Practice Design
By OSJJ
Open Source Jiu-Jitsu
THE CURRICULUM SERIES · PART SEVEN
Practice Design
Inside every teaching method there is a second set of decisions that usually get made by habit instead of on purpose: how you actually run the reps. Whole or in parts, all the detail at once or in layers, forward or backward, the same every time or varied, crammed or spread out. These choices shape whether a technique sticks as much as the method does.
Justin Hall / Co-Founder, Open Source Jiu-Jitsu
The last piece laid out the teaching methods, the different ways of getting a technique into a student. This piece goes one layer down, into a set of decisions that sit inside every one of those methods and that rarely get made deliberately. Once you have decided to show a move, how do you actually run the reps? Do you teach it whole or break it into pieces? Do you show every detail on the first pass, or add the detail in layers? Do you build it from the entry forward or from the finish backward? Do you drill the same version fifty times or vary it every rep? Do you run all the practice in one block or spread it out? How much do you correct, and how soon? These are practice-design decisions, and they change whether a technique sticks as much as the method you chose to deliver it.
Here is the thread that runs through all of them, and it is the single most useful idea in this piece: the practice that feels most productive in the room is often not the practice that transfers to a live roll. Blocked, repetitive, single-version drilling looks clean, fast, and satisfying. Everyone is hitting the move by the end of class. And then it vanishes the moment someone resists. The messier-feeling practice, varied, mixed, spread out, looks worse in the drill and holds up better when it counts. It is hard to learn this from the room alone, because the room gives the wrong feedback: it rewards the drilling that looks good today over the drilling that works next month. Learn to read past that and you have a real edge.
One honest note before the dials, because it matters and it usually goes unsaid. The counterintuitive findings below are well established in the laboratory and more mixed when tested in real sports. The likely reason is not that they are wrong, it is that the right amount of variability depends on where the student is. A beginner who does not yet own the basic pattern often needs the clean, repetitive version first. Once the pattern exists, the varied and mixed practice is what makes it durable. So none of these dials has one correct setting. They move as the student advances, and the common mistake is not choosing the wrong setting, it is never moving the dial at all.
Whole versus part
The first decision is whether to teach a technique intact or break it into pieces. This is the one you probably already make by instinct, and it is worth making deliberately.
The rule of thumb. Break it apart when it is complex, multi-step, and the pieces do not depend on continuous rhythm. Keep it whole when it is a single flowing motion that loses its timing when chopped up. A hip escape is one motion, teach it whole. A complex guard pass with three or four sequential decisions is a chain, and for many students the whole chain at once is too much to hold.
In the literature. This is the whole-practice versus part-practice question, and the rule of thumb above is the classic one: part practice pays off when a task is high in complexity and low in organization, meaning the pieces do not depend tightly on one another, and whole practice wins when the pieces are tightly coupled in time. The research sorts the ways of breaking a task down into three families. Segmentation splits it along time, first movement then second. Fractionation splits it along simultaneous components, upper body and lower body practiced separately. Simplification keeps the whole task but makes it easier, a slower partner, a bigger target. Everything in this section is segmentation. When the whole-versus-part studies are pooled there is no global winner, which is the point: it depends on the task and the learner.
The anchor example. Take a complex passing sequence. Show the whole thing start to finish and watch the room. Some students absorb the entire chain and can run it back. Others glaze over by the second step, and if you leave it whole they will never get it. Those students need it segmented: work the first movement, then add the second, then the third, joining the pieces as they go. Same technique, two completely different presentations, and the choice between them is not about the move, it is about the person in front of you. That last point is the seed of the whole next article.
How I actually run it in parts. When a technique is complicated enough to break up, here is the sequence. I still show the whole thing first, silently, a couple of times, so the room knows where all of this is going. Then I say it plainly: we are going to break this into three parts. I show the first part only, say, opening the legs and getting past the knee line, and the room drills just that. Then we come back. I show the first part and the second part together, and they drill both, joined. Then we come back again and I show all three, first through third, and they drill the whole pass. Each showing includes everything that came before it, so the pieces get joined as they are learned rather than stitched together at the end. In the literature this is the progressive-part method, sometimes called the cumulative-part method: part one, then one plus two, then one plus two plus three. It is generally considered the strongest of the part-practice variants because the joins between parts get practiced instead of left for last. Wrapped in the silent whole showing at the start and the full pass at the end, the entire structure is what the textbooks call whole-part-whole.
