🥋DoJo
What is DoJo?
Overview
DoJo is Stimpak’s algorithmic yield engine — a fully deterministic machine-learning framework that converts raw crypto market data into real, risk-managed trading strategies. Users deposit stablecoins into the DoJo Vault to earn yield from live, market-neutral strategies deployed on-chain.
📄 Litepaper 2025 Explore the full technical breakdown of Dojo’s ML pipeline.
TL;DR (for the degens): DoJo is where your USDC goes to train. Our AI does the sparring — you earn the rewards.
Philosophy
DoJo is built on two uncompromising principles:
Precision before prediction – every signal, trade, and allocation follows a deterministic path validated across houndreds of thousands of market samples.
Capital protection – strict guardrails govern exposure, drawdowns, and execution quality in real time.
Unlike most trading bots that “learn on the fly,” DoJo freezes every model constant, scaling factor, and cluster centroid once validated — ensuring the exact same logic runs live as in testing. This is institutional reproducibility, not statistical hope.
Why DoJo?
Traditional quant systems often rely on adaptive neural nets or superficial backtests that crumble under new conditions. DoJo’s edge lies in structure — a modular research architecture that discovers, validates, and deploys strategies like clockwork.
Discovers patterns – PCA and K-Means clustering reveal recurring market “micro-structures.”
Validates realism – walk-forward simulations use true price paths with execution delays and fee modeling.
Controls correlation – community detection ensures strategies aren’t secretly overlapping.
Balances portfolios – capital is allocated by stability and expectancy, not by hype.
Enforces discipline – live risk rules cap exposure, drawdowns, and daily loss limits.
How It Works (High-Level)
DoJo’s Deterministic ML Pipeline runs through seven repeatable stages:
Feature Engineering – converts raw OHLCV data into normalized indicators: volatility, momentum, structure, and higher-timeframe context.
Pattern Discovery – compresses thousands of overlapping sequences into distinct trading behaviors using PCA + K-Means clustering.
Signal Validation – stress-tests each cluster with walk-forward replays to define realistic stops and targets.
Execution Realism – models fill delays, order cancellations, and fee slippage for live-grade expectancy.
Statistical Confidence – bootstrap resampling quantifies uncertainty, generating confidence intervals and stability scores.
Portfolio Construction – groups correlated strategies and ssigns per-trade risk (0.5–2%).
Deployment & Risk Management – strategies are packaged into JSON bundles with built-in drawdown stops, loss limits, and drift detection.
What Users Get
By participating in the DoJo Vault, users gain:
Passive yield backed by real algorithmic trading performance.
Transparency — every deployed strategy is versioned and auditable.
Diversification — future multiple time horizons, instruments, and trade types.
Risk alignment — profits and protection built into the same system.
Agent compatibility — future integrations will let users launch their own DoJo-powered trading agents.
Our Edge
DoJo isn’t another signal bot or speculative quant. It’s a closed research-to-execution ecosystem that merges deterministic ML, portfolio science, and live governance.
Engineered features per bar
100 +
Historical sequences clustered
500K +
Bootstrap simulations per setup
5 000 +
Active strategy bundles
25 +
This infrastructure lets DoJo operate as an edge factory — continuously discovering and deploying statistically valid strategies across changing market regimes.
Risk Guardrails
Single-Trade Risk
≤ 3 %
Contains loss magnitude
Daily Loss
≤ 2 %
Auto cooldown trigger
Max Positions
3
Avoids overcrowding
Emergency Stop
−15 %
Capital preservation
All risk rules are enforced on-chain by the runtime manager. If performance drifts beyond statistical confidence bounds, affected strategies pause automatically.
Executive Summary
DoJo represents the evolution of Stimpak’s quantitative research into a production-grade yield system. It combines deterministic data science with on-chain execution to deliver stable, market-neutral returns — without hidden black boxes.
Our mission:
Generate consistent, uncorrelated yield through validated ML strategies.
Preserve user capital with dynamic guardrails and real-time oversight.
Bridge DeFi and data science by making institutional-quality automation accessible to everyone.
DoJo trains the data. You earn the discipline.
Last updated