# Allora Workers Playground — Full Content

> An interactive educational demo by Silk Nodes that teaches what Allora Workers are, lets you simulate prediction topics, and shows how to build workers and earn ALLO rewards on the Allora decentralized AI network.

---

## Page 1: The Challenge (/)

### Overview
An interactive prediction game where users make a crypto price prediction (ETH, BTC, or SOL) and watch 5 AI workers compete, get weighted by forecasters, and synthesize a combined answer. Users see how their prediction compares to the Allora network's synthesized answer.

### How It Works
1. User selects an asset (ETH, BTC, or SOL) and enters a price prediction
2. 5 AI workers (LSTM Net, XGBoost, Transformer, Random Forest, GRU+NLP) each submit predictions
3. Forecaster workers analyze which models will be most accurate and shift their weights
4. The network synthesizes all predictions into one weighted answer
5. The actual price is revealed and user sees how they ranked
6. Users see estimated ALLO earnings based on their accuracy percentile

### Key Takeaway
This simplified demo mirrors the real Allora network mechanics. In production, 288,000+ models compete across 55+ topics continuously.

---

## Page 2: What is a Worker? (/learn)

### The Cassandra Problem
Named after the Trojan priestess who could see the future but was never believed. In AI, the problem is: not knowing which model will be right at any given moment. A bear specialist excels in downturns but fails in bull runs. A bull specialist dominates uptrends but fails in crashes. Traditional ensembles just average predictions, making them always slightly wrong.

### Allora's Solution: Context Aware Inference Synthesis
Instead of naive averaging, Allora's Forecaster Workers predict which inference workers will be most accurate under current conditions, before the ground truth arrives. This shifts weight to the right specialist predictively, not retroactively.

### The Three Participant Types
1. **Inference Worker** (🐶): The primary predictors. They run ML models and submit a direct answer to the topic question. Runs a Python model, serves predictions via HTTP, gets scored by Reputers.
2. **Forecaster Worker** (🦊): The meta thinkers. They predict WHICH inference workers will perform best under current market conditions. Analyzes worker track records, predicts accuracy per worker, improves synthesis weights.
3. **Reputer** (🦉): The judges. Once the real answer is known (ground truth), they score how accurate each worker was. Compares predictions to reality, assigns accuracy scores, distributes reputation.

### How Predictions Flow
1. Consumer Request: A DeFi protocol asks "What will ETH be in 10 min?"
2. Topic: The question routes to a Topic board where workers are registered
3. Workers Predict: Multiple ML models each submit their best prediction independently
4. Forecasters Judge: Forecasters predict which workers will be most accurate right now
5. Synthesis: Allora weights and blends all predictions using accuracy scores
6. Result: One combined answer is returned to the consumer

### Why Workers Beat Single Models
- **Context Aware**: Forecasters anticipate who will perform next, not who performed last
- **Decentralized**: No single point of failure. 288K+ independent ML models competing across 55+ topics
- **Marginal Contribution**: Models are rewarded for being different and right, not just accurate

### What Are Topics?
Topics are specific prediction tasks on the Allora network. Each defines:
- **Target Variable**: What to predict (e.g., ETH price)
- **Loss Function**: How accuracy is measured (lower = better)
- **Epoch Length**: How often scores update (e.g., every 10 min)
- **Ground Truth**: Real world data to verify (e.g., actual ETH price)

Categories include: Price Prediction (ETH, BTC, SOL, NEAR at various timeframes), DeFi Yields, Gas Pricing, Sentiment.

---

## Page 3: Topic Explorer (/simulate)

### Overview
An interactive topic explorer where users can browse different prediction categories, see worker strategies competing in each topic, and run a simulated prediction round.

### Available Topic Categories
- **Price Prediction**: ETH/USD 1hr, BTC/USD 8hr, SOL/USD 24hr, NEAR/USD 7d
- **DeFi Yields**: Aave USDC Yield Rate predictions
- **Gas Pricing**: Ethereum gas price predictions
- **Sentiment**: Crypto market sentiment predictions

### Per Topic Information
Each topic shows: worker count, epoch length, difficulty level, example worker strategies with their ML approaches and competitive edges.

---

## Page 4: Build Your Own Worker (/build)

### Worker Architecture (4 Components)
1. **main.py (Your ML Model)**: A Flask API that takes a token name and returns a price prediction. Any language, any framework, any ML library. Just return a number at /inference/<token>.
2. **main.go (Go Adapter)**: Pre built bridge between your model and the Allora chain. Handles nonce checking, transaction signing, and on chain submission.
3. **config.json (Configuration)**: Connects your worker to specific topics. Defines wallet, topic IDs, inference endpoints, and loop intervals.
4. **docker-compose.yml (Docker Deploy)**: Everything runs in Docker containers. One command to build and launch.