Two things to notice about that. First, the practice breaks are the whole point. This is the one place in the piece where I do stop and let the room drill between showings, and the reason is length: a three-part pass with every detail is too much to hold in one sitting, so the parts have to be earned one at a time. Second, the layering method from the next section runs inside every one of those showings. When I show part one, I show it in layers, big mechanics first and finer detail on each pass. When I show parts one and two, same thing. Breaking into parts and layering the detail are not two competing ways to present a technique. Parts decide how much of the technique goes into a showing. Layers decide how much detail goes into each pass of it. On a complicated technique, I use both at once.
Sit with that for a moment, because it is easy to slide past. The decision to show a pass whole or in pieces is not a decision about the pass. The pass is the same pass. It is a decision about the student, and the moment you see it that way, something shifts in how you teach. You stop asking what is the right way to present this technique and start asking what is the right way to present it to this person. Two students, same move, two presentations, both correct. That is not a compromise or an inefficiency. It is the first, smallest instance of a principle that runs all the way up through everything that follows in this series: the people in front of you are not the same, and no single presentation reaches all of them.
Hold onto that idea. It is going to come back at the scale of the entire class in the next piece, where the question is not how you present one technique but which teaching methods you use at all. The answer there is the same as the answer here, and this small example is the proof of it in miniature.
All at once versus in layers
The whole-versus-part decision has a close cousin that governs the demonstration itself rather than the drilling: how much detail you put into each showing. You can demonstrate a technique once, with every detail in it, and send the room to drill. Or you can demonstrate it four or five times, starting with no words at all, and add a layer of detail on each pass. Coaches call this layering, or the onion, or building the technique in passes. The literature has a more precise account of why it works, below. It is one of the most useful habits I have, and it is worth making deliberate.
What it looks like. This all happens in one demonstration, with the room watching, before anyone drills. The whole thing takes me three or four minutes. It is not show, drill, show the next layer, drill again; that would take the entire class and nobody has that kind of time. It is one continuous showing that gets finer as it goes. First pass, after the silent ones: the big mechanics, start to finish. Where the body goes, what the shape is, what the technique is for. The obvious, major movements only. Then I might rotate so the room sees it from a different side, run it again, repeat the same major points, and layer in the next set of details: the grips, the first thing that makes the movement actually work. Run it again, same major points, next layer: the angle, the hip position, the timing cue. Fourth and fifth passes go micro: where the weight sits on the foot, which finger closes the grip first, the half-inch of head position that decides whether the pass holds. Every pass is the same technique, shown start to finish. Each one carries everything from the pass before it and adds one level finer. Then the room drills, and every student walks into the drill carrying the layer that met them.
Layer zero: show it before you say anything. There is a step that comes before the first layer, and it is the one I would fight hardest to keep. Before a single word of explanation, demonstrate the technique two or three times in silence. Once slow, once at real speed, maybe once more. Then sit down and start layering.
Here is why. Watch what happens when a coach walks onto the mat and starts talking on the very first demonstration. They move through the technique and narrate every detail as they go, grips, angles, weight, timing, all of it on the first pass. By the end, the room has heard all of it and registered almost none of it, and it is not because the students were not paying attention. It is because on a first viewing the student’s entire attention is spent on one question: what is this? What are we about to learn? Until that question is answered, there is nowhere for the words to go. The eyes are working and the ears are, for practical purposes, switched off. Every detail spoken over that first showing is spoken into the wrong moment.
The silent demonstration answers the question first. After two or three quiet passes, everyone in the room knows what the technique is, roughly what it looks like, and roughly where it ends. Now the ears are free, and now the layers have something to land on. It costs about ninety seconds. In cognitive-load terms, you are separating the two jobs, seeing the movement and hearing about the movement, instead of asking a student to do both at once on the first pass, when the first job takes everything they have. There is research behind that. The observational-learning literature holds that what an observer actually extracts from a demonstration is the relative motion, the coordination pattern, how the body parts move in relation to each other, and that this is the information a learner picks up first and most easily. That is the shape. The coaching-science reviews of demonstration and instruction add the caution that piling verbal instruction on top of a demonstration can narrow what the learner takes from it rather than add to it. Coaches call the practice a silent demonstration, or show before you tell. The science says the same thing in different words: let the coordination pattern go in clean, then talk.
