Learning Sciencememorytraining logspaced repetitiontechnique retention3 min read

How to Remember BJJ Techniques (and Why Most of Us Don't)

September 12, 2026

Listen to this post

0:00 / 0:00

You step off the mats with something new. A purple belt spent ten minutes on a back-take sequence — the grip, the hip switch, the moment the second hook goes in. You hit it twice in the last round. Driving home, you can still feel the timing of it.

Two weeks later you're in the same position. The purple belt is in front of you, back exposed, and your hands do something approximate. Not wrong, exactly. Just vague. The detail you were so sure you had — the hip switch, the exact grip — is not there to retrieve.

This is probably the most universal experience in jiu-jitsu. The usual explanation is that you didn't practise enough, or that you weren't paying attention. Neither is usually true. The technique was learned — what's missing is encoding: the part of learning that decides whether a detail is still retrievable a fortnight later. Understanding how encoding works changes what you do after class, and the benefit compounds. Every technique that actually stays becomes ground the next one can be built on, so a small change to how you review pays out across everything you learn afterwards. The research behind this is old, well-replicated and unusually practical.

There's a decent body of research on this. Almost none of it has been translated into what a grappler should actually do on a Tuesday night.

Forgetting is not uniform, and it is not a moral failing

In the 1880s, Hermann Ebbinghaus ran an experiment on himself that became one of the founding studies of experimental psychology. He memorised lists of nonsense syllables, then tested his own retention at intervals from twenty minutes out to a month, and measured how much time he needed to relearn each list — his "savings". The shape he found, plotted over time, drops steeply at first and then flattens. It has been known ever since as the forgetting curve.

You have almost certainly seen a version of this chart captioned with a claim that people forget 70% of new information within 24 hours. That particular number is a loose popularisation; Ebbinghaus's own data, and the replications since, are messier and more interesting than the caption.

In 2015, Jaap Murre and Joeri Dros published a replication of the original experiment in PLOS ONE — a century and a quarter later, with a modern subject and modern analysis. The curve held up. But their analysis found something the popular version leaves out: the curve is not smooth. There is a bump upward at the 24-hour point, as if some of the material was being reorganised overnight rather than simply decaying.

That detail matters for anyone who trains in the evening. It suggests the first day after a class is not a passive slide toward forgetting — it is part of the process, and it is happening whether or not you do anything with it.

Forgetting is inevitable, then, in the sense that the curve exists. What's optional is the rate.

Three things that change the rate

The learning-science literature contains three findings that are unusually practical, unusually well-replicated, and almost entirely absent from how grapplers are taught to drill.

One: producing information beats reading it. In 1978, Norman Slamecka and Peter Graf ran a series of experiments in which some participants were given a word to read, and others were given a cue that let them generate the word themselves. The generators recalled it better — consistently, across five experiments. The finding is known as the generation effect: you remember what you produce far better than what you consume.

Two: retrieving information beats reviewing it. Henry Roediger and Jeffrey Karpicke's 2006 "test-enhanced learning" experiments took students who had just read a passage and had one group restudy it while another group took a recall test instead. Then they waited. On an immediate test, restudying looked fine — often slightly better. When the delay was a week, the group that had retrieved the material rather than re-read it retained substantially more. The act of pulling something out of memory is itself what strengthens it. Restudying mostly creates fluency — the sense that you know something, which is not the same thing.

Three: spacing beats massing — but the gap has to be days, not minutes. This is where most practice in jiu-jitsu gets it wrong. A 2006 meta-analysis by Nicholas Cepeda and colleagues pooled 317 experiments on the "distributed practice" effect and reached a conclusion that is blunt in its practicality: separating learning episodes by at least a day, rather than concentrating all the work into one session, is extremely useful for maximising long-term retention. Drilling a technique fifty times in one evening is massed practice. It feels productive because performance within the session improves. That feeling is not the thing you want.

What this looks like on the mats

Put the three findings together and the implications for BJJ are specific.

Grips, positions and finishes need words, not a feeling. The generation effect says you remember what you produce. So the useful act after class is not re-reading the coach's demo in your head — it is writing or saying the technique in your own words, at a level of detail that would let someone else execute it. If you can't produce the sequence in language, you have a feeling about the technique rather than a memory of it.

This is harder than it sounds, which is the point. Try it now with the last technique you learned. If the sentence comes out as "grip, then take the back", you've found the gap.

The review has to be a question, not a summary. The retrieval-practice finding says the review should force you to reconstruct. So instead of re-reading "back take — hip switch, second hook", write the prompt and leave the answer out: opponent stands up in my half guard with an underhook, what's the first thing I do? Answer it the next day. The discomfort of not immediately knowing is the mechanism, not a sign that it isn't working.

This is also why watching instructionals can feel like learning while producing almost none of it. Recognition is cheap; retrieval is expensive. You can watch a sequence four times and feel a growing familiarity with it — that feeling is fluency, and the research above is fairly blunt that fluency is a poor predictor of whether you'll have the technique in six months. The test is not "does this look familiar?" but "can I produce it, unprompted, in the right position?" Those are different questions, and only one of them is being asked during a scroll through technique videos at midnight.

Space the reviews across days. This is the part that requires a calendar rather than good intentions. Review on day one, day three, day seven, day twenty-one. Four short reviews, spread out, will do more for retention than one long revision session — which is what the spacing literature predicts, and which is consistently misjudged by learners who feel that massed repetition is working.

And log the session the same night if you can. Not for completeness — for the 24-hour effect above. Whatever consolidates overnight appears to work with the material that's available after training. Writing the three things you actually did in that class, while they're still specific, is how the material gets into a form that overnight processing can do something with. Not everything that happens in a session is worth keeping; two or three specifics, written properly, beat a paragraph of "worked on guard". This is the same mechanism that makes logging your rolls work — the writing is the retrieval.

The five-minute version

After class, before you eat:

Write two or three techniques in your own words, at a level of detail where someone else could execute them from your description.
Turn each one into a question without an answer — a prompt you have to reconstruct rather than recognise.
Put three review dates in your calendar: tomorrow, next week, in three weeks.
Answer the prompts on those days, from memory, before checking anything.

That's the whole system. It takes less time than one round of positional sparring, and it's the difference between a technique you were shown and a technique you have.

The real-world version of this — the reviewing, the scheduling, the resurfacing of technique notes at intervals — is what a training log does. But the mechanism is old, cheap, and works with a notebook. What it doesn't do is happen by accident.

Nobody forgets a technique because they didn't care enough. They forget it because the way most people practise — see it, drill it hard once, hope — is close to the least effective schedule the research has found.

Related: Why logging your rolls makes you better — the writing is the retrieval. The spaced repetition system for BJJ technique retention — how to schedule the reviews once the notes exist. And if you're still deciding where to keep them, the 2026 BJJ training log app comparison covers the options.