After reading through most of the above, I think I’ve got it right but would like to confirm. Please correct me if anything below is incorrect:
The number 37 (normally shown in gray, but shown in black in the larger font when hovering) is the current number of leeches in the Apprentice bucket.
There are currently 3 leeches in stage 1, 14 in stage 3, and 20 in stage 4 (for a total of 37). This is shown in the tooltip when hovering.
I currently have 99 items total in the Apprentice bucket (stages 1-4). Three in stage 1, 9 in stage 2, 50 in stage 3, and 37 in stage 4. Not related to the script, but as shown on the right by WK when hovering, 3 of the 99 are radicals, 47 are kanji, and 51 are vocabulary.
So all three items in stage 1 are leeches, 14 of 50 in stage 3 are leeches, and 20 of 37 in stage 4 are leeches. In total, about a third (37/99) of the items in my Apprentice bucket are currently leeches (which is a little disheartening but sounds about right! :-).
Yep, you got everything right! It can definitely be disheartening to see that a high percentage of Apprentice items are leeches. But on the other hand, you can’t improve your situation until you know what the problem is, so I think it’s better to know. Good luck!
All that detailed breakdown actually made it a bit harder to understand. Personally, I don’t see much value in the breakdown by stage (nor the breakdown by radical/kanji/vocab, not that that has anything to do with your script).
I’m wondering if the percentage of leeches might be more (or at least as) interesting as the total number of leeches in each “bucket.” I don’t seem to remember the total number of leeches in each category, but the “about a third” realization really struck me.
Regardless, I’m hopeful that my leeches will eventually take care of themselves. Most are due to visually similar kanji (e.g.: 矢、失、夫) and homonyms (感心 vs 関心 being my current nemesis).
When I get surprised with a wrong answer or realize I keep seeing and failing the same item, I try to figure out why (the 似合い visually similar kanji script has proven invaluable). Usually, the solution is to make a new or additional mnemonic of some sort.
How does that happen with only one stage in the enlightened bucket? I’ve not unburned anything that I’m aware of. Does an item’s “leechiness” stick around for a bit if you miss an item but later successfully move it to a subsequent stage?
If I’m using the self-study script correctly, the one item seems to be the vocabulary word 考える. Here are my stats for that word:
I’m trying (but failing) to understand the getLeechScore function logic in the script. I note the current streak (4) is only slightly less than the longest streak (5) for meaning.
I’m unsure how many incorrect meaning attempts the API returns, but if it’s 21% of 43 (i.e. 100% - 79%) that is about 9. So I think the “meaning score” calculated by the script is 9 / 4**1.5 or ~9/8. Since that’s > 1 the algorithm considers it a leech. Do I have that right?
I’m just trying to get a feel for what the algorithm considers a leech. I realize it’s a heuristic, but I’m trying to get an intuitive feel for the logic.
Thanks!
[The more I think about it, by the way, the more I think I would rather see a percentage than the absolute number of leeches.]
Without digging into the exact formula, the total number of incorrect answers does matter. So if you get an answer wrong 100 times, and then get it right 7 times in a row to get it to Enlightened, it will probably still be considered a leech. I think your understanding of the formula from your example is fine. As you said, it’s just a heuristic, so it’s not going to be perfect.
In the main post of this thread there is a section that describes how the leech formula works if you are interested. This script will give you some visibility on what the leeches are. (Disclosure: I am the author of Item Inspector)
Why is that the number of leeches shown differ from what the self-study quiz gives me? The breakdown from this script shows me 408 leeches in total when added up, but the item inspector/self-study quiz gives me 417. Just curious. The filter settings are set at all items and leech threshold 1.
You can get the burned leeches out of Self-Study Quiz if you setup the SRS Level filter on top of Leech Training to exclude them. Just select all SRS Levels except burn.
This is kind of an absurd question, but how can I identify the leeches? I see the number is mapped by our function, mapItemsToSrs, but that simply returns an array with our various SRS level and the numbers of items. While that is useful, I’d like to know which specific items are leeches. To that end I’ve simply modified the script and dumped items identified in isLeech into an array and console logged them for my own info. If there’s a script that’s designed to do this, please let me know?
Otherwise, if such a script doesn’t exist, I don’t mind developing it myself. Can display the items in some kind of modal from the menu.
It was built specifically to work with the Self Study Quiz script, but in theory it should work with other scripts as well. If Self Study Quiz doesn’t give you what you want, maybe someone has a script that just lists out the items indicated by filters like the leech filter.
