividing your collection into tons of tiny presets will likely limit FSRS’s effectiveness more than it would enhance its accuracy.
    • [Switching to multiple presets when you’re used to having just one also increases the chances you’ll mis-assign a deck to the wrong one, and adds the bother of sometimes needing to make the same change in multiple places.]
  • If you are studying completely different subjects, especially subjects that differ substantially in how difficult they are for you (subjective difficulty), you might benefit from separating those to different presets. I say “might” because it’s not a one-size-fits-all situation.
  • If you have substantially different grading habits in different decks, you might benefit from separating those to different presets. Some examples might be: grading every card Again when it is introduced, or 2-button vs. 4-button grading. But still, it’s not one-size-fits-all.

So, for instance, looking at your collection – “Step 1” and “Step 2” are both medical-education related, right? It might be useful to have those in a preset together. But it seems like “Paramedic” would probably go with those too. While “Language” (language learning?) might go in a separate preset.

If you want to experiment with adding a few more presets, you can Evaluate different presets against different sets of cards to find their RMSE (essentially, error rate), but it’s a bit fiddly. (And don’t waste your time chasing itty-bitty improvements to RMSE. You’d be looking for substantial drops.)

You’re also fine to start out with just a single preset for your collection – the same parameters and Desired Retention for everything.

5

Thank you so much for the comprehensive reply, it makes a lot of sense now. Also your examples and suggestions about grouping similar decks into a common preset, makes it much clearer.

Although, after talking about the trade offs, now I am in a limbo, and have to decide whether or not to make separate presets, haha.

But thank you for the insight.

6

Here’s how much better FSRS is than SM-2 – even if you enable it with all one preset and the default parameters, it’s probably going to be better. :sweat_smile:

Another approach is to look at what your retention is like for different subjects. You mentioned 85-92% earlier – but that’s a pretty big range. If all your medical decks are around 85%, you might want to set the Desired Retention for those closer to that. You need a separate preset for that, so boom, you’ll have separate parameters too.