Predict9ja

Verifiable football prediction settlement powered by TxLINE.

Predict the match. Replay the action. Verify the result.

Live application · Demo video · Architecture · TxLINE integration · TxLINE API feedback · Operations

Overview

Predict9ja is a mobile-first football prediction platform that turns TxLINE fixture and score data into transparent demo-credit predictions, deterministic settlement and auditable receipts.

Its flagship verified replay lets a user make a pick, replay England versus Argentina from stored TxLINE observations, and inspect the source evidence used to confirm the final result. The product separates sports-data provenance, application resolution rules and account settlement.

Predict9ja is a demonstration environment. Credits have no cash value, and the application does not accept deposits, withdrawals or real-money wagers.

The problem

Most prediction interfaces show an outcome without exposing the path from source data to resolution. Users cannot easily inspect which match update confirmed the score, which rule resolved the prediction or whether settlement was applied consistently.

Predict9ja makes that path inspectable.

What Predict9ja does

• Synchronizes a football fixture catalogue from TxLINE devnet.
• Presents match-result, total-goals and both-teams-to-score demo predictions.
• Creates isolated anonymous sessions with 10,000 demo credits.
• Replays stored historical observations in provider-sequence order.
• Resolves markets with versioned deterministic rules.
• Settles an append-only demo-credit ledger idempotently.
• Generates receipts with canonical integrity digests.
• Exposes TxLINE source evidence for the authoritative final observation.
• Validates exact source predicates against TxLINE's Solana devnet program through a read-only path.
• Provides a clearly labelled synthetic fallback for provider-independent demonstrations.

Application quotes are fictional demonstration prices and are not TxLINE consensus odds.

Business and technical highlights

Business highlights

Predict9ja demonstrates a reusable foundation for white-label prediction experiences, sponsored fan campaigns, community leagues, developer-facing settlement APIs and audit tooling for sports-data products. The current build deliberately limits itself to fictional credits while proving the product mechanics, evidence model and partner-facing architecture.

Technical highlights

• Provider-sequence-aware ingestion and historical replay.
• Explicit finalisation authority rather than newest-record assumptions.
• Versioned, framework-independent market rules.
• Integer-only balances with an append-only ledger.
• Idempotent purchases, payouts and refunds.
• Canonical receipt digests and bounded proof persistence.
• Server-only TxLINE credentials and in-memory guest JWTs.
• Read-only Solana verification of exact score predicates.
• Scheduled fixture synchronization with checkpointed SSE support available to a continuously running worker.

Architecture

flowchart LR
    User[Fan or analyst] --> Web[Next.js web application]
    Web --> Services[Prediction and replay services]
    Services --> DB[(PostgreSQL via Prisma Accelerate)]

    Workflow[GitHub Actions fixture sync] --> Worker[Operator worker]
    Worker --> TxLINE[TxLINE API]
    TxLINE --> Worker
    Worker --> DB
    Worker --> Verification[Proof verification package]
    Verification --> Solana[TxLINE Solana devnet program]

    DB --> Web

Loading

This is the deployment and data-flow view. The detailed component map expands the same system into its monorepo packages and internal dependency boundaries; the diagrams intentionally use different levels of abstraction rather than representing different architectures.

The web application presents fixtures, replay controls, predictions and receipts. Worker commands own provider ingestion, historical imports, proof retrieval, verification and operator workflows. Framework-independent market rules live in the domain package, while PostgreSQL is the application system of record.

Verified replay and settlement flow

flowchart LR
    Fixture[TxLINE fixture] --> Observations[Normalized match observations]
    Observations --> Replay[Isolated historical replay]
    Pick[User pick and demo stake] --> Rules[Versioned prediction rules]
    Replay --> Rules
    Rules --> Ledger[Idempotent ledger settlement]
    Ledger --> Receipt[Prediction receipt]
    Observations --> Proof[TxLINE proof material]
    Proof --> Solana[Read-only Solana validation]
    Solana -. Verified source evidence .-> Receipt

Loading

The source proof validates the canonical match observation. Predict9ja's own rules and ledger produce the fictional application settlement. The proof does not validate the demo payout itself.

For the featured replay, sequence 962 is authoritative because it is the explicit final game_finalised observation with finalised=true. A later sequence, 963, is retained for provenance but is non-final and cannot replace the authoritative final evidence.

