EdgeLine: Autonomous Sports Trading Agent (Solana)

EdgeLine is an autonomous sports trading agent that detects and executes profitable live trades on Solana devnet with zero manual intervention. The agent continuously monitors in-play football matches, computes a mathematically defensible "fair-price" probability model using live match state, compares its outputs to the real-time TxLINE StablePrice consensus odds to find structural value (divergence edges), and automatically executes position entries and settlements directly on-chain using an Anchor Rust program.

---

Architecture Pipeline

┌─────────────────┐      ┌─────────────────────────┐      ┌───────────────────────────┐
│   TxLINE API    ├─────►│  Fair-Price Model (API) ├─────►│ Divergence Scanner (API)  │
│ In-Play Ingestion│      │  (Bayesian update prior)│      │  (Divergence > 8% Threshold)│
└─────────────────┘      └─────────────────────────┘      └─────────────┬─────────────┘
                                                                        │
┌─────────────────┐      ┌─────────────────────────┐      ┌─────────────▼─────────────┐
│  Live Dashboard ◄──────┼──  WebSockets Stream    │      │  Decision Engine (API)    │
│ (React/Next.js) │      │  (Socket.IO broadcast)  │      │  (Risk check & position)  │
└─────────────────┘      └─────────────────────────┘      └─────────────┬─────────────┘
                                                                        │
                                                                        ▼
                                                          ┌───────────────────────────┐
                                                          │   Solana Devnet Program   │
                                                          │  (record/settle position) │
                                                          └───────────────────────────┘

1. In-Play Ingestion: Sourced via HTTP polling of TxLINE live odds and scoring feeds.
2. Fair-Price Model: Treats pre-match consensus odds as a Bayesian prior and updates it dynamically based on goal differences and net red cards, scaled by match duration elapsed.
3. Divergence Scanner: Compares model outcomes to live consensus odds to flag opportunities exceeding the threshold (e.g. 8%+ edge).
4. Decision Engine: Performs sizing via a risk cap system (daily limit, per-fixture limit) and approves entry.
5. Solana Execution: Initiates Solana transactions using the wallet keypair to record the trade PDA, with auto-airdrop and retries.
6. Live Dashboard & Settlement: Pushes updates instantly to Next.js clients via WebSockets, periodically checks for finished matches, computes realized PnL, and settles them on-chain.

---

Mathematically Defensible Fair-Price Model

Rather than fitting a parameter-heavy predictive model, EdgeLine uses a Bayesian Prior-Update model. It establishes the initial prior from the consensus market price (TxLINE StablePrice), which incorporates broad market intelligence, and then applies a time-weighted update based on live match actions.

1. Urgency Weighting (Time Weight)

The significance of any live event (goals, dismissals) increases linearly as the match progresses: $$t = \min\left( \frac{\text{minuteElapsed}}{90}, 1 \right) \quad \in [0, 1]$$

2. Event-Driven Shift Inputs

• Goal Differential Shift: $$\Delta_{\text{goals}} = (\text{homeScore} - \text{awayScore}) \times 0.15 \times t$$
• Red Card Shift (Disadvantage): $$\Delta_{\text{red_cards}} = -(\text{homeRedCards} - \text{awayRedCards}) \times 0.12 \times t$$

The combined event shift represents the net change in win probability: $$\Delta = \Delta_{\text{goals}} + \Delta_{\text{red_cards}}$$

3. Probability Distribution Adjustments

If $\Delta \ge 0$ (Home team is favored by match events):

• $P_{\text{home}} = P_{\text{home_prior}} + \Delta$
• $P_{\text{draw}} = P_{\text{draw_prior}} - 0.5 \times \Delta$ (Draw compresses faster)
• $P_{\text{away}} = P_{\text{away_prior}} - 0.5 \times \Delta$

If $\Delta < 0$ (Away team is favored):

• $P_{\text{away}} = P_{\text{away_prior}} + |\Delta|$
• $P_{\text{draw}} = P_{\text{draw_prior}} - 0.5 \times |\Delta|$
• $P_{\text{home}} = P_{\text{home_prior}} - 0.5 \times |\Delta|$

4. Clamping and Normalization

To reflect epistemic uncertainty (no outcome is mathematically impossible), the probabilities are clamped to a hard floor of $0.01$ and a ceiling of $0.98$: $$P'i = \max(0.01, \min(0.98, P_i)) \quad \Longrightarrow \quad P{\text{final}} = \frac{P'_i}{\sum P'_j}$$

---

Setup & Local Run Instructions

1. Environment Variables (apps/api/.env)

Ensure you create an .env file inside apps/api/ with:

SOLANA_WALLET_PATH=/Users/basil/.config/solana/edgeline-devnet.json
SOLANA_RPC_URL=https://api.devnet.solana.com
TXLINE_BASE_URL=https://txline-dev.txodds.com
TXLINE_API_TOKEN=your_txline_token_here
USE_FREE_WORLDCUP_TIER=true
POLL_INTERVAL_MS=20000
SETTLEMENT_INTERVAL_MS=120000
MAX_POSITION_SIZE=0.5
MAX_DAILY_EXPOSURE=2.0
MAX_CONCURRENT_POSITIONS_PER_FIXTURE=1
API_PORT=3001
API_HOST=0.0.0.0

2. Fund the Devnet Wallet

Create your devnet wallet keypair and request funds:

# Generate keypair
solana-keygen new --outfile ~/.config/solana/edgeline-devnet.json --no-bip39-passphrase

# Set config to use this keypair on devnet
solana config set --keypair ~/.config/solana/edgeline-devnet.json --url devnet

# Airdrop devnet SOL
solana airdrop 2

3. Deploy the Anchor Program to Devnet

From the root directory:

# Build SBF program and generate IDL
anchor build

# Deploy vault program to Solana Devnet
anchor deploy --provider.cluster devnet --provider.wallet ~/.config/solana/edgeline-devnet.json

# Copy the real IDL to API
cp target/idl/edgeline_vault.json apps/api/src/execution/idl/edgeline_vault.json

4. Running the Agent & Dashboard

Install dependencies and launch services from the workspace root:

pnpm install

# Start NestJS backend API
pnpm dev:api

# Start React/Next.js dashboard (on http://localhost:3000)
pnpm dev:web

---

TxLINE API Endpoints Used

• Guest Token Request: POST /auth/guest/start (retrieves short-lived guest JWT)
• Static API Token Activation: POST /api/token/activate (binds signed on-chain tx signature to static API token)
• Fixture Ingestion: GET /api/fixtures/snapshot (retrieves full tournament list)
• Real-time Consensus Odds: GET /api/odds/snapshot/{fixtureId} (retrieves StablePrice odds)
• Real-time Match Events: GET /api/scores/snapshot/{fixtureId} (retrieves score updates, minutes, and card events)

---

Judging & Testing Deliverables

• Judge-Testable Status Endpoint: GET http://localhost:3001/agent/status
• Judge-Testable Simulation Trigger: GET http://localhost:3001/agent/test-event
  • Tip: Trigger the simulation via curl http://localhost:3001/agent/test-event to instantly see live-updating Socket.IO changes on the dashboard.
• On-Chain Program Address: EAVB3QfGZmMhRvYmtTrsLLuaRYbe6yBRM6JMj68R4VS3 (Solana Devnet)