Is there a particular number of cards that a deck should have for it to be deserving of its own preset?
I have decks ranging from only <100-1000+ cards and it can vary widely, and ever since I heard having separate presets is better than having the same preset for all I started allocating a separate preset for each of my decks, big and small.
This has left me with copious amounts of presets that are hard to manage. Even Optimising All Presets takes a long time (I think I have about 233 different presets at this moment).
That’s something me and LMSherlock would like to investigate, but since our dataset doesn’t have deck/preset names, we can’t find out which cards belonged to which decks, and can’t measure how much splitting things into a million presets helps.
Btw, 233 presets seems like too much. FSRS needs training data, the more the better. If you split things into too many presets, it will have too little training data in each preset.
Besides, I HIGHLY doubt that you have 233 subjects that are very different from one another, unless you are a hyperpolyglot who is learning 100 languages at once and who is also trying to get a PhD in every single field in existence and also memorize the entire Wikipedia.
Yes, not subjects but rather chapters . I know 233 is too much, question is how should I be able to judge if a deck is deserving of its own separate preset, hence why I made this topic…
Well, you could combine some of them into one preset and check if RMSE decreases. If yes, then they should be combined. But that would be too much work in your case.
. I know 233 is too much, question is how should I be able to judge if a deck is deserving of its own separate preset, hence why I made this topic…