TxLINE integration

Endpoint	Purpose
POST /auth/guest/start	Obtain an in-memory guest JWT
GET /api/fixtures/snapshot	Synchronize the devnet fixture catalogue
GET /api/scores/snapshot/{fixtureId}	Fetch controlled score snapshots
GET /api/scores/historical/{fixtureId}	Import ordered historical observations
GET /api/scores/stream	Consume score events through SSE
GET /api/scores/stat-validation	Retrieve proof material for exact score predicates

SSE support is implemented but is not continuously hosted on Vercel. Historical replay uses observations already stored by the application. Provider credentials remain server-side. TxLINE odds are not currently consumed.

TxLINE API feedback

What worked well

TxLINE's stable fixture identity, participant orientation, ordered score observations, historical access and stat-validation proof material made deterministic persistence, replay and source verification practical.

Where we encountered friction

The integration required careful handling of bounded historical availability, legitimate UNKNOWN phases, and later non-final observations that can follow an explicit game_finalised record. SSE ingestion also requires a continuously running worker rather than a request-driven Vercel deployment.

The complete experience report and improvement suggestions are documented in TxLINE API feedback.

Integrity model

1. TxLINE supplies fixture identity, participant orientation, ordered score observations and proof material.
2. Explicit game_finalised data establishes the authoritative final score.
3. Versioned application rules resolve each demonstration market.
4. Settlement writes a single ledger effect for each position.
5. A receipt records the resolution and settlement context with a canonical integrity digest.
6. Solana validation verifies the source observation predicates, not the fictional account payout.

Technology

Next.js, React, TypeScript, Node.js 22, Turborepo, pnpm, PostgreSQL, Prisma, Prisma Accelerate, TxLINE HTTP/SSE APIs, Solana devnet, Vitest, Vercel and GitHub Actions.

Repository structure

Path	Responsibility
apps/web	Public Next.js application and API routes
apps/worker	TxLINE ingestion, replay and operator commands
packages/domain	Framework-independent prediction and resolution rules
packages/txline	TxLINE transport, authentication and normalization
packages/verification	Proof normalization, digests and Solana validation
packages/db	Prisma models, persistence, settlement and receipts
docs	Architecture, integration, feedback and operations docs
scripts	Development, integration-test and runtime-smoke tooling

Getting started

Prerequisites

• Node.js 22
• pnpm 10.12.1
• Docker
• PostgreSQL 16 through the included Compose configuration

corepack enable
pnpm install --frozen-lockfile
cp .env.example .env
pnpm db:up
pnpm db:deploy
pnpm db:seed
pnpm dev

The repository documents environment-variable names in .env.example. Keep local credentials outside source control.

Common commands

pnpm dev
pnpm dev:all
pnpm db:up
pnpm db:deploy
pnpm db:seed
pnpm txline:sync-fixtures
pnpm txline:import-history --fixture-id <id>
pnpm txline:verify-proof --fixture-id <id> --sequence <seq> --stat-keys 1,2
pnpm markets:generate-all
pnpm markets:resolve --fixture-id <id> --dry-run
pnpm markets:settle --fixture-id <id>

Testing

pnpm format:check
pnpm lint
pnpm typecheck
pnpm test
pnpm build
pnpm check:web-runtime

The test strategy combines unit tests, isolated PostgreSQL integration tests, production builds and route-level runtime smoke checks. CI runs the complete verification sequence on every push and pull request.

Deployment

The public web application is deployed on Vercel. Runtime database traffic uses Prisma Accelerate, while migrations and seeds remain explicit administrative operations. A separate GitHub Actions workflow synchronizes the TxLINE fixture catalogue every 30 minutes and supports manual dispatch.

Current scope and limitations

• Demo credits have no cash value and cannot be deposited, withdrawn, transferred or redeemed.
• There is no custody, escrow or real-value settlement.
• Fixed quotes are application demonstration prices, not TxLINE odds or price discovery.
• Catalogue coverage follows the current TxLINE devnet snapshot.
• The featured replay is historical and is not presented as a currently live match.
• SSE ingestion is implemented but is not continuously hosted on Vercel.
• TxLINE proof validation establishes source-data predicates; it does not prove the application payout.
• On-chain escrow and automated real-value settlement are outside this build.

Documentation

• Architecture
• TxLINE integration
• TxLINE API feedback
• Operations

Acknowledgements

Predict9ja was built with TxLINE by TxODDS and its Solana-verifiable sports-data infrastructure. The project also acknowledges the Solana developer ecosystem and the TxODDS World Cup hackathon hosted on Superteam Earn.