Anki Forums

Alternative leech threshold: ease instead of lapses

I had been using a leech threshold of 4, because I frequently lapsed on badly formulated cards. I figured that it took too much time if I wait until they have accumulated 8 lapses. So I set the threshold to 6, later to 5, and finally to 4. This is, until I started to use https://ankiweb.net/shared/info/1672712021. With this add-on, I expect lapses to occur more frequently.

In general, I expect that setting the leech threshold according to ease instead of lapses will benefit many users, and long-term users in particular – a card might have an ease of 200%, be considered viable, but due to its year-long history have too many lapses.

Hi,

I am a long-term user with multiyear interval cards in my collection. I have been using a leech treshold of 4 as well. Usually I only leech very few cards. I think you should update your settings if you are leeching that many cards.

I disagree, that would cause the opposite effect. That approach would generate even more leeches long-term. Lapse count con only be increased, and the ease can either increased or decreased depending on the user rating being hard or easy respectively; or with add-ons. Another scenario is that I can set any starting ease in different settings group, so the current ease is not useful for card difficulty discrimination.

I leeched so many cards (relative to the total card count in my collection) at the beggining. Note that formulation of knowledge is a skill, you need time to built the skill set and accumulate the expirience to eventually get proficient with formulation of knowledge. I suggest you to spent more time when creating cards following the basic 20 rules of knowledge formation. You will spend more time initially when creating, but less time when reviewing, and reviewing again when lapsing cards. So, all the extra reviews will outweight the intial time saved.

2 Likes

Thanks for your reply! It’s freshening to hear such constructive opposition.

What kind of update do you suggest?

This is in fact what I called a benefit: Leeching more cards means less time spent on difficult cards, or detecting difficult cards earlier. It’s always possible to simplify them once they have been detected. The “leech” tag automizes the detection process.

That’s true. Still, cards that have an ease under, e.g. 170%, usually become leeches eventually. It will still take some time until leeches are accumulated – longer if learning steps are longer. The goal of my suggestion is to shorten the time until this card is detected as difficult (signified by “leech”).

I imagined that the leech ease threshold could be set separately in each deck option group, so that it would be possible to assign a lower threshold (I mean a lower ease value, i.e. a less aggressive threshold) to groups that already have a low starting ease. This way, the user can, for example, maintain the same proportionality between starting ease and leech threshold for every deck options group, even if this groups have different starting eases.

There are so many options, the first thing I’d change is starting ease and what you do with lapses (resseting or reducing interval). The big picture is to find what settings works best for you. If you have different settings groups you can compare the retention in the stats and use A/B testing. Also now you can easily filter any search in the stats. So you could filter recall cards, production cards, etc and see what setting change is getting higher retention.

Let me reword the same idea, when you lapse a card once or twice and you think that the cause is genuinely caused by the formulation of knowledge, I suggest to edit that card right away (E) and maybe split it into different cards if needed. There is no need to leech a card 4+ times to take action.

Exactly, instead of a general preference the threshold could be at a level of deck group settings. You could try to tune up the experimental ease add-on to adress this automatically.

1 Like

For starting ease, I use the add-on AvgEase. It automatically sets it to the average ease of mature cards.

For lapses, I use the add-on autoLapseNewInterval in order to target an 85% success rate.

They seem to be working well and I prefer such an automated process over testing and using the stats screen to manually calibrate.

Also now you can easily filter any search in the stats. So you could filter recall cards, production cards, etc and see what setting change is getting higher retention.

That is cool! I’m still on Anki 2.1.22 because some add-ons don’t work with higher versions.

Let me reword the same idea, when you lapse a card once or twice and you think that the cause is genuinely caused by the formulation of knowledge, I suggest to edit that card right away (E) and maybe split it into different cards if needed. There is no need to leech a card 4+ times to take action.

In my post, I had the idea in mind that I quickly and automatically tag and suspend a card that needs reworking, so I can edit it at a later time.

Perhaps it is commendable to rework cards right away. After all, the 20 Rules of Formulating Knowledge recommend detecting and eliminating interference immediately. Also, I sometimes get ideas for amelioration right after a lapse, which could fade away from working memory.

On the other hand, reworking cards later makes it easier to prioritize.

1 Like