Talent is only part of the story. Explore the market, the brand, and the relationships behind basketball’s value.
Player potentialSchool contextFan connection
THE BIGGER PICTURE+
EXPLORE THE DEMO
Your next perspective starts here.
Claude Opus 4.6
Explore an athlete’s potential.
Choose an athlete and a school scenario to build your research brief.
START WITH AN EXAMPLE
iAI research, not verified valuations. School choices are hypothetical scenarios; school factors use archived 2025 assumptions.
RESEARCH IN PROGRESS
Connecting the perspectives.
Your research request is starting.
Keep this tab open. School comparisons take longer because they include two reports.
We couldn’t complete that request.
RESEARCH BRIEF
Experimental AI analysis. Verify claims against original sources before using them in a decision.
GO A LEVEL DEEPER
Understand the thinking.
The ideas, frameworks, and technology behind the demonstration.
Claude Desktop Integration
Access the Basketball AI Analysis Platform directly from Claude Desktop using the Model Context Protocol (MCP).
Run full 12-agent analyses without opening a browser!
What is MCP
What is MCP Integration?
Model Context Protocol (MCP) is Anthropic's standard for connecting AI assistants to external tools and data.
With our MCP server, you can analyze athletes and teams using natural language directly in Claude Desktop!
Natural Language Ask in plain English
Full 12-Agent System Complete analysis
Claude Opus analysis Fast & comprehensive
Works in Chat No browser needed
Example Queries
Example Queries
Once installed, try these queries in Claude Desktop:
"Analyze Cooper Flagg's NIL value"
"What's Paige Bueckers NIL potential at UConn?"
"Compare Cooper Flagg's NIL at Duke vs Kentucky"
"Analyze Duke Blue Devils fan psychology"
Installation
Installation Options
Note: MCP integration requires the server code to be installed locally. This feature is currently available for demos, internal use, and evaluation purposes. Contact us for access to the MCP server package.
Prerequisites: Python 3.10+, Claude Desktop app, MCP server files
The installer will automatically configure Claude Desktop for you!
Step 4: Restart Claude Desktop
Quit Claude Desktop completely (Command+Q on Mac, or fully exit on Windows), wait a few seconds, then reopen it.
Option 2: Direct API Access (Coming Soon)
We're working on a hosted MCP server that won't require local installation.
This will allow you to connect Claude Desktop directly to our API with just a configuration file change.
Stay tuned!
Option 3: Enterprise/Evaluation Access
Interested in deploying this for your organization? We can provide:
MCP server package with setup support
Hosted MCP service (no local install needed)
Custom deployment to your infrastructure
Training and integration support
Contact us to discuss your needs and get started!
Testing
Testing Your Installation
Try this in Claude Desktop:
"Analyze Cooper Flagg's NIL value"
Claude will call your MCP server and return the full 12-agent analysis in ~30 seconds!
Available Tools
Available Tools
analyze_nil
Purpose: Analyze athlete's NIL market value
Agents: 12 (with coordination)
Duration: a few minutes
Returns: Financial valuation, VALORE scores, strategy, risks, opportunities
analyze_parasocial
Purpose: Analyze fan-team psychology
Agents: 6 (hybrid parallel)
Duration: 25-30 seconds
Returns: 5-dimension fan psychology analysis
compare_schools
Purpose: Compare NIL potential at 2 schools
Agents: 25 (12 + 12 + 1)
Duration: 60-90 seconds
Returns: Side-by-side comparison with synthesis
get_school_market_factors
Purpose: Get school NIL ecosystem info
Duration: < 1 second
Returns: Informational guide
Benefits
Why Use MCP Integration?
Lower Barrier: No need to navigate to a web interface - just ask Claude
Natural Workflow: Fits into your existing Claude Desktop conversations
Full Analysis: Same comprehensive 12-agent system as the web interface
Quick Queries: Perfect for rapid ad-hoc analyses during meetings or research
Multiple Access Methods: Use whichever interface suits your workflow
Troubleshooting
Troubleshooting
If Claude doesn't see the tools:
Check Claude Desktop → Help → Debug Info
Look for "basketball-ai" under MCP servers
Make sure Claude Desktop was completely quit and reopened (not just window closed)
Check logs at ~/Library/Logs/Claude/mcp*.log
Documentation
Getting Access
Request MCP Server Access
The MCP integration demonstrates our platform's extensibility and commitment to multiple access methods.