The first reason it works. Almost nobody absorbs every detail on a first showing. Working memory holds a handful of things, and a technique shown once with ten details in it hands the student ten things to hold while their body is still trying to find the gross movement. What they actually keep is the shape, and the details fall out the back. Layering respects the limit instead of fighting it. The shape goes in first, on its own, and each following layer lands on a movement the student can already do. The details have something to attach to. This is cognitive load theory applied to a demonstration. Working memory can only hold a few elements at once, and a technique with ten interacting details is high in what the theory calls element interactivity: every detail has to be held in relation to the others, which is what overloads a beginner. The specific fix has a name, the isolated-interacting elements effect: present the elements one at a time first, then present them together, and learners who would have been overloaded by the full version learn it. Layering is that effect run live on the mat, one pass at a time.
The second reason, and the bigger one. A room is rarely one level. The same class holds a white belt in week three, a blue belt with a year in, and a brown belt who has hit the technique a thousand times. Shown once with all the detail, that demonstration is pitched at exactly one of them. Layered, it reaches all of them in the same five minutes. The white belt takes the first layer, the shape, and that is the whole lesson for them, because the finer details are not yet in their awareness and would not land if you forced them. The blue belt already owns the shape and picks up the second or third layer, which may be the first genuinely new thing they have heard about this technique in months. The fourth and fifth layers are for the advanced students, the micro details that are the only reason they still need a demonstration at all. Nobody is asked to hold what they are not ready for. Each student takes the layer that meets them, and the rest passes over them without doing any harm.
On the mat. The easy mistake is to front-load, to put every detail into the first talking pass because you know it and it feels incomplete to leave it out. Resist that. Let the first pass be almost embarrassingly simple, and keep repeating the big mechanics on every pass so the beginner keeps hearing the thing they need while the detail stacks up for everyone else. It is a settings dial like every other one here. A room of pure beginners may only need two or three layers today, with the finer ones saved for the technique’s next visit. A room of purple belts can start at layer three and spend the time on the micro. And the drill afterward tells you whether you read the room right: if the white belts come out of it without the shape, the first layer was not simple enough or the later layers buried it.
Notice what this dial is really doing, because it will matter in the next piece. It is one demonstration engineered to reach several different learners at once. Hold that thought.
Forward versus backward chaining
When you do break a technique into parts, there is a second choice that is easy to miss: which direction do you build it?
Forward chaining. Start at the entry and add pieces toward the finish. This is the default, and it is how the majority of coaches teach. The student learns step one, then one plus two, then one plus two plus three. It is intuitive because it matches the order the technique happens in.
Backward chaining. Start at the finish and work backward toward the entry. The student first learns the last movement from a position where it is nearly done, then the second-to-last leading into it, and so on. It is underused and it is powerful for exactly the complex sequences that trip people up, because every rep ends in success. The student always finishes the technique, they just start it from progressively further back. It removes the experience of repeatedly failing at the end of a long chain, which is where most drilling frustration lives.
When to use which. Forward for simple sequences and for when the entry is the hard part. Backward for long chains, for finishes that need to be automatic, and for students who are getting demoralized by never reaching the end. Many of the best coaches do both on the same technique without naming it, forward on the first pass, backward when a student is stuck.
In the literature. Chaining comes out of behavior analysis, where a chain is a sequence of responses in which each step cues the next, and forward and backward chaining are the two standard ways to teach one. In motor learning it lives inside the segmentation family above. Backward chaining has the more interesting evidence: in flight-simulator training, practice that built a maneuver from the end backward transferred to the whole task better than practicing the whole maneuver from the start, and the explanation offered is the one above, every rep ends at the goal. The progressive-part method described in the last section is forward chaining with cumulative parts.
Blocked versus random
This is the dial with the strongest research behind it and the most counterintuitive result, so it deserves the most care.
What the terms mean. Blocked practice is drilling the same thing over and over, twenty armbars in a row. Random practice is mixing several things up, an armbar, then a triangle, then a sweep, in an unpredictable order, so the student has to reload the movement each time instead of repeating it on autopilot.
The finding. This is the contextual interference effect, and the Shea and Morgan experiment from 1979 is the one everyone cites. In those experiments, students who practiced in a blocked schedule performed better during the practice session, and students who practiced in a random schedule performed worse. Then, on tests days later, it flipped. The random group retained more and transferred better to new conditions. The effort of constantly switching, which feels like it is slowing you down, is what builds the durable version of the skill. This has been replicated many times, and the discomfort of it is the mechanism, not a side effect.
