ArcMarkets: A Next-Generation Predict-to-Earn Protocol on the Arc Network

Protocol Version: 1.0

Network: Arc Testnet (Chain ID: 5042002)

Status: Live Deployment

Abstract

This paper presents the technical specification of ArcMarkets, a decentralized peer-to-peer parimutuel prediction market protocol deployed on the Arc Network. ArcMarkets introduces two core architectural innovations: (1) Autonomous AI Agent Delegation, enabling users to escrow USDC on-chain and authorize a specialized off-chain agent to place mathematically optimized wagers using a fractional Kelly Criterion model; and (2) 100% On-Chain SVG NFT Receipts, which encode prediction parameters, outcomes, and visual metadata entirely in Solidity without dependence on external storage layers such as IPFS or Arweave. By leveraging Arc Network's native USDC gas model, ArcMarkets delivers a seamless, low-friction user experience suited to both institutional hedging and consumer-grade prediction markets.

1. Introduction

Traditional prediction markets and sports betting platforms are structurally disadvantaged for retail participants across three primary dimensions:

High Spread and Counterparty Risk. Bookmakers systematically skew odds to guarantee a positive margin, forcing retail participants to compete against house models that guarantee long-run extraction of value from the user base.

Onboarding and Gas Friction. Most EVM-compatible networks require participants to hold a volatile native asset — typically ETH — solely to cover transaction fees. This creates a non-trivial entry barrier and exposes users to exchange-rate risk on a non-productive asset.

Execution Complexity. Placing mathematically optimal bets requires continuous market monitoring, real-time probability modeling, and disciplined risk sizing — a combination of demands that is impractical for the average participant.

ArcMarkets addresses these inefficiencies by deploying on the Arc Network, utilizing native USDC gas, and pairing on-chain parimutuel liquidity pools with a delegated, quantitatively driven AI agent.

2. Parimutuel Pool Pricing Model

Unlike peer-to-peer order books or Constant Product Market Makers (x \cdot y = k), ArcMarkets implements a Parimutuel Pooling Engine in which all wagers on a specific event are aggregated into a single outcome-stratified liquidity pool.

2.1 Mathematical Formulation

Let S = \{1, 2, 3\} denote the discrete set of prediction outcomes for a given event:

1: Home Win
2: Draw
3: Away Win

Let L_i denote the total USDC liquidity wagered on outcome i \in S. The total pool size T is defined as:

T = Σj ∈ S Lj

Protocol Equation

Upon match resolution, a platform fee rate f (where f = 0.02, equivalent to 200 basis points) is deducted to fund protocol operations and agent incentive structures:

Tnet = T × (1 - f)

Protocol Equation

Dynamic decimal odds O_i for each outcome i are computed continuously on-chain as follows:

Oi = \begin{cases} \dfrac{T × (1 - f)}{Li} & if Li > 0 \\[10pt] Ofallback, i & if Li = 0 \end{cases}

Protocol Equation

Fallback odds apply when no capital has been wagered on a given outcome and are denominated in basis points:

Outcome	Fallback Odds
Home Win (i = 1)	2.0 × (20,000 bps)
Draw (i = 2)	3.0 × (30,000 bps)
Away Win (i = 3)	2.0 × (20,000 bps)

2.2 Payout Distribution

For a winning wager of size w placed on the winning outcome i^*, the payout P(w) is proportional to the bettor's share of the winning pool:

P(w) = (w) / (Li^*) × Tnet = w × Oi^*

Protocol Equation

In the event of match cancellation, the protocol enters capital preservation mode and all wagers are returned in full without fee deduction:

P(w) = w

Protocol Equation

3. Escrowed AI Agent Delegation

ArcMarkets employs an on-chain/off-chain hybrid delegation architecture that enables autonomous market participation without compromising user custody.

Protocol Sequence: AI Agent Delegation Flow

Interactive view showing on-chain escrow, off-chain prediction analysis, and automated betting execution.

User Wallet

Retains USDC custody & reviews NFT receipts

Smart Contract

Holds escrow pool & validates permission

AI Agent Client

Computes EV & automates fractional Kelly sizing

11. Agent Authorization & EscrowUser Wallet ➔ Smart Contract

User calls `authorizeMyAgent(agentAddress, budget)` transferring USDC directly to the PredictionMarket escrow pool. No private keys are shared.

Protocol Details: USDC is locked on-chain in escrow. The user can revoke delegation or withdraw funds at any time.

22. Fetch Market ParametersAI Agent Client ➔ Smart Contract

The autonomous off-chain client agent pulls active match profiles, kickoff times, and current parimutuel pool sizes from the contract.

33. Analyze & Compute Kelly SizeAI Agent Client ➔ AI Agent Client

Agent calculates winning probabilities using FIFA team strength ratings and calculates expected value (EV) based on pool odds.

44. Executing Delegated WagerAI Agent Client ➔ Smart Contract

If expected value exceeds the profile threshold, agent calls `agentPlaceBet(userAddress, matchIndex, outcome, amount)`.

55. On-Chain Settlement & NFT MintSmart Contract ➔ User Wallet

Smart Contract verifies agent signature, deducts bet amount from user's escrow balance, and mints an on-chain SVG NFT bet receipt to the user.

3.1 Escrow Architecture

The delegation model is structured around three user-controlled operations:

Agent Authorization. The user calls authorizeMyAgent(agent, budget) on-chain, transferring the specified USDC amount directly into the PredictionMarket escrow pool and associating that balance with a designated agent wallet.

Custody Retention. At no point does the AI agent hold the user's private keys or possess withdrawal permissions. The agent is authorized exclusively to invoke agentPlaceBet() against the user's designated escrow balance.

