Diving Deep into BJJ through Models: A Novel Approach to Strategy Development & Technique Navigation
All models are approximations. Assumptions, whether implied or clearly stated, are never exactly true. All models are wrong, but some models are useful. So the question you need to ask is not “Is the model true?”
(it never is) but “Is the model good enough for this particular application?”
- Statistical Control By Monitoring and Adjustment (Alberto Luceño & María del Carmen Paniagua-Quiñones)
In the competitive world of video games, there’s a Korean saying, “모르면 맞아야지,” which loosely translates to “If you don’t know, you’ll just have to suffer.” The proverb rings true in many arenas of life, particularly in Submission Grappling, where if you don’t know how to swim with the tides, you will inevitably sink in the ocean of pressure, pins, and submissions in ways you can’t imagine.
As a know-nothing novice filled with zeal to get better at Brazilian Jiu-Jitsu (BJJ), I found myself blending the timeless art of submission grappling with graphical modelling. This tale is about transforming a personal challenge into a trailblazing project, mixing my fresh excitement for BJJ with the mission to leverage technology to map the art of grappling’s complex beauty.
From Parking Lots to the Mats
My adventure into practicing martial arts didn’t start in a traditional dojo with a professor/sensei but in the parking lots and parks next to UC Berkeley (shoutout Underhill). It was here, through sparring with a couple of UFC fanatics, that my interest in combat sports ignited. In this concrete complex, we would glove up with shin pads, focused on sharpening iron with iron and not getting hit. These unrefined, thrill-packed sessions were the stepping stones into mixed martial arts, eventually guiding me to BJJ’s doorstep. I stumbled upon a sport that prizes strategic thinking over sheer force, a shift that completely altered my outlook on martial arts and solving problems. I quickly realized that simply emulating what I saw in YouTube is not enough for efficiently learning how to grapple, so when I had the chance to join a BJJ gym back at home, I had to jump the gun.
The Epiphany of Human Chess
The most common analogy you would hear about BJJ is that it is like human chess. But it is evident to all beginners that rolling is not as evident as looking at a chessboard made of 64 black and white squares, rather a more complex system that requires a nuanced understanding in control of limbs, leverage, and biomechanics. When you’re getting smashed by fierce blue belts, your wits are not necessarily the first line of defense, denoted by helpless flailing and squirming. It would be nice to look at any entanglement and recognize the dynamic like looking at a chessboard. A mere three weeks into learning BJJ with a bucket load of taps, I was convinced that there was a glaring opportunity gap: the need for a system that demystifies the intricate dance of grappling, translating its fluid movements into an accessible format. This challenge sparked a daring idea — a product that acts like a compass for BJJ players, helping them visualize and plan their moves with clear precision. This tool aims to transcend traditional training limitations by offering a dynamic visualization platform where techniques and strategies become clear, actionable knowledge. It promises not just to enhance the learning curve for beginners but to provide seasoned practitioners with a laboratory for strategy experimentation and refinement. Besides, not everyone has the money to buy instrucitonals.
Capturing the Spirit of Grappling in Models
The main hurdle was how to encapsulate BJJ’s spontaneous and fluid essence while coming up with a robust taxonomy for thousands of different techniques and situations. Initially, my background in operations research allowed me to think of Markov Chains to model the processes of a BJJ match. With its nodes and arcs, it seemed promising to model the big picture of the path to submission, traveling through various nodes, representing the outcomes of techniques. However, the stochastic model would require assigning arbitrary weights and transition probabilities, of which there is no convincing way of assigning until empirical computation through thorough data collection is complete. Also, it would be more important to define conditions that allow transitions between nodes, rather than transition probability itself, for understanding what it takes to dominate an opponent and remain in control.
Hence, I’ve come to a decision to build a deterministic model rather than a stochastic one. I narrowed my options on using Finite State Machines (FSMs), Statecharts, and Petri Nets as a prototype for exploration and strategy development modules, enabling the mapping of BJJ’s positions and transitions in an organized fashion focusing on categories of possible control positions, transitions (e.g. passing guard, takedowns, sweeps) and prerequisites (e.g. grips, hooks, wedge). This method offers a blueprint for users to follow, linking one movement to the next with strategic purpose.