The honest caveat. That effect is strong in the lab and weaker in real sports. Reviews of applied and sport studies, including the more recent meta-analyses, find the effect is smaller and less consistent with real athletes and real skills than in the laboratory tasks it was discovered on. The best reading of the evidence is that it depends on the learner. A beginner who cannot yet perform the movement at all gets little from random practice, they need enough blocked repetition to own the basic pattern first. Once the pattern exists, mixing things up is what makes it hold. So blocked is a beginner tool, random is what you graduate to, and the error is never graduating.
On the mat. Early in learning a technique, blocked reps to get the shape. Then, sooner than feels comfortable, start mixing it with other techniques, other positions, other partners, so the student has to find it rather than repeat it. If drilling still looks perfectly clean after a few weeks, that is usually the sign it has stayed blocked too long.
What it looks like in a class. Say the class has three techniques in it this week: an armbar from mount, a hip escape from under mount, and the scissor sweep. The blocked version is the one the majority of academies already run. Ten minutes of armbars, partner lies there, twenty reps each. Ten minutes of hip escapes. Ten minutes of sweeps. Clean, orderly, and by the end of each block the room is hitting the move on autopilot. The random version uses the same three techniques and the same thirty minutes, but the reps are shuffled. The simplest way to run it is a call-out: partners in position, you call the technique, they hit it once, reset, you call a different one. Armbar. Sweep. Escape. Sweep. Armbar. Nobody knows what is coming, so every rep starts from a cold reload instead of a warm repeat. A version that needs less of your voice is the carousel: three stations on the mat, one technique per station, pairs rotate every rep or every two reps instead of every ten minutes. And the version that fits a real curriculum best is the mix-in: drill this week’s technique blocked for the first stretch, then for the last ten minutes fold in the techniques from the last two or three weeks and run them in random order, so the new move gets learned in a blocked schedule and stored in a random one.
There is a middle setting with its own name, serial practice: the techniques rotate in a fixed order, armbar, escape, sweep, armbar, escape, sweep. The student still has to switch every rep but can see the switch coming. It produces some of the random benefit with less of the chaos, and it is a sensible first step for a room that has only ever drilled in blocks. What you will notice the first time you run any of these is that the drilling looks worse. Reps get sloppier, the pace drops, students hesitate for a beat before each one. That is the effect working, not failing. The cost shows up in the drill and the payoff shows up in the roll.
The same dial at the scale of the curriculum. Blocked versus random is a within-class decision, but it has a twin one level up, and you have already met it in Part Two. Take a modular rotating curriculum of the kind Gracie Combatives uses, and that many academies, ours included, use a version of: thirty-six lessons, each with two or three techniques, and the techniques inside a lesson are not connected to each other or to tomorrow’s lesson. Inside each lesson the drilling is blocked, each technique in its own clean slot. But across the days it is closer to random: the material arrives in unconnected pieces, in whatever order the student happens to walk in, and the student has to reload each one from scratch. A positional curriculum runs the other way. Inside a block everything is chained to the position, control feeding submission feeding escape, so the reps within the block are constantly switching between related things, and the gap between positions is where the isolation happens. Neither is right, and I am not going to argue here for which produces faster learning, because that is a curriculum question and it belongs in the second half of the series. What the practice-design lens shows is only where each structure has set this dial: the modular structure gets the clean, blocked acquisition of individual techniques, and leaves the connections between them to the student. The positional structure gets the connections built in, and gives up some of the clean isolation. Which one your curriculum should be is the ascent half of this series. The point here is only that the dial exists at both scales, and the tradeoff is the same at both.
Constant versus variable
Closely related to the last dial, and the one most worth building a habit around, because it is the one that most directly prepares a student for the mess of a live roll.
What it means. Constant practice is drilling one technique in one configuration, the scissor sweep from exactly one angle, every rep. Variable practice is drilling the same technique across different conditions, the same sweep from ninety degrees, from forty-five, from across the chest, three, four, five configurations of the one movement.
In the literature. This is the variability of practice hypothesis, which comes out of Schmidt’s schema theory. The theory holds that what a learner stores is not a fixed movement but a rule, a schema, relating the conditions to the movement that worked, and that practicing across varied conditions builds a stronger rule than practicing one condition many times. The prediction that variable practice transfers better to novel versions of the skill has been tested extensively, most strongly in children and in tasks where the test condition was never practiced. Differential learning, discussed below, is Schöllhorn’s separate program and is not the same thing.
