🎯 Sports Betting Claude
A Claude Project skill system that turns "what bet should I make today?" into a structured, data-driven answer.
What this is
A collection of skill files designed to be loaded into a Claude Project. Once set up, you can ask Claude a single question:
"What bet should I make today?"
And get a structured answer: sport, market, line, book, stake size in units, EV%, confidence tier, and reasoning — every time.
The system is built around five layers that mirror how professional bettors think:
Data Ingestion → Edge Detection → Bankroll Sizing → Sport Context → Output
⚡ Quick start (5 minutes)
1. Clone the repo
git clone https://github.com/YOUR_USERNAME/sports-betting-claude.git
cd sports-betting-claude
2. Create a Claude Project
-
Go to claude.ai → click Projects in the sidebar
-
Click New project
-
Name it something like
Sports Betting System
3. Add the skill files to your Project
In your Claude Project, open Project instructions and paste the contents of PROJECT_INSTRUCTIONS.md.
Then upload these files to your Project's knowledge base (drag and drop or click "Add content"):
Start with these core files:
-
skills/data-ingestion/SKILL.md -
skills/edge-detection/SKILL.md -
skills/bankroll-management/SKILL.md -
skills/performance-tracking/SKILL.md -
skills/shared/anti-bias-checklist.md -
skills/shared/output-format.md
Then add the sports you bet on:
-
skills/sport-specific/nfl/SKILL.md -
skills/sport-specific/nba/SKILL.md -
skills/sport-specific/mlb/SKILL.md -
skills/sport-specific/nhl/SKILL.md -
skills/sport-specific/ncaa/SKILL.md
4. Set your bankroll in the Project instructions
Edit the PROJECT_INSTRUCTIONS.md you pasted and fill in:
BANKROLL: $[your starting bankroll]
UNIT SIZE: $[bankroll × 0.02] (we recommend 2% per unit)
BOOKS AVAILABLE: [list the sportsbooks you have accounts with]
5. Ask your first question
Start a new conversation in the Project and type:
"What bet should I make today?"
📁 Repo structure
sports-betting-claude/
├── README.md ← you are here
├── PROJECT_INSTRUCTIONS.md ← paste this into Claude Project settings
├── LICENSE
├── CONTRIBUTING.md
│
├── skills/
│ ├── data-ingestion/
│ │ └── SKILL.md ← how Claude fetches lines, injuries, news
│ │
│ ├── edge-detection/
│ │ └── SKILL.md ← EV calculation, vig removal, edge threshold
│ │
│ ├── bankroll-management/
│ │ └── SKILL.md ← Kelly criterion, unit sizing, stop-loss rules
│ │
│ ├── sport-specific/
│ │ ├── nfl/SKILL.md
│ │ ├── nba/SKILL.md
│ │ ├── mlb/SKILL.md
│ │ ├── nhl/SKILL.md
│ │ └── ncaa/SKILL.md
│ │
│ ├── performance-tracking/
│ │ └── SKILL.md ← bet log format, ROI tracking, calibration
│ │
│ └── shared/
│ ├── anti-bias-checklist.md ← cognitive bias detection
│ └── output-format.md ← standardized bet slip format
│
├── docs/
│ ├── how-edge-detection-works.md
│ ├── kelly-criterion-explained.md
│ ├── reading-line-movement.md
│ └── glossary.md
│
└── examples/
├── sample-daily-output.md ← example of what Claude returns
├── sample-bet-log.md
└── sample-bankroll-tracker.md
📊 What a daily output looks like
📅 DATE: [Today]
🏦 BANKROLL: $2,450 | UNIT: $49
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔵 BET #1 — HIGH CONFIDENCE (4 units)
Sport: NBA
Market: Spread
Pick: Celtics -4.5
Book: DraftKings (best available)
Line: -110
Stake: $196
EV: +4.2%
Reasoning: Sharp reverse line movement from -5.5 to -4.5 despite 68% public on
Boston. Tatum O/U minutes reduced, suggesting precautionary rest
management — load still expected. Closing line target: -5.
🟡 BET #2 — MEDIUM CONFIDENCE (2 units)
Sport: MLB
Market: Total (Over)
Pick: Cubs/Cardinals O8.5
Book: FanDuel
Line: -115
Stake: $98
EV: +2.1%
Reasoning: Wrigley day game, wind blowing out 12mph to left-center. Both starters
have struggled in day games (ERA 0.8 pts higher). Park factor elevated.
⛔ PASS: 4 other games reviewed, no qualifying edge found today.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📋 LOG THESE: [format for your bet log]
🔬 How the edge detection works
Every bet goes through this filter before Claude will recommend it:
1. Find the true line (best available across books, no vig)
2. Calculate implied probability from your model
3. Compare: if your P(win) > market P(win) + 2% threshold → edge exists
4. Apply Kelly: stake = (edge × bankroll) / odds
5. Reality-check against sport-specific context (Layer 4)
6. Run anti-bias checklist (Layer 5)
7. Output or discard
No edge = no bet. The system will explicitly tell you when to pass.
⚠️ Honest expectations
What this system does What it doesn't do
Structures your research process Guarantee winners
Removes emotional bias Replace your own judgment
Finds positive EV opportunities Beat every line every day
Tracks your real edge over time Work without your input on bankroll/books
Teaches you to think like a pro Access live odds APIs automatically
20% monthly bankroll growth is achievable in winning stretches but not guaranteed month-over-month. Professional bettors average 5–10% ROI over a season. This system gives you the framework — variance is still part of sports betting.
🛠️ Contributing
See CONTRIBUTING.md. PRs welcome for:
-
New sport-specific skill files (soccer, tennis, MMA, golf)
-
Better data sources
-
Improved bias detection heuristics
-
Community-tested prompt improvements
📜 License
MIT — use it, fork it, improve it. See LICENSE.
🙋 FAQ
Do I need Claude Pro? Yes. Claude Projects require a Pro or Team subscription.
Does this place bets automatically? No. Claude gives you the recommendation; you place the bet manually. This is intentional — always review before acting.
What books does this work with? Any book you have access to. The system compares lines across whatever books you list in your Project settings.
How do I track my bets?
Use the examples/sample-bet-log.md as a template and paste your results back into the chat so Claude can update its performance model.
Can I add my own data sources? Yes — the data ingestion skill file lists the search patterns Claude uses. Fork and extend them.