[Question] Distribution of Difficulty is skewed towards 10 if only using Again and Good buttons?

[kuroahna]

Research

Enter an [x] character to confirm the points below:

• I have checked the FAQ and could not find an answer to my question
• I have read the wiki and still felt confused
• I have searched for similar existing questions here

Question

I've trained my data using FSRS4Anki v3.14.5 Optimizer, and got the difficulty distribution

difficulty
1     0.000725
2     0.006948
3     0.004429
4     0.005472
5     0.108589
6     0.017503
7     0.027641
8     0.231351
9     0.019130
10    0.578213

Notice that most of my cards have difficulty of 5, 8, or 10. I believe this has to do with the fact that I only press Again (1) and Good (3) buttons. I do not use the Hard (2) or Easy (4) buttons.

I remember in our previous discussion about ease hell, you mentioned

I'm working on the updating function of difficulty. It is still heuristic and magic. But I think I will figure it out.
And for the ease hell problem. It is difficult for the algorithm to determine if the interval is too long or too less if you only use again and good. Maybe I need to add mean reversion to the difficulty updating function for this case.

The mean reversion works fine for the scheduler, but for the optimizer, I feel like if we're only using Again (1) and Good (3), the difficulty distribution will naturally towards the right (difficulty of 10).

Perhaps this is impossible to fix and that it is expected for users to use the Easy (4) button when grading, but I'd like to open a discussion here and get confirmation

[L-M-Sherlock]

The optimizer also has mean reversion:

[kuroahna]

Oh oops, I missed that, thanks!

So with the mean reversion, it'll tend towards back to the average difficulty of 5. But if you only use the good and again buttons, won't it be impossible for it discover "easy" cards?

For example, let's say I learn a new card and I press good on it 10 times in a row without pressing again. Wouldn't the difficulty be stuck at 5? Is it possible for the algorithm to consider that the card is easy for me and hence assign it a lower difficulty?

Or is this a case where the user truly should just use the easy button?

[L-M-Sherlock]

The difficulty is a relative indicator in FSRS. It isn't like the ease in Anki's built-in algorithm, which is the ratio of two successive intervals. The difficulty is adaptive, so just let FSRS control the difficulty. If you only use again and good, the optimizer has considered it. I recommend keeping your habit because the law of your review behaviors has been recorded in the review logs. Keeping self-consistency will make FSRS more adaptive to you.

[kuroahna]

Thanks for answering my question! I think I'll close the issue