The example. I do not teach the scissor sweep from one angle. I teach it from three or four, sometimes five, based on the different ways the position actually shows up. Ninety degrees. Forty-five. Across the chest. Each one is a real, legitimate version of the same sweep, and the student learns all of them as options.
Why it works. The sweep never shows up in a live roll in exactly the configuration you drilled. A student who drilled one angle owns one angle and is lost the moment the geometry shifts. A student who drilled five owns the sweep itself, the principle underneath the angles, and can find it wherever it appears. Same tradeoff as blocked versus random: the varied practice looks messier in the drill and holds up far better live, because the student was never building a single fragile path, they were building the movement.
Not to be confused with differential learning. Variable practice is still direct instruction. You prescribe every version, and every version is real and meant to be kept. Differential learning, which we covered in the last piece, goes a step further, deliberately introducing imperfect, exaggerated, never-repeated variations that are not meant to be kept at all, only to disturb the movement so the body learns the principle. Three points on a line: one version, several good versions, deliberate perturbation. Teaching more angles is the upgrade almost any coach can make tomorrow.
Let me stay on this one longer, because it is one I use constantly and one I want a coach to walk away actually using. Here is where the angles come from. Show the scissor sweep from one angle and, before the room has even paired off, a hand goes up. But what if he does this? What if he leans back? What if he goes this way? Most coaches hear it most classes, it is usually a white belt, and it is usually a little annoying. It is also a fair question, and the honest answer is: then it does not work, not that way. If your opponent changes his angle or his weight distribution, you have to change your angle and your weight distribution to match. The variations I teach are the answers to those what-ifs, worked out in advance and shown as versions of the sweep instead of argued about one rabbit hole at a time.
That does two things at once. It keeps the room out of the rabbit holes, because the what-if has already been shown as a version rather than debated as an exception. And it teaches something bigger than the sweep: the technique is not a fixed shape, it is a relationship to what the opponent is doing, and when he changes, you change. The reason it works goes deeper than giving them more options. When a student drills the sweep from ninety, forty-five, and across the chest, they are not learning three sweeps. They are triangulating on the one thing all three share, the actual mechanic that makes the sweep work regardless of angle: the hip position, the timing of the leg drive, the way the top player’s weight has to be committed. The angles are scaffolding. Vary enough of the surface and the student is forced to find the principle underneath, because the principle is the only thing that stays constant. That is why variable practice transfers, it teaches the invariant, not the instance.
There is a practical way to build this into any technique, and it is simple. Ask what actually varies when this move shows up live, and drill across that. For the sweep it is the angle and the opponent’s posture. For a guard pass it might be whether they have an underhook, whether their knees are in or out, whether they are on their side or flat. For a takedown it is the grip and the stance. Pick the two or three dimensions that really change in a roll, and run reps across them. Three angles is better than one. Five is better than three, up to the point where the student cannot keep the pattern at all, which is your signal that they needed more blocked reps first.
And that last line is the honest limit, the same one that governs every dial in this piece. A brand-new student who cannot yet hit the sweep from any angle gains nothing from five. They need the one clean version until it exists. Then, and this is the part that too often never happens, you open it up. The sequence is not variety instead of repetition. It is repetition to build the pattern, then variety to make it real.
Here is the link forward, said plainly. Varying how you present one technique so that it reaches the student in front of you, five angles so the sweep survives a real roll, is the same act as varying which teaching methods you use so that a class reaches everyone in it. One is a single move. The other is the whole hour. The principle is identical: the world does not show up in one configuration, and neither do your students, so you build for the variation on purpose. The next piece takes that principle to full scale.
Massed versus distributed
A quieter dial, but a real one, and it runs at two scales: how you schedule reps inside a class, and how you schedule a technique across weeks.
What it means. Massed practice crams the reps together, one long session on a technique. Distributed practice spreads the same reps across multiple sessions with time between them.
The finding. Spreading practice out produces better long-term retention than cramming it, even though cramming feels like more progress in the moment. This is the spacing effect, one of the most reliable findings in all of learning research. It goes back to Ebbinghaus in the 1880s and has been confirmed across hundreds of studies of memory. For motor skill the terms are massed versus distributed practice, and the pooled evidence there points the same way: distributed practice wins for retention, with the largest advantage on simple tasks and a smaller but still real one on complex tasks.
