…Anki originally fully included any extra delay in answering in the next interval calculation. I think in 2.0 it was reduced to full/half/quarter, and more recently hard has stopped adding any delay at all. It’s mainly trying to prevent people from just completely resetting their cards, as seems to be a common desire when returning from a break: Due times after a break - Anki FAQs
In my point of view, it is really sad that Anki stopped considering a quarter of the delay to the calculation of the Next Interval for cards rated Hard. Despite understanding the preoccupation with people resetting cards completely and the aim to offer them an alternative, this new behavior has some downsides.
If we consider very overdue cards (this is the matter in this discussion), there are few (if any) theoretical justifications for completely ignoring the really used interval in the calculation of the new interval of a successful review. Users (at least me), when afraid of too aggressive intervals growth, usually used Hard when the content of very late reviews was remembered. But now pressing Hard appears to bring an unreasonable penalty, and the gap between Hard and Good became enormous.
One of the greatest improvements of Anki scheduler in comparison with the original SM-2 is exactly the computing of delays in next interval calculations. I feel like Anki unwittingly made a step back in this design choice!
It’s unlikely that you’ll find much support for reconsidering a decision that was already a fading memory when Damien wrote that 4 years ago. And especially now that the SM-2 algorithm has been surpassed in most ways by FSRS, which uses a completely different system for determining new intervals after delayed reviews.
Isn’t the real issue that it was pretty much guesswork? Whether to add the full, half, quarter, or no delay sounds like it was a rough approximation along the lines of “that seems fair.” If it was hard to remember the answer after that period of time, and we know the next interval is already going to grow based on the card Ease – not having it grow even more than that doesn’t seem like a penalty, or all that unreasonable.
When pressing hard, the interval grows not according to the card Ease, but according to the “Hard Interval” (1.2x by default, 1.3x maximum). This is a very small growth when you do not consider the overdue period. For Instance, a card with an interval of 7 days, rated a year late, will receive a new interval of 9 days. I know it is an edge case (but it happens a lot for forgotten collections), and, for me, this is kind of unreasonable.
I understand this is a 4-years old change, but it changed a 15 years standard behavior. All memory models and schedulers “guess” many things; how reasonable are these guesses makes all the difference! Anki SM-2 is yet a very nice (and configurable!) algorithm, that I still use for some specific purposes (e.g. For multiple-choice questions, that I want to have an initial interval of 50 days, and that I do not care about the memory stabilization decay for very mature cards). I am sure it is still going to be used in the years to come.