The Rope Walk

Host: Today's guest represents a growing breed of trader: the quant who didn't start on Wall Street. Captain Yuki has a PhD in statistical physics, spent five years in academia, and taught himself to code in his thirties. He now runs a fully automated trading system that executes 50,000 trades a year. But the road from Excel to algorithm was not smooth. Yuki, welcome.

Yuki: Thank you. I should mention that I knew nothing about markets when I started. My first trade was in 2015. I bought Apple because I liked my iPhone. That's the level I was at.

Host: How did a physicist become a trader?

Yuki: Burnout. I was a postdoc at a university, studying phase transitions in complex networks. Fascinating stuff, but the pay was terrible and the job security was nonexistent. A friend showed me his brokerage account. He had turned $50,000 into $120,000 in two years with a simple momentum strategy. I thought: I can model this. I model complex systems for a living. How hard can financial markets be?

Host: Famous last words.

Yuki: Exactly. I spent six months building an Excel model. Every weekend, I'd download price data for the S&P 500 constituents, calculate 20-day and 50-day moving averages, rank them by momentum, and manually enter orders on Monday morning. It was absurd. I was spending 15 hours a week on something a computer could do in 15 seconds. But it worked. I made 18% in 2016. I was hooked.

Host: When did you automate?

Yuki: 2017. I learned Python from YouTube tutorials. I built my first script that connected to my broker's API. It would scan the market at 9:30 AM, generate signals, and email me a list. I still had to manually enter the trades, but the analysis was automated. That saved me 10 hours a week. Then I thought: why am I still entering trades manually?

Host: Full automation.

Yuki: January 2018. I wrote a script that not only generated signals but placed orders. It would calculate position size, submit the order, set stops, and send me a confirmation email. The first day it ran, I sat at my desk refreshing my email every 30 seconds. I couldn't believe I was letting a script handle real money. But it worked. It bought three stocks. Two went up. One went down. Net: +$400. I made $400 while eating breakfast.

Host: Then came the bug.

Yuki: March 14, 2018. I had added a new feature: dynamic position sizing based on volatility. The idea was to reduce size when the market was volatile and increase it when calm. I had tested it in backtests. It looked good. But I made a unit conversion error. The volatility was calculated as a percentage — 0.015 for 1.5%. My position sizing formula expected a decimal, but I had passed it the percentage multiplied by 100. So instead of sizing at 1%, it sized at 100%. The script saw a low-volatility signal and went all-in.

Host: How much did it buy?

Yuki: It bought $890,000 worth of a mid-cap semiconductor stock. My account was $450,000. It used 2× margin. I didn't even know I had margin enabled. I was in a meeting when the order filled. I got the confirmation email and almost threw up. The stock was up 2% when I saw it. I sold immediately. After slippage and commissions, I lost $47,000. In 90 seconds. Because of a decimal point.

Host: What did you change?

Yuki: Everything. I implemented a three-layer safety system. Layer one: hard position limits. No single trade can exceed 5% of account equity, regardless of what the algorithm says. Layer two: daily loss limits. If the system loses more than 2% in a day, it shuts down until I manually restart it. Layer three: the human circuit breaker. Every order above $10,000 requires my phone approval via two-factor authentication. The algorithm can suggest. It cannot execute large orders without me.

Host: Do you still trade manually?

Yuki: Never. I haven't placed a manual trade in four years. My system runs 8 strategies across equities, futures, and FX. It executes roughly 50,000 trades per year. My job is not to trade. My job is to monitor the system, update the models, and ensure the safeguards are working. I spend maybe 10 hours a week on it. The rest of the time, I'm a stay-at-home dad. That's the real benefit of automation: it gives you your life back.

Host: What would you tell someone who wants to automate their trading?

Yuki: Start small. Automate the analysis first, then the execution. Use paper trading for at least six months. And never, ever trust a new piece of code with real money until it has survived a stress test. My $47,000 lesson was expensive. But it was cheaper than the alternative: continuing to trade manually and burning out, or deploying unchecked code and losing everything.