If you're interested in evaluating this feature or integrating it into your workflow,
please reach out and we'll provide the necessary files and setup support.
What you'll get: MCP server code, installation scripts, full documentation,
and setup assistance to get you up and running with Claude Desktop integration.
Business Value
Why This Matters
Multiple Access Methods: We're not locked into a single interface. Our platform can be accessed via:
This web interface (zero setup)
Claude Desktop integration (natural language)
Direct API calls (for custom integrations)
This architectural flexibility means you can integrate our 12-agent analysis system into
your existing workflows - whether that's chat, custom dashboards, CRM systems, or mobile apps.
The analysis power stays the same; only the interface changes.
About This Prototype
The Basketball AI Analysis Platform is a cutting-edge demonstration of multi-agent artificial intelligence systems applied to sports analytics. This prototype showcases how specialized AI agents can work together to produce deep, nuanced analysis.
Core Capabilities:
Parasocial Analysis (6 Agents) - Examines the psychological and social relationships between fans and basketball teams across 5 dimensions
NIL Valuation (12 Agents + VALORE Framework) - Evaluates athletes' Name, Image, and Likeness market potential through comprehensive multi-dimensional analysis enhanced with behavioral science scoring and goal-oriented coordination
Powered by AWS Bedrock
This platform leverages AWS Bedrock, Amazon's managed service for foundation models, to access state-of-the-art large language models:
Claude Opus 4.6 - The Opus model available in this VVG AWS account, powering both the 6-agent parasocial analysis system and the 12-agent NIL valuation system with industry-leading coordination capabilities
Technical Innovation
Rather than relying on a single AI model, this prototype demonstrates multi-agent orchestration, where specialized agents collaborate:
Data Gathering Agents - Collect and organize information from multiple sources
Analysis Agents - Apply domain expertise to evaluate specific dimensions
Synthesis Agents - Combine insights into coherent, comprehensive reports
Orchestrator Agents - Coordinate workflows and manage agent interactions
Why Multi-Agent Systems?
Multi-agent systems offer significant advantages over single-model approaches:
Specialization - Each agent is optimized for a specific task
Scalability - Agents can work in parallel for faster results
Modularity - Individual agents can be upgraded independently
Note: This is a research prototype designed to explore the potential of multi-agent AI systems in sports analytics. Results are generated by AI models and should be interpreted as experimental demonstrations of the technology.
Parasocial Analysis
Parasocial analysis examines the one-way relationships that fans form with sports teams, exploring the psychological and sociological dynamics that create deep emotional connections despite the lack of direct personal interaction.
Theoretical Foundation
This analysis is grounded in established academic frameworks:
Parasocial Interaction Theory (Horton & Wohl, 1956) - How media creates the illusion of intimate relationships
Social Identity Theory (Tajfel & Turner, 1979) - How group membership shapes self-concept
Uses and Gratifications Theory - Why people seek out specific media experiences
Emotional Contagion - How emotions spread through communities
Place Attachment Theory - The role of physical spaces (arenas) in identity
Five Analytical Dimensions
1⃣ Emotional Connection
Attachment triggers, rituals, emotional investment patterns, and future-oriented affect
2⃣ Player Personas
Hero narratives, character roles, media framing, and relatable vs. aspirational figures
3⃣ Fan Identity
Tribal markers, symbols, collective identity formation, and in-group dynamics
4⃣ Media Narratives
Storytelling patterns, framing effects, legacy themes, and conflict resolution arcs
5⃣ Engagement Patterns
Social media behavior, ritual participation, two-way interaction, and content creation
6-Agent Parallel Architecture
Our parasocial analysis employs a parallel 6-agent system:
Phase 1: Data Agent (4-6 seconds)
Gathers and organizes web data, news articles, social media content, and fan sentiment into structured categories for analysis.
Player Personas Agent - Hero narratives, character roles, media framing
Fan Identity Agent - Tribal markers, community bonding, in-group dynamics
Media Narratives Agent - Storylines, framing effects, legacy themes
Engagement Patterns Agent - Social media, rituals, two-way interaction
Phase 3: Synthesis Agent (6-10 seconds)
Combines all five dimension analyses into a cohesive, well-formatted report with executive summary, structured sections, tables, and key takeaways.
