I have 980 review with fsrs the problem is one card raise rmse 1% with the card
0.4892, 0.7611, 1.4768, 2.2975, 7.2405, 0.4793, 1.4525, 0.0622, 1.5263, 0.1507, 1.0054, 2.0977, 0.0010, 0.4572, 2.4189, 0.3273, 2.8932, 0.4643, 0.7995 with 3.93 rmse
without the card
0.4055, 0.6202, 1.1733, 1.7946, 7.1730, 0.5373, 1.4308, 0.0725, 1.5727, 0.2741, 1.0573, 2.0938, 0.0010, 0.4552, 2.4162, 0.3964, 2.9452, 0.4761, 0.8043
rmse 2.8% so should I delete that card
1 Like
I cannot answer your question (someone else might) but I’d like to know how you somehow found a card that had this kind of effect on your RMSE.
Also, are you talking about a new card you just added or a card that was already part of your collection?
was already one of my cards (its on young phase) its hard card very hard the new parameters seems that it doesn’t changed that much but the rmse raised significantly
I’m sure we’ll need to know more about your collection to answer your question.
- How are you adding and removing that one card from this optimization? Moving the card to a different deck in another options group, filtering that card out in the Deck Options window, suspending the card, or something else?
- Are you determining that you have 980 reviews from looking at your Stats? Or is that how many reviews FSRS says that it is optimizing/evaluating?
- How many cards are included in that 980 reviews? [or asked another way: How many active Review cards do you have in that preset?]
- How many of those 980 reviews are from this one card? [or asked another way: How much does the FSRS reviews count change with/without it?]
- Are the parameters and RMSE that you’re getting with the card from a fresh optimization you’re doing just now, or an optimization from before – a week ago, a month ago, etc.?
1 Like
It’s clearly possible there’s one outlier card that’s FSRS isn’t quite good with but the question is: did you try to sync just before removing the card from that preset?
will I decided not to delete the card since it affects the no. and the new parameters after deleting the card seems very near and yes I tried to sync
You didn’t answer these questions. I can assume you’re removing it somehow, because I can see the change in the review count. But it’s not clear from your screenshots whether this is a fresh optimization or whether you’re even re-optimizing when you add/remove this one card.
You didn’t answer #3 either, but the screenshots show that whatever you’re removing contributes 37 of the 984 countable reviews – that’s almost 4% of the data FSRS has. I think it’s reasonable to think that removing that card could have an out-sized effect on your parameters – especially if the card has a higher number of reviews than any other card – and especially if you have a small collection. (Is this under 100 cards?)
Figuring out whether a card is a leech and how to fix it is more significant than just how it affects FSRS. If you want to learn this card, you have more work to do, so you don’t need to waste time trying to game FSRS. As you accumulate more data, each single card will matter less, and FSRS will even out.
1 Like
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.