📝 Blog
Community for Vibe Coders

Built with AI,
shipped with pride.

You built something with AI — now show it off. Get real feedback, collect upvotes, and join the builders turning ideas into working products over the time. You have a working product ? Great, find here the right customer.

95projects
75builders
423upvotes
144members
⚡ Code Check

Run a Code Check.
Get your Vibe Score.

Paste your live URL or public GitHub repo and get a score from 0–100 with clear fixes.

Run a Code Check
My builds
Under Review
Saved projects
Share your build ✦

Built something with AI? Post it and get feedback from the community.

Trending tools
Top builders
Coming soon

Code Check is coming soon

Check your live site and public GitHub repo for leaked keys, security headers, SEO and speed, with clear fixes.

Messages
No messages yet
or
⌕
Community

Forum

Discuss vibe coding, share tips, ask questions. 👇🏼

New Thread
Title
Body

@kodji7202 on CheckMyVibeCode

ScripFotbal

🚀 ScripFotbal

What it is: Statistical football analysis platform. Calibrated probability models trained on 24,693 matches across 13+ top leagues. Outputs per-match win/draw/over/BTTS probabilities, fair odds, edge vs. market, Kelly stake, and AI-reasoned narrative. Informational only — not a betting operator, not a tipster. What it is NOT: - Not "safe bets" — no guaranteed picks - Not a bookmaker — cannot place bets on site - Not hype — shows losses, drawdowns, full public backtest How it works: 1. Data ingest: API-Football feeds: fixtures, live scores, odds (10+ bookmakers), lineups, injuries, standings, xG. 2. Feature engineering (36 features, Phase1B set) - Elo ratings per team (updated match-by-match) - Rolling xG last 10 matches - Momentum deltas (form trend) - Rest days + European fatigue flags - H2H stats, home advantage, league context - Lineup intel (absences weighted by impact) 3. Probability stack - Base: XGBoost ensemble (1X2 + Over 2.5 + BTTS heads) - Poisson: Dixon-Coles bivariate for score distribution - Stacking: 5-fold walk-forward CV, isotonic regression calibration - Anchor: Pinnacle closing-line reference to reduce model bias 4. Edge computation - Vig-free fair odds from sharp books - edge = model_prob × best_odd − 1 - Kelly stake = (prob × odd − 1) / (odd − 1) (fractional 1/4 displayed) 5. AI reasoning (DeepSeek V3.2) Layer on top. Feeds model probabilities + context (derby, motivation, congestion, lineup, H2H) into structured prompt. Returns: - Match narrative (Romanian, sharp tone) - Edge validation per market (is this edge real or statistical noise?) - Trap warnings (volatility, market-already-priced scenarios) - Conviction scores 0-100 per market - Recommended vs. avoid markets - Stake modifier 0.5×-2.0× 6. Smart Pick Highest-conviction market per match. Logged automatically with timestamp + reasoning. Reconciled with actual result within 3h of final whistle via API-Football. Displayed publicly on /performanta. Transparency model - Public walk-forw

By in website

Upvotes: 5

Build time: 3 weeks

Cost: $200

Tools: Claude Code

View Project