Expected Output
A comprehensive analysis report (~5,000-8,000 words) that includes:
Executive summary of key findings
Detailed analysis of each dimension with data citations
Tables summarizing parasocial mechanisms
Theoretical grounding for all observations
Actionable insights about fan psychology
Typical Duration: A few minutes | Model: Claude Opus 4.6 | Architecture: 6 agents (1 + 5 parallel + 1)
NIL Analysis
NIL (Name, Image, and Likeness) analysis evaluates an athlete's market potential in the new era of college sports where student-athletes can profit from their personal brand. This analysis uses the VALORE Framework (Valuation Agent-Led Operations for Recruitment Economics), integrating behavioral science with multi-agent AI.
Why Current NIL Valuations Fail
Traditional NIL valuation methods demonstrate fundamental limitations:
Surface-Level Metrics - Focus on follower counts and engagement rates, which correlate poorly with actual purchasing behavior
Missing Behavioral Factors - Overlook psychological relationships (parasocial bonds) and social identity benefits that drive real marketing value
Single-Model Limitations - Monolithic AI systems attempting multiple domains achieve mediocre performance due to conflicting optimization requirements
No Authenticity Assessment - Unable to distinguish genuine fan connection from manufactured engagement
The VALORE Framework: Behavioral Science Integration
Our system addresses these limitations by integrating three quantifiable behavioral science metrics:
Parasocial Relationship Strength (PSR) - Measures emotional intensity, interaction frequency, content resonance, loyalty indicators, and temporal consistency of fan-athlete bonds
Authenticity Score - Evaluates cross-platform consistency, value alignment, behavioral congruence, communication style, and temporal stability to distinguish genuine personality from manufactured personas
Social Identity Value (SIV) - Quantifies self-esteem enhancement, social connection facilitation, and aspirational motivation that fans derive from athlete association
These metrics predict actual marketing effectiveness, sustained engagement, and partnership success far better than surface-level follower counts.
The NIL Landscape
Since July 2021, NCAA athletes can monetize their name, image, and likeness through:
Endorsement Deals - Brand partnerships and sponsorships
Social Media - Sponsored content and influencer marketing
Our NIL valuation uses a sophisticated 12-agent parallel system for comprehensive, multi-dimensional analysis enhanced with behavioral science and goal-oriented coordination:
Consistency - Same inputs → similar calculations → predictable outputs
Explainability - See exactly how numbers were derived
Auditability - Verify mathematical logic
Constraint Boundaries - Values stay within realistic ranges
Enhanced Explainability
Every valuation includes:
Top 3 "WHY" - Drivers of Valuation:
Primary Driver - The main factor determining value (quantified)
Secondary Driver - Supporting factor with measured impact
Tertiary Factor - Additional contributor with demonstrated influence
Top 3 "WHAT WOULD HAVE TO CHANGE" - Sensitivity Analysis:
Critical Assumption - Key variable that would materially alter valuation if changed
Performance Dependency - Athletic factors with quantified risk impact
Market/External Factor - External conditions affecting value stability
This approach transforms valuation from a "black box" to a transparent, falsifiable financial model.
NIL Valuation Metrics
Our analysis produces quantified assessments across key areas:
Market Value Estimation - Current and projected NIL value with formula-based calculations
Brand Strength Score - Overall brand equity and recognition
Social Influence Metrics - Reach, engagement rate, audience quality with platform multipliers
Risk Assessment - Potential reputation and performance risks
Growth Opportunities - Identified areas for brand expansion with temporal projections
Expected Output
A comprehensive NIL valuation report that includes:
Executive summary with overall NIL assessment
Financial analysis with formula-based calculations shown step-by-step
Top 3 "Why" explanations - Quantified drivers of the valuation
Top 3 "What Would Change" sensitivity analysis - Key assumptions and their impact
Brand strategy recommendations with tier positioning
Social influence metrics with platform-specific breakdowns
Risk assessment and mitigation strategies
Opportunity analysis with temporal projections
VALORE Framework behavioral scores (PSR, Authenticity, SIV)
Typical Duration: A few minutes | Model: Claude Opus 4.6 | Architecture: 12 agents (11 + coordination) in parallel
Medium Article Callout
Published Research
"Why Current NIL Valuations Fail — and How Multi-Agent AI Fixes Them"
Introducing VALORE: Where specialized AI agents negotiate like experts to value athlete influence using behavioral science
Multi-agent AI systems represent a paradigm shift from monolithic single-model approaches to collaborative networks of specialized AI agents working together to solve complex problems.
