Collections
Curated reading paths through the blog, ordered from basics to advanced.
Backtesting Without Fooling Yourself
A step-by-step path from what your backtest really optimizes to proving an edge survives overfitting, multiple testing, and live execution. Read top to bottom — each part builds on the last.
- 01 目標函數設計：你最佳化的那個指標，正悄悄替你選好了策略
- 02 證明多時間框架回測中沒有look-ahead：擾動未來，證明過去看不到它
- 03 前視偏差：一根 K 線的錯誤如何從純噪聲中憑空製造出 15 的夏普比率
- 04 Walk-Forward 最佳化：唯一誠實的策略測試方法
- + 7 more
High-Performance Backtest Engines
How to build a backtest engine that runs hundreds of times faster without changing a single PnL number — data layout, caching, adaptive resolution, and architecture, from first speedups to production internals.
- 01 回測引擎速度階梯：筆記本 CPU 上 298 倍提速，PnL 精確到最後一筆交易
- 02 框架稅：當你的回測庫比手寫的 pandas 迴圈還慢
- 03 聚合 Parquet 快取：如何將多時間框架回測加速數百倍
- 04 雙軸參數空間：為什麼你的大部分參數掃描幾乎是免費的
- + 7 more
Complex Arbitrage in Rust
A six-part build-up of multi-leg crypto arbitrage — from negative-cycle detection to the linear algebra, copulas, and machine learning behind it, ending in low-latency Rust execution.
- 01 套利檢測的圖演算法：從 Bellman-Ford 到 RICH
- 02 期貨-現貨套利：從期現套利到 DeFi-CeFi
- 03 矩陣、張量與熱帶代數：用於套利檢測的線性代數
- 04 套利中的 Vine Copulas：高維依賴關係建模
- + 2 more
Order Book & Market Microstructure
How the order book really works — accessing the data, reading queue position, rebuilding bars from order flow, and modeling it with deep learning and Hawkes processes.
- 01 CCXT：WebSocket 訂單簿方法實際工作原理
- 02 演算法交易中的訂單型別：從追價限價單到虛擬訂單
- 03 牆內排隊：訂單簿密集區的掛單位置分析
- 04 演算法交易的K線型別與聚合方法
- + 8 more
Portfolio Construction & Risk
From Markowitz to production HRP + CVaR: how to allocate across crypto assets, model tail dependence with copulas, and size positions without blowing up.
- 01 馬科維茨投資組合理論之加密貨幣篇：從零到英雄
- 02 12種投資組合最佳化演算法比較：HRP、Black-Litterman、NCO及其他
- 03 我們的內部演算法揭秘：HRP + 多空 + 基於Hull-White的CVaR
- 04 Copula模型：加密投資組合聯合風險建模
- + 1 more
Statistical Arbitrage & Pairs Trading
Trade the spread between correlated assets — from the distance approach to cointegration and Kalman filters, then dynamically combining mean reversion with momentum.
- 01 配對交易中的距離法：Rust實現與分析
- 02 加密貨幣市場的統計套利與配對交易：從協整到卡爾曼濾波
- 03 訊號相關性：需要監控多少個交易對
- 04 統計套利中均值迴歸與動量策略的動態結合：數學基礎與實踐實現
- + 1 more
Deep Learning for Markets
Neural forecasting for crypto — transformers, diffusion models, and foundation models, and how conformal prediction keeps their uncertainty honest.
- 01 用於多時間跨度組合預測的時序融合 Transformer
- 02 擴散模型對抗加密貨幣無政府狀態：為什麼DDPM比你的占星師更能預測比特幣崩盤
- 03 Kronos：讓K線圖說Transformer語言的基礎模型
- 04 用於風險感知倉位管理的保形預測
AI Agents for Trading
The agentic-AI stack for markets — multi-agent frameworks, open-source hedge funds, and LLMs that mine alpha from earnings calls.
