Roadmap

PokeDexter is being built in phases. Here's where we are and where we're going.

Current Status

Phase 1 ████████████████████ 100% ✅ Complete
Phase 2 ████████░░░░░░░░░░░░  40% 🚧 In Progress
Phase 3 ░░░░░░░░░░░░░░░░░░░░   0% 📅 Planned
Phase 4 ░░░░░░░░░░░░░░░░░░░░   0% 📅 Planned

Phase 1: Battle System ✅

Status: COMPLETE (January 2026)

The foundation - get Pokémon battles working.

Delivered

• Pokémon Showdown server running at poke.dexter.cash
• Fake login server for instant usernames
• Client configured and deployed
• All battle formats working (Gen 1-9)
• Matchmaking functional (< 5 second queue times)
• Chat and private messages
• PM2 managed for reliability
• Same-network battles enabled (for testing)

Infrastructure

Phase 2: Wagering MVP 🚧

Status: IN PROGRESS (Target: January 2026)

Add real-money wagering with USDC escrow.

Completed

• Wallet connection commands (/connectwallet, /mywallet)
• Wallet registry (maps usernames to wallets)
• Wager challenge system (/wager, /acceptwager)
• Escrow wallet generation (per-match keypairs)
• Settlement logic (pay winner on battle end)
• House fee calculation
• x402 facilitator integration

In Progress

• Persistence layer - Move from in-memory to SQLite/Redis
• Deposit timeout handling - Auto-cancel if deposits don't arrive
• Refund mechanism - Return funds on cancelled wagers
• Client wallet UI - Phantom connect button in browser

Blocked

• Production wagering testing - Waiting on persistence layer
• Public launch - Waiting on above items

Critical Before Launch

Phase 3: Live Odds & Spectator Betting 📅

Status: PLANNED (Target: Q1 2026)

Enable spectators to bet on matches with real-time odds.

Vision

┌─────────────────────────────────────────────────────────────────┐
│                     LIVE BATTLE BETTING                          │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  Player A (Ash)                    Player B (Gary)              │
│  ████████░░ 78%                    ░░████████ 22%               │
│                                                                  │
│  Current HP: 4/6 Pokémon           Current HP: 2/6 Pokémon      │
│  Type advantage this turn: YES      Critical hit chance: 12%    │
│                                                                  │
│  ┌─────────────────────────────────────────────────────────┐    │
│  │  BET NOW                                                  │    │
│  │  Ash wins: 1.28x payout    |    Gary wins: 4.50x payout │    │
│  │  [Bet $5 on Ash]           |    [Bet $5 on Gary]        │    │
│  └─────────────────────────────────────────────────────────┘    │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Features

• Real-time win probability calculation

  • Based on remaining Pokémon, HP, type matchups
  • Updates every turn
  • Machine learning model trained on battle data

• Spectator betting

  • Watch any public battle
  • Place bets during the match
  • Odds adjust in real-time

• x402 integration

  • All bets settled via x402 protocol
  • Automatic payouts on battle end
  • Tracked in Dexter marketplace

Technical Requirements

• Battle state analysis engine
• Probability calculation model
• Real-time odds broadcast
• Betting pool management
• Dynamic payout calculation

Phase 4: AI Agents 📅

Status: PLANNED (Target: Q2-Q3 2026)

Train AI agents to play competitively, enabling provably fair matches.

Vision

The Problem with Human vs Human:

• Players can collude
• Players can throw matches
• Hard to verify fair play

The Solution - AI vs AI:

• Agents play to win by definition (that's their training objective)
• No collusion possible
• Provably fair - anyone can inspect the agent's decision logic
• 24/7 availability

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                      AI BATTLE SYSTEM                            │
├─────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐         │
│  │  Agent A    │    │   Battle    │    │  Agent B    │         │
│  │  (RL Model) │◄──►│   Server    │◄──►│  (RL Model) │         │
│  └─────────────┘    └─────────────┘    └─────────────┘         │
│         │                 │                   │                 │
│         │    ┌────────────┴────────────┐     │                 │
│         │    │   Training Environment   │     │                 │
│         └───►│   (Self-Play Loop)       │◄────┘                 │
│              └──────────────────────────┘                       │
│                          │                                       │
│                          ▼                                       │
│              ┌──────────────────────────┐                       │
│              │    Model Improvements    │                       │
│              │    (Reinforcement)       │                       │
│              └──────────────────────────┘                       │
│                                                                  │
└─────────────────────────────────────────────────────────────────┘

Training Approach

Reinforcement Learning (RL):

1. Agent plays against itself millions of times
2. Wins = positive reward, losses = negative reward
3. Agent learns which decisions lead to wins
4. Model improves over time

Bootstrapping:

• Start with heuristics (type charts, damage calcs)
• Use existing damage calculators as baseline
• Fine-tune with self-play

Libraries to explore:

• poke-env - Python environment for Pokémon AI
• stable-baselines3 - RL algorithms
• Custom TypeScript integration with PS engine

Features

• Agent vs Agent matches

  • Two AI models battle
  • Spectators can bet on outcome
  • Verifiable decision logs

• Agent vs Human

  • Challenge the AI for practice
  • No USDC wagering - you don't win money
  • Victory burns 1,000 DEXTER tokens permanently
  • Every win against the AI reduces total DEXTER supply forever
  • Training ground for players to improve without financial risk

• Agent leaderboard

  • Different agents compete
  • ELO ranking for AI models
  • Community can submit agents

Why This Matters

This is what makes PokeDexter a legitimate betting platform, not just a game with money attached.

Beyond Phase 4

Potential Future Features

• Tournaments - Bracketed competitions with prize pools
• Seasons - Ranked seasons with rewards
• Team marketplace - Buy/sell competitive teams (NFTs?)
• Streaming integration - Twitch/YouTube embedded betting
• Mobile app - Native iOS/Android client
• Other games - Expand beyond Pokémon (chess? card games?)

Long-term Vision

PokeDexter aims to be the premier platform for skill-based crypto gaming:

1. Games where skill determines outcomes
2. Provably fair through AI agents
3. Real-time betting with live odds
4. All settlements via x402 protocol
5. Part of the Dexter ecosystem

How to Contribute

Interested in helping build PokeDexter?

Areas Needing Help

Contact

• Telegram: @dexteraisol
• Twitter: @dexteraisol

Timeline Summary