POP FRAMEWORK: Picture the Outcome First
POP Framework: Picture the Outcome First
Picture: What We're Building
A 12-agent NIL valuation system that produces coordinated, conflict-free recommendations rather than "it depends" answers. When 5 specialized analysis agents have different perspectives (e.g., Financial Analysis suggests $800K, Risk Assessment says $600K), a Coordination Agent intelligently resolves conflicts using confidence assessment and dynamic weighting to deliver one unified recommendation stakeholders can act on.
Obstacles: The Multi-Agent Challenge
Problem: Traditional multi-agent systems produce conflicting outputs. Each specialized agent has deep expertise but narrow perspective. Financial agents are bullish, risk agents are conservative, brand agents focus on intangibles. Result? Five different valuations with no clear answer—just like having 5 consultants who can't agree.
Path: Our Solution
Goal-Oriented Coordination: We added a 12th agent that acts as an intelligent mediator. It detects conflicts ("Financial says $800K but Risk says $600K"), assesses each agent's confidence in their data, applies dynamic weighting formulas, and synthesizes a coordinated recommendation with transparent reasoning. Users see HOW agents collaborated, not just the final answer.
Current State: Complex NIL valuations require multiple specialized perspectives—financial analysis, brand strategy, social influence, risk assessment, and opportunity analysis. No single LLM or agent can master all dimensions simultaneously.
Problem
The Conflict Challenge: Specialized agents produce different valuations. Financial analysis: $800K (based on metrics). Risk assessment: $600K (accounting for uncertainty). Brand strategy: $750K (market positioning). Without coordination, you get 5 answers instead of 1 actionable recommendation.
Implication
Business Impact: Stakeholders can't make decisions with conflicting recommendations. "It depends" isn't an answer for a recruit choosing between Duke and Kansas, or an athletic director allocating NIL budget. Conflicting agent outputs create analysis paralysis and erode trust in AI systems.
Need-Payoff
Our Solution: The Coordination Agent (agent #12) intelligently resolves conflicts by assessing confidence, detecting contradictions, applying dynamic weighting, and producing one coordinated recommendation with transparent reasoning. Result: Actionable insights stakeholders can trust and act on.
What is Goal-Oriented Coordination?
The Coordination Agent: Intelligent Conflict Resolution
After 5 specialized NIL analysis agents complete their work, the Coordination Agent performs 4 critical functions:
Conflict Detection: Identifies disagreements between agents (e.g., valuation ranges that don't overlap, contradictory risk assessments)
Confidence Assessment: Evaluates each agent's confidence based on data quality, specificity of findings, and clarity of reasoning (0.0-1.0 scale)
Dynamic Weighting: Applies mathematical formulas to weight agent outputs based on confidence scores, data quality, and relevance to the specific athlete context
Coordinated Synthesis: Produces one unified recommendation with explicit reasoning showing HOW conflicts were resolved and WHY certain agents had more influence
Result: Instead of 5 different perspectives, users get one coordinated answer with full transparency into the decision-making process.
Agent Types & Roles
Orchestrator Agents
Role: Coordinate workflows, manage inter-agent communication, synthesize final outputs
Example: NILOrchestratorAgent manages the 12-agent workflow
Search/Data Agents
Role: Gather information, execute searches, collect data from multiple sources
Example: CoordinationAgent resolves disagreements between 5 NIL analysis agents
Behavioral Science Agent
Role: Calculate quantifiable psychological and social metrics
Example: VALORE Framework agent (Parasocial Strength, Authenticity, Social Identity Value)
Synthesis Agents
Role: Combine information, create coherent narratives, format final reports
Example: ComparisonSynthesisAgent for school comparisons
Execution Patterns with Goal-Oriented Coordination
Hybrid Execution (Parasocial - 6 Agents)
Data Agent → 5 Dimension Agents ‖ → Synthesis Agent
Data gathering is sequential, then 5 specialized dimension agents run in parallel, followed by sequential synthesis. Best for workflows with a parallel middle phase.
