API Items.previousSuccess wrong criteria?


The documentation says:

  • previousSuccess
    boolean, whether the last review had a score greater than 1

I am seeing “False” for things I know weren’t wrong last time I studied them. I’m not 100% sure what they have in common, but I suspect from this sequence:

part last previousSuccess reviews successes
rune 1538432468 False 41 35
rune 1538432462 False 36 30
rune 1538432458 False 123 102
好朋友 rune 1538432458 False 1 0

I “just learned” 好朋友 (from a Mandarin Companion book list) and when I got the hinted presentation that did not count as “correct” and so each character in that word got dinged. Historically I also think it’s unlikely I’ve gotten 好 wrong 21 times

Another example (including 子 that I’m sure I have not “missed” 16 times):

part last previousSuccess reviews successes
rune 1538432404 False 157 141
rune 1538432400 False 56 48
rune 1538432396 False 67 59
没面子 rune 1538432396 False 1 0


Hmm, nice catch. I will look into this a bit more when I have a chance. The previousSuccess property isn’t really used for much SRS calculation wise, but is something that should be accurate (especially for people using it for 3rd party API purposes).



Advice for anyone who hits this issue: Turn off “Teaching mode” (looks like this is one of those per-browser settings). With teaching mode off you have control over the outcome the first time you see a word or character. So if it’s a word and it re-uses a character you can get it right (even if you don’t know which variant and you have to hint it out, you can override the result to neutral at least). I just verified by learning 留下 and then fetching via the API to see previousSuccess is correct for both 留 and 下. Plus 留下 itself is “1 for 1” instead of starting “0 for 1”. Maybe that will also help with scheduling efficiency when learning words containing all characters I know.


closed #4

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.


unlisted #5