The coaching method

A method for the decision a coach actually makes: not what to teach next week, but what this person in front of you cannot yet do, which of those problems is worth the next month, and how to build practice that tells you whether it changed.

The theory of control explains the exchange between two grapplers. The coaching method answers a different question:

What should a coach repeatedly do to change a particular athlete's probability of winning under declared constraints? The word that matters is repeatedly. The method is a closed loop, run again and again, and it is universal in its process rather than in what it produces: two coaches running it on two athletes should arrive at different technical content by the same kind of process.

The loop1–2Define the task, observe the athlete.What are they trying to win, under which rules, against whom; and what actually happens when they grapple.3–4Diagnose, then prioritise.Which class of problem is limiting them; which of the candidates is worth working on, compared without collapsing the reasons onto one scale.5–6Construct and compress the solution.Refine, import, synthesise or prune; then put it in a form the athlete can use.7–8Choose the mode and the autonomy; design representative practice.How much instruction, how much athlete ownership, and a task that puts the problem under pressure.9–11Execute, assess, read the field.Gather evidence while practising; assess in a way that changes decisions; treat rolling and competition as feedback.12Revise, and go again.Update the athlete model, the opponent model and the current policy, and select the next target.The twelve steps of the method, grouped. It starts from the task and the athlete, never from a technique, and it ends by revising the model that produced the last decision.Model the athlete, and separately the opponent

The loop runs on a model of the athlete, and the model is deliberately not a number. It holds what this athlete is physically constrained from doing, what they can execute reliably, what they recognise, how they select, what technical repertoire and connected systems they have, what they can reach, force and preserve, how robust they are under resistance and disruption, how their game transfers across rulesets, their fatigue and injury constraints, their learning state and capacity for autonomy, and their preferences, style and technical history. No single figure stands in for that profile, and the method treats an athlete who has been reduced to one as unmodelled.

The opponent gets a model of their own, named or as a population: likely attacks, controls, transitions and reactions, strategic routes, how they exploit the rules, their constraints, and the uncertainty around all of it. Where no named opponent exists, the field the athlete is likely to meet stands in. The two models are kept apart because a plan built from one of them alone is a plan for a different fight.

An athlete keeps getting passed from half guard. The instinctive response is to teach a new half-guard technique, and the method forbids that instinct until the failure has been placed. For any observed failure, the question is which of nine classes is limiting the athlete: availability, capability, observation, selection, enforcement, viability, the opponent’s contribution, value and context, or robustness.

Placed, the same failure becomes four different problems. The underhook was never available because the passer’s crossface arrived first: an availability problem, and the fix is earlier. The underhook was available and the athlete cannot yet execute it against a heavy partner: a capability problem. They can execute it and did not see the moment: an observation problem. They saw it and chose to hold the knee shield instead: a selection problem. From outside, all four look like the same passed guard. A coach must not assume a new technique is the answer before determining which class, or which combination, is actually limiting performance.

Once the problems are placed, one has to be chosen, and the method is exact about how. Nine considerations enter the judgement: the competitive value if the problem improved; how often and how badly it fails; how reachable meaningful improvement is; the time it would take against the athlete’s horizon; how far the improvement transfers across likely opponents; what other capabilities it depends on; the fit with this athlete; the training-time and fatigue cost; and the risk. None of them is combined into a weighted score, a fixed numerical weighting or a universal ranking. There is no mastery score, no gate-debt total and no readiness percentage, and the theory rules those out rather than merely omitting them.

What replaces the score is a non-dominated set. The coach first removes candidates that are inadmissible under the declared constraints or clearly dominated by another candidate on every consideration that matters. If one admissible candidate is clearly preferred, it is the target. If several remain genuinely incomparable, the coach makes an explicit provisional judgement among them, records why, what is uncertain, and what evidence would change the choice, and revisits it when the evidence arrives. “Highest-value realistically changeable problem” means the preferred admissible target that process produces, and it never means that every coaching problem sits on one hidden scale.

Two things follow that a coach can feel tomorrow. Priority is value- and constraint-driven, never syllabus-gap-driven: “we have not covered turtle yet” is not a reason. And the decision about what to work on belongs to this loop, not to the curriculum, the rank, or the templates.

For a selected problem the coach may refine a solution the athlete already has, import a known technique or concept, synthesise a combination, delete a fragile branch, build a response-conditioned system, or create a variation specific to an opponent or a ruleset. Where none of that is adequate the method runs a research loop: define the technical problem, inspect the constraints on both athletes, search existing exemplars, keep the provenance, test the mechanics, test the opponent’s responses, reject the non-viable branches, integrate the useful ones, stress-test them in representative practice and reassess in rolling. Several valid solutions may coexist for one problem and one athlete.

The solution then has to be put in a form the athlete can use, and the method treats every such form as an interface rather than as a fact about grappling. The same underlying problem can be presented as any of these: