Spaced repetition: how it works and how Łatweprawko uses it

Spaced repetition works because memory strengthens when you retrieve something after a break, not when you read it a fifth time in a row. The condition: you review by testing yourself and the intervals grow after every successful review. In Łatweprawko the FSRS-6 algorithm schedules them for you.

Why one night of cramming does not work?

The test question catalogue is public. The Ministry of Infrastructure publishes it as an Excel file; the July 2026 version has several thousand questions, and that is the version we use in Łatweprawko. In the test at the WORD (Wojewódzki Ośrodek Ruchu Drogowego, the regional driving test centre) you get 32 of them, drawn in real time, and you have 25 minutes for everything.

Memorising several thousand questions in one evening is unrealistic, but the catch is not only in the number. Memory grows stronger when you have to pull something out of it. When you read the same question five times in a row, after the second time your brain gets the signal "I know this" and stops working. When you come back to it after a few days and have to reproduce the answer from your head, the memory trace is consolidated.

Two effects described in the psychology of learning rest on this: the spacing effect and the testing effect. The learning mode in Łatweprawko uses both. First, though, a few words on how we know this.

The spacing effect: what Cepeda's meta-analysis showed?

Cepeda and co-authors in 2006 gathered 839 measurements from 317 experiments described in 184 articles. They compared massed practice, that is reviews one after another, with distributed practice, that is the same reviews separated by breaks. Spaced reviews won.

The second conclusion matters more to us. The gap between reviews that gives the best result depends on how long you have to remember the material: the further away the final test, the longer the breaks that pay off. So there is no single good interval for everyone. A review schedule should fit the breaks to the question and to you, rather than apply a rigid calendar.

The testing effect: why a test teaches better than reading?

Roediger and Karpicke in 2006 checked what fixes material for longer: reading it again or trying to reproduce it from memory. Testing yourself won. The very act of searching for the answer in your head is learning.

Dunlosky and co-authors in 2013 rated ten popular study techniques for usefulness. The summary is below.

Highlighting and rereading got a low rating, even though these are what students reach for most readily. Studying with catalogue questions is a practice test, and the review algorithm adds spacing over time. So you get both techniques from the top row of the table at once.

How FSRS-6 works in Łatweprawko's learning mode?

The learning mode in the online theory course schedules reviews with the FSRS-6 algorithm. It is an open spaced repetition algorithm, the successor to SM-2 known to Anki users. We compute everything on the server, so your review plan does not depend on the device you study on.

After every answer the algorithm receives a rating. You do not give it yourself, as with classic flashcards where you click "easy" or "hard". The rating comes from three things:

1. Correctness: whether the answer you marked matches the catalogue.
2. Hesitation: whether you changed your answer before confirming.
3. Reaction time: how long it took you to answer.

A correct answer after a long think and a change of mind is a different rating from a correct answer straight away. On this basis FSRS-6 estimates how stably you remember a given question and sets the date of the next review. After a successful review the interval gets longer, after a mistake it gets shorter. We will not give specific numbers of days, because they depend on your ratings and differ between questions.

The target is a retention of 0.9. The algorithm schedules a review for the day on which the chance that you still remember the question drops to about 90%. An earlier review would be a waste of time, because you would remember anyway. A later one would mean you had already forgotten some questions and were learning them from scratch.

What a study day looks like?

You enter the learning mode and see today's queue. It contains the questions whose review date has just come up, plus new questions; you can narrow your studying with a topic filter. There are at most 20 new questions per session. The limit protects you from taking on material you then cannot review on time: every new question means future reviews in the queue.

You answer. On a mistake you get an explanation referring to the rule and similar questions from the same topic. The point is to fix the rule, not the wording of one question. The catalogue contains many versions of the same rule in different phrasing, and someone who only remembers "here I click YES" loses confidence with a twin version.

You finish when the queue is empty. The next day the algorithm sets a new one. A short session every day gives a better result than one several-hour session at the weekend, because that is exactly what the spacing effect looks like in practice.

The readiness indicator: how to read it?

In the learning mode you see a test readiness indicator. It is not the percentage of questions covered. The algorithm knows from your history the chance of remembering each question, and on that basis the server simulates 2000 times what you would remember today and what you would not (the Monte Carlo method). One number emerges from these simulations.

The indicator rises when you work through new topics and when reviews succeed. It falls when you skip a few days, because the chance of recalling questions decreases with time, even if you have done nothing new. It is not a guarantee of your result at the WORD, only an estimate based on what you have shown while studying. Treat it like a weather forecast: useful, but not certain.

What the algorithm will not do for you?

Two gaps matter most. First: the algorithm decides when you come back to a question, but why the answer is this and not that you learn from the explanation with its legal basis. Without it you learn questions, not rules, and you fall into the typical traps we described in the summary of the most common mistakes in the theory test.

Second: in the learning mode you do not train the test's time limit. In the test you have 20 seconds to read a basic question and 15 seconds to answer, and 50 seconds in total for a specialist question. You train the pace in the timed category B mock test, with the same rules as at the WORD: 32 questions, 25 minutes, a pass mark of 68 out of 74 points.

How to start?

1. Do one mock test without preparation. The result will show how much you already know from the course and from watching the road.
2. Switch on the learning mode. Every day the review queue plus up to 20 new questions.
3. On every mistake read the explanation and go through the similar questions straight away.
4. Every few days do a timed mock test to get used to the clock.
5. Watch the readiness indicator, not the number of questions covered.

If your test is in two weeks, take the 14-day study plan for the theory test and lay the daily queue from the learning mode over it.

One piece of advice to finish. On a day when you only have ten minutes, do just the review queue and skip the new questions. A review done on time is worth more than new material that, without reviews, will evaporate soon anyway.