The above graph is from SuperMemo.com under the section “Expected increase in memory stability.”
I get that higher retrievabilities mean higher chance the card won’t be sent back to learning state. That’s why we choose 80-90% (it’s not near 100% because we need manageable reviews as well). But my thinking is that we are forgetting about the fact that lower retrievabilities yield higher increases in stability. When you find the balance between chance of lapsing and increase in stability, that’s what gives you the maximum E (S inc) value. This graph shows that lower retrievabilities give the fastest path to attain the goal of near 100% stability (lower retrievabilities give you the maximum E (S inc) value regardless of stability level).
Edit: I think I am close to understanding why I’m wrong. Search the page I linked for the phrase: " (if we were only ready to sacrifice highretentionlevels)." I don’t quite understand why Wozniak put this sentence in so if someone could help me explain this phrase I think it would lead to me understanding why I’m wrong.
It’s all about mature card(interval > 21 days). The ratio of mature cards / total cards must be high enough to attain your “memory stability” goal.
High retention rate keep your mature card at mature state (not send them back to learning state), and promote your learning card to mature state. That is. high retention rate create virtuous circle. While low retention rate demote your mature card to learning card, and lead to “ease hell” as Anki user like to say.
I get that higher retrievabilities means higher chance the card won’t be sent back to learning state but my thinking is that we are forgetting about the fact that lower retrievabilities yield higher increases in stability. When you find the balance between these two factors, that’s what gives you the maximum E (S inc) value. This graph shows that lower retrievabilities give the fastest path to attain my memory stability goal (lower retrievabilities give you the maximum E (S inc) value regardless of stability level).
search the page for this exact phrase:(if we were only ready to sacrifice high retention levels). I don’t understand exactly why he added that phrase. If you could figure out what exactly he means, I think this will show why I’m wrong.
Retention only matters at test time. That’s why I bump up to 90+%. But when its far from the test, why not build the strength of the memory faster with less work than if you used 80-90% retention. And then, when you bump up, because the memory is more stable/stronger, you will be doing less reviews at 90+% retention. What is wrong with this thinking?
First of all, only a part of learners study solely for a test. But even those who do will benefit from the knowledge outisde the test because it makes it easier to learn and understand related things.
Also, you don’t take into account the extra study time before a test. Your source only says that studying with a lower retention rate is more efficient, but not that it’s still more efficient if you have to bump up the retention later by additional studying.
