GrappleMind Sign up for early access

GrappleMind / How it works

How GrappleMind analyzes a BJJ match

4 min read5 sections
The short answer

Upload a match, and GrappleMind marks every guard pass, sweep, takedown, submission attempt and turning point, so your review starts where the learning is.

The problem: rewatching instead of learning

Every competitor knows the routine. You film the match, sit down to review it, and spend most of the time waiting for something to happen. Your memory of the match is not reliable either: adrenaline makes it subjective, the moments you remember are rarely close to what actually happened, and you miss the ones that decided it.

We call that first watch the staring phase. The real work only starts in the second phase, critical viewing: rewinding one exchange, seeing what happened, and finding the hole, the mistake or the missed opportunity.

What GrappleMind does: it removes the staring phase. You open a match and go straight to the moments that matter, each one marked and explained, and every insight you take from it becomes something to train.

How GrappleMind recognizes what happens in a match

The analysis is not a generic video model guessing at grappling. It works from a rulebook written by competitors.

1. A rulebook written by a black belt world champion

Amit Burshtein, black belt, UWW World Champion and IBJJF No-Gi European Champion, wrote precise definitions for every position and transition the engine looks for: more than 35 event types, from guard passes, sweeps, takedowns, mount, knee on belly and back takes to submission attempts, near passes, leg lock exchanges, guard pulls, scrambles, resets and penalties.

The definitions are exact on purpose. A sweep needs a real guard and real height, not just a change of top position. Danger is not a pass. An opening only counts if it actually took hold. These are the same distinctions a referee and a high-level coach make.

2. Every moment is marked, typed and attributed

A video AI model reads the whole match against that rulebook. Each event comes back with its start and end time, its type, which athlete did it, the score when the referee gave one, and a short description of what happened, so you can jump straight to it.

The engine follows a clear order of priorities: first the event, then the athlete, then the score. Identifying correctly that something happened matters most. That is why it works on imperfect footage too.

3. A second, sharper look at how it ended

Finishes happen fast. After the full pass, GrappleMind cuts the end of the match at full resolution and takes a focused second look: how the match ended, and who finished whom. If the two passes disagree, the moment is flagged for review instead of being guessed.

4. It learns from champions’ corrections

World-class competitors review analyzed matches and correct what the engine got wrong. Every real correction, across every reviewed match, is carried into future analyses, so the same mistake is not repeated.

What you get from it

  • Efficient rewatching. Every point, attempt and turning moment, marked on the match timeline.
  • A fight breakdown from your point of view. What worked, what to improve, the key turning points, and why you won or lost.
  • Insights that become training. Each conclusion becomes a task you can sort into folders, for example a folder that groups similar moments across matches.
  • A plan to follow. A weekly training plan, a streak, and a countdown to your next competition.
  • A coach in the loop. Connect your own coach, or a pro coach through the app, to review your matches with you.

Which matches can you analyze?

Any match you can film.

  • Competition and training. A tournament final or a hard round at your academy.
  • Gi and No-Gi. The engine knows which one it is watching and does not invent lapel or collar grips in a No-Gi match.
  • IBJJF and other rulesets, like AJP and ADCC. The score follows the referee when it is visible, and the events are found either way.
  • Any footage. Official streams, or a phone filming from the stands, cropped or partly blocked.

GrappleMind marks what both athletes did. If your opponent passes your guard, you see that moment too, because that is exactly what you need to learn from.

Questions people ask

How does GrappleMind recognize what happens in a BJJ match?

A video AI model watches the whole match against a detailed rulebook of more than 35 event types written by black belt world champion Amit Burshtein. Every event is timestamped, typed and attributed to an athlete, a second high-resolution pass checks how the match ended, and every correction made by expert reviewers is fed back into future analyses.

What problem does GrappleMind solve?

Rewatching a full match to find the few moments that matter takes time, and memory of a match is unreliable. GrappleMind finds those moments for you, so your review starts with the critical exchanges and ends with clear things to train.

Which matches can I analyze?

Any BJJ match: competition and training rolls, Gi and No-Gi, IBJJF and other rulesets like AJP and ADCC, from an official stream or a phone filming from the stands.

Your next match starts with the right breakdown.

Upload a match, and GrappleMind marks every moment that mattered.

Sign up for early access