The following endeavor involves several underlying assumptions that make these mathematical models suitable for capturing the complexity and fluidity of the game. These assumptions facilitate the translation of physical movements and strategies into a structured, analyzable form, aiming to enhance understanding, teaching, and strategic planning in BJJ.
Here’s a look at some of these core assumptions:
1. Discrete States and Transitions
- Both FSMs and Statecharts operate under the assumption that the dynamics of BJJ can be discretized into a finite set of states and transitions between these states. This assumption simplifies the continuous and often complex flow of BJJ into a manageable set of components that can be analyzed and understood systematically.
2. Hierarchical and Complex Behaviors
- Statecharts extend the concept of FSMs by introducing the assumption that BJJ dynamics can exhibit hierarchical and parallel states. This reflects the idea that certain positions or situations in BJJ can encompass nested levels of complexity, where multiple sub-positions or actions might occur concurrently or within a broader strategic framework.
3. Concurrent States and Transitions
- Petri Nets assume that BJJ dynamics can involve concurrency, enabling the modeling of multiple, simultaneous transitions. This is particularly relevant in BJJ, where a practitioner might set up several offensive or defensive maneuvers at the same time, or where actions and reactions between two practitioners occur in parallel.
4. Conditions and Effects
- These models assume that transitions between states in BJJ are conditional, often requiring specific criteria or triggers (e.g., an opponent’s action, kuzushi, or the availability of a limb). This reflects the strategic depth of BJJ, where practitioners must continuously adapt to the evolving conditions of engagement.
5. Determinism and Non-determinism
- While FSMs traditionally assume a deterministic progression from one state to another, the application of these models to BJJ also has to account for the non-deterministic nature of human competition, where the same starting position might lead to various outcomes depending on the practitioners’ actions, skills, and decisions. The use of non-deterministic or probabilistic models acknowledges the unpredictability inherent in BJJ encounters.
6. Quantification and Analysis of Dynamics
- By using these frameworks to model BJJ, there’s an assumption that aspects of BJJ can be quantified and analyzed using these discrete models. This includes the ability to analyze the efficiency, effectiveness, or predictability of certain positions or sequences, aiming to derive insights that can inform training and competition strategies.
Using frameworks like FSM, Statecharts, and Petri Nets to model BJJ dynamics assumes that the physical and strategic aspects of the martial art can be abstracted into mathematical terms, allowing for a systematic exploration of its principles. While these models cannot capture the full breadth of human intuition and adaptability, they offer valuable tools for deconstructing and understanding the complex interactions that define BJJ, providing practitioners with structured approaches to analyze and improve performance.
The Tech Stack
In order to turn these models into tangible products, I started imagining what the technical requirements for the product feature would be. I would know my product succeeded in its mission if it is able to map the progress of a grappling match in a cohesive and intuitive manner. So, I chose React for its nimble ability to update and display user interfaces, and Django for its solid and swift development traits. This duo promises a sturdy base for crafting an app rich in data that could adapt and grow alongside its user base. However since I don’t come from an extensive background in software development, Generative AI, namely ChatGPT, emerged as a critical partner for product development. It was instrumental not just in coding but also in the cyclical process of enhancing the product’s design and features. This highlights the transformative impact of technology in bringing creative visions to life, no matter the starting point.
Looking Forward
As the product’s base features begin to solidify, attention turns to polishing the model, developing UI/UX, integrating feedback from users, and delving into advanced functionalities like computer vision. The aim is to forge a tool that simplifies BJJ for newcomers and offers fresh insights to veteran enthusiasts. This endeavor celebrates the union of personal passion and existing technology. It serves as a conduit between the tangible and the imaginable, opening new avenues to engage with and comprehend submission grappling.
I’m inviting BJJ practitioners, developers, and product managers to be part of this voyage. Your perspectives, skills, and input are vital in sculpting a tool poised to redefine our engagement with Brazilian Jiu-Jitsu.
Keep an eye out for future updates, deep dives into the tech behind the scenes, and conversations at the crossroads of martial arts and technology.