- 01 基於代理型AI的投資組合管理革命
- 02 TradingAgents：模擬對沖基金的多智慧體AI交易框架
- 03 AI4Finance Foundation：FinGPT、FinRL和FinRobot量化交易生態系統
- 04 AI Hedge Fund：AI分析師投票決定交易的多智慧體基金
- + 1 more
QuestDB for Algorithmic Trading
Stand up a time-series stack for trading on QuestDB — from architecture to the SQL that matters, to a production deployment.
- 01 QuestDB 演算法交易實戰：讀懂市場語言的架構設計
- 02 QuestDB 演算法交易實戰：改變遊戲規則的 SQL 擴充
- 03 QuestDB 演算法交易實戰：從訂單簿到生產架構
Low-Latency Trading Infrastructure
The plumbing under an HFT stack — how components talk (WebSocket, FIX, gRPC, Aeron), messaging on Aeron and Zig, and a C++ FIX/FAST scalper.
- 01 演算法交易系統中的資料通訊：技術綜述
- 02 Aeron：驅動半個HFT行業的訊息傳遞系統揭秘
- 03 ZigBolt：為什麼我們用 Zig 從零打造了自己的 Aeron，實現了每條訊息 20 納秒延遲
- 04 使用 FAST/FIX 開發簡單 C++ 剝頭皮交易機器人：分步指南
Execution & Market Making
From TWAP/VWAP and Almgren-Chriss to TCA, slippage models, and the Avellaneda-Stoikov market maker — how to turn a signal into fills without paying the spread twice.
- 01 TWAP、VWAP與POV對比：如何選擇執行基準（以及何時它們會對你撒謊）
- 02 去掉玄乎其辭的 Almgren-Chriss：一個下午就能實現的最優執行模型
- 03 切片內部:排程器與交易所之間的子訂單戰術
- 04 加密市場的智慧訂單路由：一筆訂單，十二個場所，沒有 NBBO
- + 7 more
Onchain & DeFi Quant
Quantitative DeFi from first principles — MEV and sandwich attacks, atomic arbitrage, liquidation cascades on lending protocols, and LP profitability on Uniswap v3.
- 01 MEV 剖析：三明治攻擊、搶先交易與記憶體池的黑暗森林
- 02 MEV 供應鏈：PBS、MEV-Boost 與區塊市場
- 03 鏈上套利：原子循環、閃電貸與你必須贏得的拍賣
- 04 鏈上清算：Aave 與 Compound 的機制，以及圍繞它們的機器人生意
- + 5 more
Causality & Lead-Lag in Markets
Who moves whom — Granger causality, transfer entropy, PCMCI discovery, and the causal ML toolbox (double ML, causal forests, synthetic control) applied to crypto markets.
- 01 Toda-Yamamoto vs Differenced Granger: Does the BTC Lead-Lag Survive?
- 02 Transfer Entropy: Which Way Does Information Flow Between Crypto Assets?
- 03 PCMCI: Causal Discovery in Multivariate Crypto Time Series
- 04 交易中異質治療效果的因果森林
- + 3 more
Volatility Modeling & Regimes
Forecast volatility with the GARCH family — asymmetry and leverage effects, dynamic correlations, vol targeting — then detect market regimes with HMMs and trade adaptively.
- 01 GARCH(1,1)：預測加密貨幣波動率
- 02 非對稱與厚尾GARCH：EGARCH、GJR與Student-t
- 03 DCC-GARCH：配對交易與組合風險的動態相關性
- 04 波動率目標化與基於GARCH預測的交易
- + 1 more
Modern ML for Trading
The full gradient-boosting-to-neural-operator toolkit — ensembles, AutoML, probabilistic forecasting, distillation and pruning for latency, plus physics-informed architectures like Neural ODEs and Fourier operators.
- 01 用於報酬方向的 XGBoost：類別不平衡與決策閾值
- 02 整合方法：結合弱學習者以獲得穩健的 Alpha
- 03 用於系統化交易流程的 AutoML
- 04 Multi-Task Learning for Simultaneous Price, Volume, and Volatility Prediction
- + 11 more