What it looks like in a class. Inside one hour, massed is thirty straight minutes on the scissor sweep, then rolling. Distributed is the same thirty minutes of sweep reps broken into three ten-minute pieces with something else between them: ten minutes of sweep, ten of positional rounds from mount, ten more of sweep, ten of guard retention, ten more of sweep. Same rep count. The second version feels less focused and produces a worse-looking last round of sweeps, because the student has to pick the movement back up each time. That pickup is where the retention comes from. Across weeks it is the same shape stretched out: a technique taught once in a two-hour deep dive versus the same technique taught in twenty-minute pieces across four classes over two weeks. The deep dive looks like more, and the student walks out feeling like they have it. The four-class version is what they still have a month later.
The same dial at the scale of the curriculum. This is the dial that most clearly climbs out of the classroom and into curriculum design, and it is where the positional approach has its biggest built-in risk. A position-of-the-month structure is massed practice at the scale of a season: a month of deep, connected, excellent work on side control, and then side control is not seen again for the better part of a year. The depth is real. So is the gap, and the gap is long enough that it stops being spacing and becomes forgetting. The distributed version of the same positional curriculum keeps the position set but tightens the interval, so each position comes back before it fades, at the cost of never going quite as deep in any one visit. The rotating curriculum has the opposite profile: short visits, tight intervals, thin depth. None of these is wrong. Each one has simply set this dial somewhere, usually without noticing, and the setting decides what the students retain. When we get to building a curriculum in the second half of the series, the interval between visits to a position is one of the first numbers you will have to choose on purpose.
Progressive resistance
The last dial is the one that ties practice design to everything else in the methodology section: how you ramp from cooperative drilling to full resistance.
The idea. A technique learned only against a compliant partner is not yet a technique. It has to survive resistance to be real. But full resistance from the first rep prevents the student from ever getting the movement in the first place. So there is a progression to design: start with no resistance to get the shape, then light resistance, then specific resistance aimed at the exact thing being learned, then live.
The connection. This is where practice design meets the constraints-led approach and the implicit tools from the last piece. Errorless learning is progressive resistance in its purest form, design the early reps so failure is nearly impossible, then remove the scaffolding. A constraint game is a controlled step up the resistance ladder. The dial you are turning is how much the partner fights back, and the craft is turning it at the right pace, fast enough to build real skill, slow enough that the student keeps succeeding.
In the literature. Progressive resistance is a coaching term, not a scientific one, and there is no single study that owns it. It sits on top of several things that are. The ladder itself is scaffolding, in the original educational sense of the word: support that is deliberately withdrawn as competence grows. Levels one and two are simplification, the whole task made easier. Level one is close to errorless learning, where skills built with few early errors have been shown to hold up better under pressure. Level three is a constraint in the constraints-led sense. And the reason level four cannot be skipped is what the skill-acquisition literature calls representative learning design: practice has to contain the information the real performance environment contains, and a partner who does not fight back removes most of it.
Here is what that progression actually looks like, using a single technique, a knee-slide pass, run across roughly two weeks. It is not a rigid schedule, the pacing moves with the student, but the shape is the point.
Level one, no resistance. The bottom player does nothing but hold the position. The top player runs the pass start to finish, uncontested, until the movement is clean and they are not thinking about it. This is where the shape gets built. It might take one class or three. Do not rush past it, but do not live here.
Level two, light resistance. The bottom player resists at maybe a quarter effort, enough that the top player has to actually do the technique rather than just move through it, but not enough to stop them. The pass still succeeds almost every time. The student is learning what the movement feels like against a body that is not cooperating.
Level three, specific resistance. This is the level that gets skipped most often, and it is the most valuable one. The bottom player resists hard, but only with the one thing the technique is designed to beat. If the pass is about winning the knee line, the bottom player fights only the knee line, full effort, and gives up everything else. Now the student is getting maximum reps against the exact problem they are learning to solve, without the noise of a full defense. This is a constraint game in everything but name.
Level four, live. Full resistance, real roll, the pass has to actually work. By the time a student arrives here they have built the shape, felt it against a body, and beaten its specific defense hundreds of times. The live version is the test, not the classroom.
Two things to notice. First, the ladder is not one class, it is a technique’s life across several. Second, the levels do not replace each other, a student who has reached level four still benefits from returning to two or three when they hit a wall. The dial goes both ways. The coach’s job is to read which rung each student is standing on and set the resistance there, which is exactly the skill of watching and cataloging that the last piece described.
