I need repeatable historical data, parameters, trades, and diagnostics — not just one performance number.
← All projects
Case study / 2026
TMF Quant Trading Platform
An end-to-end engineering project for researching and operating Taiwan Micro TAIEX Futures strategies. The platform separates market-data ingestion, analysis, backtesting, replay, and trading workflows so each can evolve independently.
View public shared core on GitHub ↗The public core contains shared engineering components, not the complete production platform or private strategies.
The story
Suppose I have a trading strategy, and its backtest looks profitable.
The tempting next step is to let it trade. But a profitable backtest is only the beginning — it does not tell me whether the strategy can survive real data, execution, failures, and risk.
The same logic must work with replay and paper trading before it ever reaches a broker.
Stale data, disconnects, permissions, recovery, and a kill switch matter as much as the strategy itself.
So the real question becomes:
How do I turn a strategy idea into a system that can be tested, observed, and controlled before execution?
01ResearchIdea · Parameters
→ 02BacktestTrades · Diagnostics
→ 03ReplayHistorical events
→ 04PaperOrders · Risk
→ 05ControlAccess · Kill switch
The actual result
The strategy is no longer a backtest script. It becomes one part of a controlled trading platform.
Interactive result
Try the interactive strategy demo.
No login is needed to explore a fixed synthetic backtest, its chart, simulated trades, and strategy diagnostics. Full platform features require Email verification and account approval.
No login · Synthetic data
01 Interactive Strategy Demo
Fixed synthetic case · Backtest chart · Strategy diagnostics. No real trading or investment performance.
Try Interactive Demo Email verification · Approval
02 Full Platform Access
Verify your Email, then use “申請開通” to request access. An administrator reviews the account before full features are available.
Request Full Platform AccessSystem architecture
Separate the flow. Share the domain.
A layered view of the implemented system boundaries—not a claim about traffic volume or performance.
SOURCEMarket DataProvider · Tick stream
PROCESSEvent PipelineQueue · Candle aggregation
DOMAINStrategy EngineAnalysis · Composition
EXPERIENCETrading PlatformLive · Replay · Backtest · Paper
FOUNDATION Risk · Access Control · Observability · Recovery
Detailed architecture
Full system architecture
The complete path from market data and strategy decisions to controlled execution, risk management, recovery, and deployment.
Overview
One product, multiple engineering concerns.
An end-to-end engineering project for researching and operating Taiwan Micro TAIEX Futures strategies. The platform separates market-data ingestion, analysis, backtesting, replay, and trading workflows so each can evolve independently.
Problem
Problem
Strategy research often becomes fragmented across data scripts, one-off backtests, notebooks, and separate trading interfaces. This project explores how those workflows can share a coherent system boundary.
Goals
Goals
Create a maintainable path from market data to strategy analysis, historical replay, event-driven backtesting, risk controls, paper trading, and operational visibility.
Key Engineering Decisions
Key Engineering Decisions
Keep tick callbacks lightweight, aggregate candles in workers, separate market-data providers from broker accounts, reuse strategy analysis across live, replay, and backtest contexts, and isolate user-owned strategy data from shared market data.
AI-native Development Workflow
AI-native Development Workflow
Requirements and acceptance criteria are decomposed into focused changes. AI assists implementation while architecture choices, code review, tests, deployment checks, debugging, and acceptance remain explicit engineering responsibilities.
Features
Features
Live one-minute candle updates, historical replay, strategy backtesting, parameterized strategies, reusable strategy composition, paper-trading workflows, access control, and a unified trading workspace.
Reliability & Observability
Reliability & Observability
Operational state covers market freshness, provider connectivity, queue depth, WebSocket connections, database write latency, paper-order events, resource health, and a kill switch for halting new trading activity.
Security & Access Control
Security & Access Control
Cloudflare access controls protect the application boundary. Application permissions distinguish administration, market access, backtesting, strategy ownership, and trading capabilities. Provider and system settings stay hidden from general users.
Deployment
Deployment
The product runs independently at tmf.milespapa.com. Its application deployment, access controls, and operations remain decoupled from this Milespapa brand site.
Engineering Challenges
Engineering Challenges
Key challenges included exchange-time candle aggregation, reconnect behavior, mobile chart usability, data-provider decoupling, user data ownership, and safe behavior when market data becomes stale.
What I Learned
What I Learned
Shipping a usable platform requires more than a working strategy. Clear service boundaries, operational states, permissions, failure behavior, mobile constraints, and repeatable validation shape the product just as much as feature code.
Future Improvements
Future Improvements
Potential directions include deeper strategy research tooling, expanded market-data support, stronger automated verification, and more mature production operations. Scope and metrics will be documented only after implementation and validation.
Links