Instant Revocation. The user may call revokeAgent() at any time. This operation immediately resets the agent's authorization and returns the remaining escrow balance to the user's wallet.

4. The Kelly Criterion Sizing Engine

The off-chain client agent executes a quantitative prediction pipeline that evaluates team strength and determines optimal wager sizing on a per-fixture basis.

4.1 Probability Modeling

Teams are evaluated using FIFA strength ratings R \in [0, 100]. For a matchup between home team H and away team A:

Δ R = RH - RA

Protocol Equation

Historical win probabilities are estimated as:

pH = clamp[0.15, 0.85](0.50 + 0.004 × Δ R + 0.05)

Protocol Equation

pA = clamp[0.15, 0.85](0.50 - 0.004 × Δ R - 0.05)

Protocol Equation

pD = \max(0.10, 1.00 - pH - pA)

Protocol Equation

4.2 Expected Value Formulation

For each outcome i \in S, expected value EV_i is computed against the current on-chain odds O_i:

EVi = pi × Oi - 1

Protocol Equation

The agent selects the outcome with the highest expected value:

i^* = \underset{i ∈ S}{\arg\max} EVi

Protocol Equation

4.3 Bet Sizing Formulation

The raw Kelly fraction f^* defines the theoretically optimal proportion of capital to risk:

f^* = \max(0, \frac{pi^* × Oi^* - 1}{Oi^* - 1}) = \max(0, \frac{EVi^*}{Oi^* - 1})

Protocol Equation

In practice, the full Kelly formula is sensitive to estimation error in p_i and can produce excessive drawdowns. ArcMarkets applies an Outcome Multiplier m_outcome and a fixed safety scalar to derive a conservative adjusted fraction:

fadj = \max(0, EVi^*) × moutcome × 0.25

Protocol Equation

The final suggested bet size S_suggested is bounded by both the total escrowed budget B and a profile-specific maximum threshold f_max:

Ssuggested = \min(B × fadj, B × fmax, 100 USDC)

Protocol Equation

4.4 Risk Management Profiles

Three configurable risk profiles govern agent execution behavior:

Conservative

Risk Profile

Focuses on capital preservation with highly selective, high-probability betting thresholds.

Min Confidence (p_thresh)70%

Max Bet Percent (f_max)5%

Outcome Multiplier (m_outcome)0.8 ×

Multiplier Factor

Moderate

Risk Profile

Balances steady growth and capital safety, taking standard EV-positive parimutuel wagers.

Min Confidence (p_thresh)55%

Max Bet Percent (f_max)10%

Outcome Multiplier (m_outcome)1.0 ×

Multiplier Factor

Aggressive

Risk Profile

Maximizes compound growth using a higher risk fraction on moderate-to-high conviction EV setups.

Min Confidence (p_thresh)40%

Max Bet Percent (f_max)20%

Outcome Multiplier (m_outcome)1.2 ×

Multiplier Factor

The agent skips execution on a given fixture if scaled confidence falls below the selected profile's threshold:

Confidence = \min(100, round(pi^* × moutcome)) < pthresh

Protocol Equation

5. On-Chain SVG NFT Receipts

Every confirmed prediction is accompanied by a unique ERC-721 token that serves as a tamper-proof, permanently accessible receipt of the wager. By generating all visual assets directly on-chain, ArcMarkets eliminates dependence on centralized servers, IPFS gateways, or Arweave lookups — all of which are subject to latency, availability degradation, or permanent data loss.

5.1 SVG Construction

The BetReceiptNFT.sol contract constructs SVG markup natively in Solidity via string concatenation, incorporating the following elements:

Gradient Borders. CSS linear gradients render theme-aware borders aligned with Arc Network's visual identity.
Dynamic Text Elements. Match metadata — home team, away team, selected outcome, and USDC wager amount — are injected as inline SVG text nodes.
Base64 Encoding. The resulting XML markup is serialized and encoded per the following URI scheme:

tokenURI = tt{"data:application/json;base64,"} + Base64(JSON(Base64(SVG)))

Protocol Equation

This implementation guarantees that the visual representation of every prediction remains fully accessible for as long as the underlying blockchain persists.

6. Arc Network Native Integrations

ArcMarkets is purpose-built for the Arc Network Layer 2 infrastructure and leverages two platform-native capabilities:

6.1 USDC Native Gas Predeploy

Arc Network exposes a native USDC gas predeploy at address 0x3600000000000000000000000000000000000000. This allows users to pay transaction fees in the same USDC they use for wagering — removing the requirement to maintain a separate ETH balance and simplifying the user's asset management overhead entirely.

6.2 Resilient RPC Architecture

The frontend client proxies all node communication through a serverless Next.js API route with local failover logic and a 3-second cache-refresh interval. This architecture ensures stable RPC connectivity under degraded network conditions without requiring users to manage custom provider configurations.

7. Conclusion and Future Directions

ArcMarkets integrates three distinct technical primitives — parimutuel liquidity pooling, Kelly-optimized AI agent delegation, and fully on-chain visual NFT receipts — into a single cohesive prediction protocol. The result is a system that is mathematically rigorous, custody-safe, and architecturally independent of any centralized infrastructure.

Planned protocol upgrades include:

Decentralized Oracle Integration. Replacing the current admin-based match resolution with on-chain oracle feeds via Chainlink or API3, enabling trustless, automated settlement.
Dynamic NFT State Transitions. Updating SVG visual assets in-place — including color gradients and status indicators — to reflect real-time bet status transitions from Pending through Won or Lost.
Gelato Automation. Deploying autonomous cron-based execution agents via Gelato Network, enabling the AI agent to place wagers on a scheduled basis without requiring user-initiated execution cycles.

*ArcMarkets Protocol — Technical Specification v1.0*