Let’s get learning steps wrong! (it’s very easy)
Let’s just preface this by saying that 15m, 20m or 30m is ideal both with FSRS and with the old algorithm too. But who cares about what is “ideal” or “optimal”? We’re here to get things WRONG! Here are 12 ways to get the learning steps in Anki wrong:
1) 2d. This can cause the interval for Hard to be longer than the interval for Good.
2) 1d. Believe it or not, this also can cause the same problem as above, because it can somehow turn Hard into 2d, don’t ask how.
3) 12h. This can cause Hard and Good to be equal. It’s not as bad as Hard > Good, but still undesirable.
4) 18h. This can cause Again, Hard and Good to be equal to each other.
5) 10m 10m. This also can cause Again, Hard and Good to be equal to each other.
6) 1m 15m. This will make you review a new card twice per day. FSRS doesn’t take same-day reviews into account (FSRS-5 does, though), so the extra step is just a waste of time. The more short steps you have, the more time you waste, since FSRS won’t use those reviews, and a year from now on it won’t matter whether you reviewed this card 1 or 2 or 3 times on your first day of seeing it, regardless of which algorithm you use. This is the least wrong way of using learning steps out of all the wrong ways listed in this article, though it’s still suboptimal.
7) 30m 15m. This will cause Again to be longer than Hard, which in turn will be longer than Good aka Again > Hard > Good.
8) 15m 1d. This will cause your first interval after you press Good to be one day long instead of allowing FSRS to choose the best first interval for you. The same consideration applies to the old algorithm too, though it’s more important for FSRS.
9) 12h 1d. This will combine the problems of number 3 and number 8 together.
10) 18h 1d. This will combine the problems of number 4 and number 8 together.
11) 1m 15m 1d. This will combine the problems of number 6 and number 8 together.
12) A special award goes to the learning steps of a certain user: 1m 10m 1d 2d 4d 8d 16d 32d 64d 99d. At this point it doesn’t matter whether you are using FSRS or the old algorithm, your learning steps are basically your own new algorithm now, and an extremely inflexible one.