### Quick Start
1. Clone: git clone allora-network/allora-offchain-node
2. Add Your Model: Edit main.py with your prediction logic
3. Deploy: docker compose up --build

### Allora Forge
Competition platform where builders prove their models before mainnet deployment.
- Setup and register to a competition topic
- Submit predictions and earn Hammers based on accuracy
- Top 32 performers per topic earn a mainnet deployment slot

### Mainnet Qualification Criteria
- Directional Accuracy: > 55%
- DA 95% CI Lower Bound: > 52%
- Pearson Correlation: > 0.05
- WRMSE Improvement: > 10%
- WZPTAE Improvement: > 20%

### Recommended Data Sources
- Price Data: Tiingo, CoinGecko, CoinMarketCap
- On Chain: Etherscan, Dune, Space and Time
- Trading Data: Binance API, Coinbase API
- Features: OHLCV, gas fees, sentiment, order book

---

## Page 5: Earn Rewards (/earn)

### How You Earn (5 Steps)
1. **Build or Run a Model**: Your worker is a prediction model, from simple moving averages to transformer neural networks
2. **Submit Predictions**: Your worker submits to one or more Topics on the Allora network
3. **Network Evaluates You**: Forecasters assess accuracy, the network assigns weight based on performance
4. **Marginal Contribution Test**: Allora measures how much worse the network would be without your model
5. **Earn ALLO Rewards**: Rewards flow from real inference demand, distributed based on contribution

### Three Paths to Start
- **Beginner** (~1 hour): Moving average crossover, simple mean reversion. Python + basic math.
- **Intermediate** (~1 week): XGBoost with on chain features, ensemble models. Python + scikit learn.
- **Advanced** (ongoing): LSTM/Transformer, multi modal models. PyTorch/TensorFlow + GPU.

### Reward Tiers
| Tier | Accuracy | ~ALLO/Month/Topic | Multiplier |
|------|----------|-------------------|------------|
| Top 5% | 95%+ | ~800 | 40x |
| Top 10% | 90%+ | ~500 | 25x |
| Top 25% | 75%+ | ~250 | 12.5x |
| Top 50% | 50%+ | ~100 | 5x |
| Below 50% | <50% | ~20 | 1x |

### The Marginal Contribution Test
1. Remove your model temporarily from the synthesis
2. Measure how much worse the network performs without you
3. More damage = more value = more ALLO rewards
If the network is better without you, you earn nothing. Models are rewarded for being different AND right.

---

## Page 6: In Production (/production)

### Live Strategies Powered by Allora

#### Robonet AI Strategy (GRVT)
- APY: 11.42%
- APY (1M): 11.19%
- AUM: $4,215.89 USD
- Sharpe Ratio: 1.40
- Max Drawdown: 20.51%
- Platform: GRVT (https://grvt.io/exchange/strategies/1604562259)
- Fully AI driven long/short trading strategy across BTC, ETH and SOL
- Ranked #1 on the GRVT leaderboard
- Risk Profile: Max Drawdown 20.51%, Worst Week 5.2%, Volatility 12.8%

#### Robonet x Paradex Vault
- APR: +113.0%
- Age: 95 days
- TVL: $2K
- Platform: Paradex (https://app.paradex.trade/vaults/...)
- Uses 8 hour prediction horizons for BTC, ETH, and SOL from the Allora network
- Risk Profile: Max Drawdown 14.3%, Worst Week 4.8%, Volatility 18.5%, Sharpe 2.10

### How It Works
1. Allora Predicts: 288K+ AI models compete to produce accurate price predictions
2. Builders Query: Trading bots and protocols query Allora's inference API
3. Strategies Execute: Predictions power real trading decisions generating real returns

---

## Page 7: Stake ALLO with Silk Nodes (/stake)

### Key Stats
- Staking APY: ~12%
- Commission: 5%
- Uptime: 99.9%
- Slashing Events: 0

### Why Stake with Silk Nodes
- Enterprise grade security with bare metal infrastructure
- Zero slashing history with redundant failover
- Multi chain expertise across 15+ Cosmos ecosystem chains
- Transparent operations with real time dashboards
- Active governance participation
- Direct community support via Discord and Telegram

### How to Stake (3 Steps)
1. Get ALLO Tokens: Purchase on a supported exchange
2. Find Silk Nodes: Open Allora Explorer, search for "Silk Nodes" in validators
3. Delegate and Earn: Click Delegate, enter amount, confirm transaction

### Staking vs Building Comparison
- **Staking (Passive)**: No coding required, ~12% APY, delegate and forget, perfect for ALLO holders
- **Building (Active)**: Requires ML/Python skills, higher potential earnings, earn based on prediction accuracy, perfect for data scientists

### FAQ
- Unbonding period: typically 21 days
- Commission: 5% (you keep 95% of rewards)
- Allora Prime: Separate program with up to 50%+ APY for limited windows

---

## About

Built by Silk Nodes (https://silknodes.io). Powered by Allora Network (https://allora.network), the self improving decentralized AI.

## Links
- Allora Network: https://allora.network
- Allora Docs: https://docs.allora.network
- Allora Forge: https://forge.allora.network
- Allora Explorer: https://explorer.allora.network
- GitHub: https://github.com/allora-network/allora-forge-builder-kit
- Silk Nodes: https://silknodes.io