• Deep Learning for Trading
  • GitHub: Deep Trader
    • Keywords: Direct Policy Optimization, Differentiable Backtesting
  • GitHub: Fast Trading Simulator
    • Numba accelerated trading simulator
  • GitHub: Crypto Data Downloader
    • High-speed crypto data downloader via concurrent API requests
  • GitHub: Crypto Monitor
    • Crypto market monitor via WebSocket

Archived Predecessors

Yinhan Tradebot Web App

• 2024-04-15 - 2024-08-26: My personal homemade tradebot achieved +300% profit with maximum -28% drawdown. I decided to convert it to a public App supporting multiple users. The name Yinhan (Serena) is from a girl that is really important to me.

• 2024-10-15 - 2024-12-14: The App was deployed running algorithm MSB0 with high-risk parameters aiming at quick return. It achieved +100% profit initially but later faced a severe drawdown of -60%, as the bot was shorting during a strong uptrend.

• 2024-12-19 - 2025-07-19: I switched to algorithm MSB1, it has the same mechanism as MSB0, but with much safer parameters. It performed steadily for 6 month, achieving +600% profit, until it crashed with a -30% drawdown in July 2025.

• I decided to stop the bot after the -30% drawdown, as it exceeded my statistical expectation and tolerance. This led me to start Serena Trading Research, the successor project that aims at: better modularity, partially open source and most importantly: much more robust trading algorithms. (as I realized that minimizing drawdown is far more important than maximizing profit!)

• More screenshots from the App:
  • A table of trades and a graph showing the entry and exit point (in the image below, shorting EPICUSDT during a strong uptrend led to a -40% loss)