Sixth Sense: Comprehensive Assistant Coach

🏆 Cloud9 x JetBrains Hackathon Submission

Category 1: Comprehensive Assistant Coach

Sixth Sense is a data-driven assistant coach for VALORANT that provides personalized player insights, automated macro game review, and predictive "what-if" analysis - exactly as specified in the hackathon requirements.

---

⚡ Key Features

1. Personalized Player/Team Improvement Insights

Example Output:

DATA: Team loses approximately 78% of rounds when f0rsakeN dies first (without KAST)

INSIGHT: f0rsakeN's opening duel success rate heavily impacts the team. They died first in 8 rounds, potentially costing ~6 rounds.

RECOMMENDATION: Review f0rsakeN's opening pathing and ensure trade support is always available.

2. Automated Macro Game Review

• First Blood Impact Analysis (KAST-style)
• Round-by-round breakdown with kill distribution
• Death Trap identification using ML clustering
• Weapon effectiveness analysis
• Player matchup patterns (Nemesis detection)

3. Predictive "What-If" Analysis

• "What if we avoided the main Death Trap areas?"
• "What if our most vulnerable player improved opening duels?"
• "What if we traded effectively after first blood deaths?"

Each prediction includes:

• Current state analysis with actual data
• Predicted improvement percentages
• Confidence level
• Implementation strategies

---

Unique Differentiator: Visual Map Analysis

Unlike typical analytics tools, Sixth Sense provides spatial visualization on actual map layouts:

• Auto-calibrated Map Images aligned with GRID coordinates
• ML-Powered Death Trap Detection using DBSCAN clustering
• Kill Zone Mapping to identify successful positions
• Interactive Filtering by player, map, and sensitivity

Coaches can literally SEE where problems occur on the map.

---

Quick Start

Step 1: Install Dependencies

# Create virtual environment (recommended)
python -m venv .venv
.\.venv\Scripts\Activate.ps1  # Windows PowerShell

# Install required packages
pip install -r requirements.txt

Step 2: Get Match Data (Optional - sample data included)

# Set your GRID API key
$env:GRID_API_KEY="<your_key>"

# Download a match
python grid_client.py <SERIES_ID>

Skip this step if using the included real_match.jsonl sample data.

Step 3: Process Kill Data ⚠️ REQUIRED

python process_kills.py

This parses real_match.jsonl → data/kills_data.csv

You must run this before launching the app!

Step 4: Launch the Dashboard

streamlit run app.py

Adding More Matches

python process_kills.py <input.jsonl> matches/<match_name>.csv

Example:

python process_kills.py game2.jsonl matches/VCT_Game2.csv

---

Sample Insights Generated

Critical Impact Analysis

 KAST Impact Analysis

DATA: The team loses approximately 78% of rounds when f0rsakeN dies first

FINDING: f0rsakeN died first in 8 rounds (33.3% of all rounds)

IMPACT: This potentially cost the team ~6 rounds

STRATEGIC INSIGHT: f0rsakeN's opening duel success rate heavily impacts 
team performance. When they die "for free" (without Kill/Assist/Survive/Trade), 
the team has only a ~22% chance of winning that round.

Death Pattern Analysis

 DATA: 35.2% of all deaths (127/361) occur in 3 identifiable death traps

 INSIGHT: These concentrated death zones indicate predictable positioning 
that opponents exploit.

📈 RECOMMENDATION:
1. Review the Death Map for exact locations
2. Develop alternative approach paths
3. Use utility to clear or block these areas

---

️ Project Structure

Sixth_Sense/
├── app.py                 # Main Streamlit application
├── grid_client.py         # GRID API client (streaming downloads)
├── process_kills.py       # Kill data extraction (generator-based)
├── add_match.py           # Helper to add new matches
├── utils.py               # Shared utilities & caching
├── requirements.txt       # Python dependencies
├── assets/                # Map images
│   ├── Ascent.png
│   ├── Bind.png
│   ├── Split.png
│   └── Abyss.png
└── matches/               # Additional match data

---

Technical Highlights

• Memory Efficient: Generator-based JSONL processing
• Optimized I/O: Batched CSV writing, streaming downloads
• Smart Caching: LRU cache for data, TTL cache for Streamlit
• ML Analytics: DBSCAN clustering for death pattern detection
• Real Data: All insights calculated from actual match events

---

Supported Data

• VALORANT match data from GRID API
• Kill/death events with coordinates
• Round information
• Weapon and headshot data
• First blood tracking

---

Hackathon Alignment Checklist

Requirement	Implementation
Personalized Player Insights	KAST-style analysis, Data→Insight format
Team Improvement Insights	Death patterns, matchup analysis
Automated Macro Game Review	Round analysis, first blood impact
Predict "What-If" Scenarios	5 data-backed prediction scenarios
GRID Data Integration	Download, process, visualize pipeline
Provide Data/Reasoning	Every insight shows supporting data

---

Developed for the Cloud9 x JetBrains Hackathon 2026