Contact person: Sports Prediction Website with a Bot

Phone:Show

Email:Show

Location: Nadia, India

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"Project Overview
We are building a sports prediction website where users make football (soccer) predictions and write analyses. To enrich the community, we want to introduce AI-powered personas (bots) that act like real users.

These bots will:
• Use the SportMonks API (fixtures, advanced stats, odds) to generate probabilities for standard prediction types (1X2, Over/Under, BTTS, Correct Score, etc.).
• Interpret probabilities differently based on behavioral styles (e.g. Safe Analyst, Aggressive Gambler, Trend Chaser, Underdog Hunter).
• Generate longform analyses in a target language (English, German, Turkish, etc.) explaining why they made a prediction — without ever mentioning betting odds explicitly.
• Post predictions & rationale on the site via browser automation (Playwright), so they behave exactly like a normal user.

⸻

Key Features
• Persona Engine: Configurable personas with styles (safe, balanced, aggressive, contrarian, trend-based, etc.), posting windows, languages, and voice/tone.
• Probability Engine:
• Compute model probabilities using Poisson (λ from xG), Elo, and rolling form stats.
• Pull SportMonks odds → normalize → blend with model probabilities (time-to-kickoff dependent).
• Enforce clash rules & probability floors based on persona style.
• LLM Integration (OpenAI API): Generate rationale/analysis text (strict JSON contract → safe parsing).
• UI Automation: Playwright posts predictions + rationale through the website’s user interface.
• Dashboard ([login to view URL] + Postgres):
• Manage personas (CRUD) and their configs.
• Shadow / Canary modes (draft vs live posts).
• Scheduler (posting windows, quotas, blackout hours).
• Runs Explorer: view each bot’s picks, probabilities (model/market/final), rationale, and status.
• Basic metrics: win rate, total points, points by prediction type.
• Scalable Job Queue: Redis + BullMQ workers with concurrency controls and retries.
• Health/Logs: view queue depth, worker status, and failure artifacts (Playwright [login to view URL]).

⸻

Deliverables
• Bot platform with all features above, deployed (Vercel + Neon Postgres + Redis worker node).
• Documentation: environment variables, how to add new personas/styles, how to add new prediction types.
• Codebase clean, modular, and production-ready.

⸻

Tech Stack (preferred)
• Frontend/Dashboard: [login to view URL] + Tailwind
• Backend/API: Node.js (TypeScript) + Prisma + Postgres
• Job Queue: Redis + BullMQ
• Browser Automation: Playwright
• Data: SportMonks API (paid key provided)
• LLM: OpenAI API

⸻

Candidate Requirements
• Strong background in Node.js/TypeScript and React/Next.js.
• Experience with queue systems (BullMQ, Redis, or similar).
• Solid knowledge of browser automation (Playwright/Puppeteer).
• Comfort working with sports data APIs (SportMonks a plus).
• Understanding of probability models (Poisson, Elo) is a strong bonus.
• Must deliver clean, documented, production-ready code.
• Fluent English communication.

⸻

Scope & Timeline
• Estimated scope: ~150 hours for first version.
• Timeline: ~3–4 weeks (flexible).
• Future extensions: bot learning loops, advanced analytics, multi-language support, and replay tools." (client-provided description)