CombatScore — AI coaching platform for combat athletes

The problem

Combat athletes train hard but the feedback they get is generic — the same YouTube drills and blog posts that apply to everyone. The signal in every session log — which positions you survived, what you failed to escape, where your energy dropped — goes unused. Gym owners have the same problem at scale: they can see who's attending, but not who's plateauing, who's about to quit, or who's ready for the next stripe.

What we built

CombatScore is a data-driven AI coaching platform for BJJ, MMA, Muay Thai, wrestling, judo, boxing, and similar disciplines. Athletes log training sessions in under sixty seconds; the AI analyzes their last twenty sessions to surface technique gaps and generate a personalized game plan before they step onto the mat.

Gym owners get a parallel toolkit: at-risk student alerts (based on attendance frequency and injury signals), per-student AI reports on strengths and belt readiness, a class planner that generates 60-minute lesson plans from the gym's actual attendance and goals, and a branded storefront at yourgym.combatscore.app with integrated Stripe Connect billing.

Technical highlights

Offline-first mobile sync

Athletes log in the middle of training — often in gyms with unreliable wifi. The mobile app writes locally first and syncs when connectivity returns, with conflict resolution that preserves every session even when the same user logs from multiple devices. Voice logging via on-device transcription means you can describe a roll in thirty seconds without typing.

AI coaching pipeline (Grok / xAI)

Raw session data — positions, durations, energy and focus ratings, success and failure rates per technique — is aggregated into structured context and sent through Grok with prompts tuned specifically for combat sports. The output is specific enough to be actionable ("your bottom side control escapes succeed 22% of the time; here are three drills that address the most common failure mode") without hallucinating techniques that don't exist.

Stripe Connect for multi-tenant gym billing

Gyms bill their own students through the platform, keep the majority of the revenue, and we collect a small platform fee. This is the same direct-Stripe pattern Four Thirteen Studios uses for client work — which is precisely how we learned to trust it. No third-party billing platform in the middle, no percentage skimmed off every membership.

Multi-tenant branded subdomains

Every gym gets its own instance at yourgym.combatscore.app — their logo, their colors, their students, their revenue. One deployment, hundreds of apparent tenants, all sharing the same upgrade path. It is platform architecture, not just app architecture.

Stack

• Next.js + React web dashboard
• Native iOS and Android mobile apps
• Supabase for Postgres, Auth, and Realtime (the platform itself holds SOC 2 Type II)
• Grok (xAI) for AI coaching
• Stripe Connect for multi-tenant billing

Outcome

CombatScore is live at combatscore.app with athletes and gyms actively using the platform across BJJ, MMA, judo, wrestling, and striking disciplines. The platform serves dozens of gyms and hundreds of active athletes today, with thousands of sessions logged and AI game plans generated. Branded subdomains are running for paying gym operators across multiple states; granular figures are available on request under NDA.

Why it matters to clients

This is the proof we can actually ship: a shipped multi-tenant SaaS product built on the same tools and billing model we use for client work. When Four Thirteen Studios quotes you a fixed price for a web or app build, we're quoting from direct experience shipping the same kind of thing for ourselves.