Inside the Mind of an 18-Year-Old Quant: How Jorge Marino Built Algorithmic Trading Bots That Returned 746%

Marino: "I run three distinct strategies in parallel. The first is momentum detection with volume confirmation — it's looking for assets where price is breaking out but more importantly, where volume is at least 2.5x the 20-period average. The second is whale wallet tracking — I monitor the top 100 non-exchange wallets for each major asset. When I see accumulation patterns, my algorithm scores it and enters if multiple conditions align. The third is sentiment arbitrage — I scrape crypto Twitter, Reddit, and Discord, run NLP to gauge sentiment, and trade against extreme fear or greed."

Marino: "I never use leverage. That's rule one. Rule two is position sizing — no single position starts at more than 15% of portfolio. Rule three is systematic stops — I don't set emotional stops, my algorithms do. If a position drops 12% from entry, it's automatically liquidated. No exceptions. That's why my max drawdown is only 12.3% even during volatile periods."

Marino: "September 15th, 2025. My whale tracker flagged 17 wallets that had accumulated over 50,000 SOL each in a 48-hour period. These weren't exchange wallets — they were fresh wallets with no previous history. That's typically either institutional accumulation or a very confident whale. My sentiment module also showed developer activity on Solana had increased 40% month-over-month. The combination triggered a strong buy signal. I entered at $34.20."

"What's interesting is that I didn't sell at $100 or $150. My algorithm uses a trailing stop once a position is up 25% — it locks in profits while letting winners run. That's how I captured the move to $178."

Marino: "The core is Python — I use Web3.py for blockchain interaction, CCXT for exchange APIs, and TensorFlow for some of the predictive models. The infrastructure runs on a dedicated server in my room — nothing cloud-based because I don't want latency or dependency on third parties. I built a custom dashboard with React and Node that visualizes all the signals in real-time."

"The most sophisticated part is the NLP engine for sentiment. I fine-tuned a BERT model on six months of crypto Twitter data. It's not perfect, but when it hits extreme readings, it's remarkably accurate."

Marino: "Because I know my emotions are the enemy. The hardest trade I ever made was not selling when I was up 300% on Solana. Everything in me wanted to take profits. But the algorithm said hold, and I trusted it. That's the whole point — you build the system, then you get out of its way."

"I check the system twice a day — once in the morning before school, once at night. If there's an alert, I investigate. Otherwise, I let it run. Most people overtrade. I try to do the opposite."

Marino: "Eventually, yes. But I'm in no rush. I want to go to college first — probably computer science and finance. I've talked to a few people at Northwestern and UIUC about their quant finance programs. The goal is to keep building, keep refining the algorithms. By the time I graduate, I want to have a fully automated fund that can manage outside capital. But for now, it's just me, my servers, and my code."