Back to blog

The Leech Protocol: When to delete a flashcard you keep failing

How to identify leeches in your flashcard deck and a specific Rule of 3 for deleting them to save your study schedule.

Hello there! If you are reading this, you probably have a handful of flashcards that just will not stay in your head. You see the card, you feel that brief flash of recognition, and then you blank. You hit "Again" for the tenth time this week.

My name is Daniel. I'm a third year engineering science student at UofT, specializing in machine intelligence. I also built Nebulearn. I work on the app every day, which is a clear conflict of interest, but I still use Anki for some of my older, 2,000 card decks from freshman year.

The most painful part of spaced repetition isn't the volume of cards. It is the "leeches." These are the cards that suck the life out of your study session because they refuse to be learned. Most students think the answer is to "try harder" or spend more time staring at that specific card.

The actual answer is usually to delete it.

What exactly is a leech

In the world of spaced repetition, a leech is a card that you have failed so many times that it is statistically unlikely you will ever remember it in its current form.

In Anki, there is a literal "leech threshold" (usually set to 8 lapses by default). Once a card hits that number of failures, the app tags it or suspends it. It stops showing up in your queue because the algorithm has realized that showing it to you again is a waste of your time.

The problem with a leech isn't just that you don't know the answer. The problem is the cost of re-learning. Every time you fail a card, you aren't just "not earning" progress. You are adding a heavy tax to your future self. You have to review that card tomorrow, and the day after, and the day after that. If you have ten leeches in a deck, they can account for 30% of your daily review time while providing 0% of the actual knowledge.

Encoding failure vs memory failure

Before you go on a deleting spree, you have to understand why the card is failing. There is a massive difference between forgetting something and never having learned it in the first place.

Memory failure is when a card was atomic and well-written, you knew it cold for three weeks, and then you just happened to forget it. This is normal. The algorithm will reset the interval, and you will likely get it right next time.

Encoding failure is when the card was never "sticky" to begin with. This happens when:

  1. The card is too complex (a "wall of text").
  2. You don't actually understand the underlying concept.
  3. The card relies on "pattern matching" (you recognize the shape of the text, not the meaning).

If you are failing a card repeatedly, it is almost always an encoding failure. You are trying to memorize a string of characters without a mental hook. No amount of "Again" presses will fix a card that was fundamentally written poorly.

The Rule of 3 for deleting cards

While Anki waits for 8 lapses, I prefer a much more aggressive protocol when I am in the middle of a heavy semester at UofT. I call it the Rule of 3. It helps me decide whether to fix a card or kill it before it ruins my morning.

Failure CountAction
1st LapseStandard failure. Hit "Again" and move on.
2nd LapseWarning. Look at the card. Is it too long? If yes, split it into two.
3rd LapseThe Leech. Delete the card immediately or "Ignore" it.

If I fail a card three times in a short window, I accept that the current card is a failure. I have two choices: I can go back to the original textbook or PDF to actually learn the concept, or I can decide that this specific fact isn't worth the "tax" it is putting on my schedule.

Most of the time, the fact isn't that important. We often make cards for tiny details out of a sense of completionism. The Rule of 3 gives you permission to let go of the low-yield details that are clogging your brain.

How your app handles (or hides) leeches

The Rule of 3 works best when your software actually tracks your failures. If your app doesn't tell you a card is failing, you'll just keep hitting the wall without realizing why your sessions are getting longer.

If you are using Quizlet, you might not even realize you have leeches. Since the free "Learn" mode is session-scoped (it focuses on mastering a set in one sitting), you might "learn" a card today just by seeing it five times in ten minutes. But because Quizlet retired its Long-Term Learning feature, there is no system to catch that card when it inevitably falls out of your head next week. You end up in a cycle of "re-learning" the same set every time you have an exam.

Anki is better at this because it tracks the "lapses" count. However, Anki is a power tool. If you are on an iPhone, AnkiMobile costs about $25. While it is worth it for med students with 10-year decks, it can feel like overkill if you just want to get through a midterm without your queue exploding.

I work on Nebulearn, and we built the "Ignore" feature specifically for this Rule of 3 workflow. When I'm importing a 50-slide PDF for a midterm, I don't have time to fix every bad card the AI or my own tired brain generated. I need a way to skip the junk without deleting the whole set. Our study-apps-that-save-the-most-time breakdown shows how this 'skip' logic is the main difference between a 20-minute session and a 2-hour slog.

When you hit a card that is clearly a leech, you don't have to delete it and lose the data forever. You can just hit "Ignore." This removes it from your active study queue so it stops wasting your time, but it stays in your library. If you realize later that you actually need that fact, you can go back, fix the wording, and re-enable it.

Re-encoding the cards you can't afford to lose

Sometimes a leech is a core formula for a final that you simply cannot delete. In those cases, the card needs a total rebuild because the original 'hook' didn't work. (for example, a core formula for a final), you cannot just put it back in the queue. You have to change it.

The best way to fix a leech is to add context or imagery.

If the card is "What is the capital of Kazakhstan?" and you keep failing it, stop trying to memorize the word "Astana." Instead, change the card to include a personal hook or a mnemonic. Or, better yet, break the card down. If you are failing a card because it has four different list items on the back, that is a "list leech." Turn that one card into four separate cards.

You can also use the "Extra" section in your flashcard app to paste a screenshot from the lecture slide where the concept was first explained. Sometimes just seeing the original context is enough to "re-encode" the memory correctly.

The discipline of a clean queue

The goal of spaced repetition isn't to have the biggest deck. The goal is to have the most efficient deck.

Every time you delete a leech, you are making your future study sessions faster and less frustrating. When your queue is filled with high-quality, atomic cards that you actually understand, studying feels like a breeze. When it is filled with leeches, studying feels like a chore that you want to procrastinate on.

Be ruthless. If a card hasn't stuck after three tries, it is probably the card's fault, not yours. Delete it, re-write it, or ignore it. Your schedule will thank you.

study-mechanicsankimemory