Advantage: Faster than pure sequential (a few minutes), specialized analysis Trade-off: Moderate complexity
Full Parallel + Coordination (NIL - 12 Agents) ADVANCED
Coordination Phase: After analysis agents complete, the Coordination Agent detects conflicts (e.g., Financial says $800K, Risk says $600K), assesses each agent's confidence (0.0-1.0), applies dynamic weighting formulas, and produces one coordinated recommendation showing HOW conflicts were resolved.
Advantage: Fast completion (a few minutes), comprehensive coverage, conflict-free coordinated recommendations Trade-off: Highest complexity, most resource intensive, but delivers actionable unified insights
Model Selection & Deployment
This platform uses Claude Opus 4.6 for all agents:
Agent Optimization - Explicitly designed for multi-step, real-world agent applications
Long-Horizon Tasks - Built for complex, multi-stage workflows with peak accuracy
200K Context - Larger context window enables comprehensive analysis across all agents
Superior Coordination - Industry-leading capabilities for multi-agent orchestration
Using a single high-performing model across all agents simplifies architecture while maintaining excellent results.
Deployment Architecture: The VVG deployment uses HTTPS hosting and background analysis jobs. The page checks progress while private AWS Lambda workers run Claude Opus through Amazon Bedrock.
Why Multi-Agent Systems with Coordination?
Specialization
Each agent optimized for a specific task with tailored prompts and parameters—no single LLM can master all dimensions
Coordination NEW!
Intelligent conflict resolution produces one unified recommendation instead of conflicting perspectives
Depth & Breadth
Multiple specialized perspectives yield richer, more nuanced analysis with transparent reasoning
Scalability
Parallel execution enables faster processing and better resource utilization (a few minutes)
Modularity
Individual agents can be upgraded, replaced, or added without rebuilding the entire system
Transparency
Users see HOW agents collaborated, WHY conflicts occurred, and HOW they were resolved
Full System Architecture
The Basketball AI Analysis Platform now includes 3 major multi-agent systems:
System
Agents
Pattern
Duration
Parasocial
6 agents
Hybrid (1 → 5‖ → 1)
25-30s
NIL
12 agents (11 + coordination)
Parallel (5‖ → 5‖ → 1 behavioral)
20-25s
School Market
4 agents (1 orchestrator + 3 search)
Hybrid (static → 3‖ → synthesis)
5-10s
Comparison
1 synthesis agent
Strategic synthesis
5-8s
School Market Multi-Agent System
SchoolMarketOrchestrator
Role: Coordinates hybrid data gathering (static + AI)
Phase 0: Check cache, load static lookups
Phase 1: Deploy 3 search agents in parallel (if gaps)
What it doesn't do: Generic advice like "consider your preferences" or "it depends on your goals"
The Path: Architecture Evolution
Our journey from simple to sophisticated multi-agent coordination:
v1.0 - Sequential Processing (40-50s): 3 agents running one after another—slow but simple
v2.0 - Hybrid Parallel (25-30s): Upgraded to 6-agent hybrid (1 → 5‖ → 1), cutting time nearly in half
v3.0 - VALORE Framework: Added Behavioral Science Agent for quantifiable psychological metrics (Parasocial Strength, Authenticity, Social Identity Value)
v4.0 - Goal-Oriented Coordination (NOW): Added 12th agent for intelligent conflict resolution—no more "5 different answers, you decide"
Infrastructure: VVG HTTPS hosting with asynchronous jobs keeps long-running Opus analyses responsive
Result: A production-ready 12-agent system that delivers coordinated, actionable recommendations stakeholders can trust and act on.
Key Insight: Multi-agent systems with goal-oriented coordination represent the future of AI applications. Specialization without coordination creates conflicting outputs. Coordination without specialization lacks depth. Together, they enable actionable insights that single-model approaches cannot match. This prototype demonstrates practical implementation patterns that can be adapted to virtually any domain requiring complex, multi-dimensional analysis with unified recommendations.
The School Market Factors Database is a structured system that quantifies how different schools' characteristics affect NIL valuations. Instead of relying purely on LLM intuition, we anchor valuations to measurable, transparent factors with explicit multipliers.