How much you say, and when
The last dial is the one that runs against a coach’s every instinct, and it has some of the strongest evidence of any of them: how often you give feedback, and how soon.
The instinct. Correct constantly. Every rep, every error, the moment it happens. It feels like coaching. It feels attentive and generous. And in the short term it works, the student’s reps get cleaner right away.
The finding. Frequent feedback improves performance during the session and degrades what the student retains after it. This is the guidance hypothesis, and it has held up across decades of motor-learning research. The clearest single demonstration is the classic Winstein and Schmidt study, where learners given feedback on half their trials retained the skill better than learners given feedback on every trial. When a coach corrects every rep, the student stops building their own ability to detect and fix errors, because there is no need to, the coach is doing it for them. Withdraw the coach and the skill collapses, because the student never learned to feel what was wrong. Reduced feedback, correcting a subset of reps, or letting several go by before saying anything, produces better retention and transfer than correcting everything, precisely because it forces the student to develop their own internal sense of the movement. The literature has names for the versions: reduced relative frequency (feedback on fewer trials), summary feedback (several trials, then one comment on all of them), bandwidth feedback (say nothing unless the error leaves an acceptable range), and self-controlled feedback (the learner asks for feedback when they want it, and learns better than a learner given the same feedback on the coach’s schedule).
The honest caveat. As with the rest of this piece, it depends on the learner. A complete beginner on a complex movement needs more feedback early, they cannot detect errors they do not yet know exist. The reduced-feedback effect is strongest once the student has some grasp of the movement and on simpler tasks. So the dial fades: more correction early, then deliberately less, even when it feels like you are neglecting them.
On the mat. Let reps go by. When you see an error, do not always name it, let the student hit it two or three more times and watch whether they feel it themselves. When you do correct, correct the one thing that matters most, not everything you noticed. And ask before you tell: what did you feel there. A student who can answer that question is building the internal error detection that outlasts your presence in the room. A coach who answers it for them every time is building dependence that looks like progress.
This dial also reframes the great-coach technique from the lesson-plan piece. Cataloging a student’s struggles during discovery and constraint work, then spending that knowledge later in direct instruction, is reduced and delayed feedback done on purpose. The silence during the games is not neglect. It is the student building their own sense of the problem before you give them the answer, which is exactly when the answer sticks.
The through-line, and where this goes
Every dial in this piece points the same direction. Whole, everything at once, forward, blocked, constant, massed, and constant correction are the settings that feel productive, and they are the right settings for a beginner who does not yet own the movement. Part, layered, backward, random, variable, distributed, and reduced feedback are the settings that feel harder and hold up better, and they are where a student needs to be moved as soon as the basic pattern exists. The dials are not right or wrong. They are dials. The whole skill is knowing where the student is and moving the setting on purpose, instead of leaving every dial where it started because the room keeps saying the beginner settings are working.
There is a framework that says all of this in one place, and it is the scientific name for moving the dial. It is called the challenge point framework. It separates a task’s nominal difficulty, how hard the task is in itself, from its functional difficulty, how hard it is for this learner, at this skill level, under these practice conditions. Learning happens best at an optimal challenge point where functional difficulty is high enough to force problem solving and not so high that the learner drowns, and that point moves as the learner improves. Every dial in this piece is a way of adjusting functional difficulty: random practice, variable practice, reduced feedback, and live resistance all raise it, and the framework predicts exactly what the caveats above say, that beginners learn best with those settings turned down and advanced students with them turned up. The memory researcher Robert Bjork named the general principle desirable difficulties: conditions that make practice harder and performance worse in the session, and learning better after it. That is the through-line of this piece, said by someone with a laboratory.
And notice the deepest thing all of this reveals, because it is the door into the next piece. The same complex pass needs to be shown whole to one student and segmented to another. The same demonstration, shown five times in layers, hands the white belt the shape and the purple belt the fine detail in the same five minutes. The same sweep needs five angles because the position varies. The same technique needs blocked reps early and mixed reps later. Every one of these decisions is a version of the same fact: the people in front of you are not the same, and no single way of running a rep reaches all of them. That is true at the scale of one technique, and it turns out to be true at the scale of the whole class. If you have to vary how you present a single move to reach everyone, then you have to vary which methods you use to reach everyone. That argument is the next piece.
Open Minds. Open Mats. Open Source.