The Core Question
"What would Cooper Flagg's NIL value be at Duke vs. Rutgers vs. Kansas?"
Different schools offer vastly different market advantages. A 5-star recruit at Duke (major media market, historic program, strong NIL collective) has access to opportunities that simply don't exist at smaller programs.
For schools not in the database, our SchoolDataAgent dynamically gathers market data using static lookup files (DMA rankings, conference data, Fortune 500 HQs) + LLM estimation for missing fields.
Confidence Score: 95%+ for cached schools, 60-85% for dynamic lookups
The Six Market Factors
Each school's composite multiplier is calculated from six weighted factors:
Factor
Weight
What It Measures
Example
Media Market
20%
Nielsen DMA ranking, TV market size
Rutgers (NYC, DMA #1) = 1.5x
Program Prestige
25%
Championships, Final Fours, tournament history (20yr)
Elite program prestige (1.3x), huge fan base (1.4x), strong NIL collective (1.3x)
Rutgers
1.13x
NYC media market (1.5x), local sponsors (1.5x), but weaker program history (0.8x)
Kansas
1.12x
Championship pedigree (1.3x), passionate fan base (1.3x), but smaller market (0.6x)
Gonzaga
0.87x
Small media market (0.4x), smaller sponsors (0.6x), mid-major conference (0.9x)
Insight: The archived Duke factor is approximately 39% higher than Gonzaga’s (1.21 ÷ 0.87 ≈ 1.39). This illustrates the prototype’s assumptions; it is not evidence of a 39% difference in athlete earnings.
Archived School Data
This deployment uses eight school scenarios retained from the 2025 prototype. The factors are illustrative assumptions, not independently verified current values. Schools outside this set receive a qualitative report without a numeric school factor.
A scenario factor is not a measured probability or a forecast of earnings. Current sponsorship terms, audience metrics, program conditions and athlete fit need separate verification.
Use Cases
Transfer Portal Decisions
"Where would my NIL value be highest if I transfer?"
Sponsorship Analysis
"Which schools offer the best local sponsor connections for my brand?"
Agent/Advisor Tools
"Compare market opportunities across multiple school offers"
Recruitment Strategy
"Quantify the NIL market advantage of our program vs. competitors"
ComparisonSynthesisAgent (NEW!)
When comparing two schools (e.g., "Cooper Flagg at Duke vs. Kansas"), a specialized ComparisonSynthesisAgent provides strategic, actionable recommendations—not just "School A is higher than School B."
What It Does:
Identifies PRIMARY DRIVERS: "Duke's 8% advantage is driven by its 1.4x fan base (vs Kansas 1.3x), NOT media market (both 1.0x)"
Quantifies Impact: "$150K-200K annual difference; Duke's fan base = 300K more Instagram followers = $60K-100K in influencer deals"
Analyzes Trade-Offs: "At Duke, you GAIN $150K in NIL but SACRIFICE starting minutes (7 returning guards)"
Provides Conditional Recommendations: "If starting guard → Duke (fan engagement). If bench player → Kansas (lower cost of living)"
"Duke offers $150K-200K more annually, driven PRIMARILY by its 1.4x fan base advantage (vs Kansas 1.3x). For a starting point guard, this translates to 300K+ more Instagram followers worth $60K-100K in influencer deals. HOWEVER, Duke's higher cost of living ($1,800 rent vs $900) and roster competition (7 returning guards) reduce the net advantage to $100K-125K. Choose Duke if you're a projected starter prioritizing maximum earnings. Choose Kansas if you're a role player prioritizing playing time and lower living costs."
What It DOESN'T Do: Generic advice like "consider your personal preferences" or "it depends on your goals." Every recommendation is quantified, specific, and actionable.
Key Insight: By anchoring NIL valuations to explicit, measurable school characteristics AND providing intelligent strategic synthesis, we move beyond subjective guesses to data-driven market analysis with actionable recommendations. Athletes, agents, and schools can make more informed decisions based on transparent, auditable factors.
Future Development: The database will expand to all 50 Power 5 schools, then to strong mid-majors. As more schools are analyzed, the system self-improves through caching. Dynamic lookups enable instant analysis of 350+ D1 schools today.
