Our Machine Learning Investment Approach

We've spent years figuring out how algorithms can actually help with investment decisions. Not magic formulas or guaranteed returns—just a practical system that combines data analysis with real market understanding.

What started as experiments in 2019 has turned into a framework we trust. We're teaching it because we think more people should understand how machine learning works in finance without the hype.

Napaporn Kositchaiwat, Lead Research Analyst
"

Most people think machine learning will predict the future. It won't. What it does is help you see patterns you'd miss otherwise and make more informed choices. That's valuable, but it's not magic.

Napaporn Kositchaiwat
Lead Research Analyst
6
Years Refining Models
47
Data Sources Integrated
1,200+
Hours Testing Strategies
89
Model Iterations

Four Core Principles

These aren't revolutionary ideas. They're just what we've learned works when you're building systems that need to perform consistently over time.

1

Data Quality Over Quantity

We used to think more data was always better. Turns out, clean, relevant data beats massive datasets full of noise. We spend probably 60% of our time just making sure what goes into the models is actually useful.

Example: We track 47 sources but only use about 22 regularly—the ones that consistently add value to specific strategy types.
2

Transparent Decision Logic

Black box algorithms make people uncomfortable, and honestly, they should. If you can't explain why a model made a recommendation, you can't trust it when market conditions change.

Our models document decision factors so you understand what influenced each analysis—not just "the computer said so."
3

Continuous Model Evolution

Markets shift. A model that worked last year might underperform this year. We update our approaches regularly based on performance feedback and changing market dynamics.

Between January 2024 and March 2025, we adjusted our volatility assessment model three times based on emerging patterns.
4

Risk-Aware Analysis

Returns are meaningless without context about risk. Every analysis we do includes clear information about potential downsides and uncertainty levels—because those matter as much as upside potential.

We flag when confidence is low rather than forcing predictions, which means sometimes the answer is "wait for better information."
Data analysis workspace showing multiple financial indicators

Multi-Source Analysis

Combining traditional metrics with alternative data streams for fuller market perspective

Machine learning model testing environment

Rigorous Backtesting

Every strategy gets tested across multiple market conditions before real application

Investment strategy documentation and review process

Documentation Standards

Clear records of model assumptions, limitations, and decision factors

Learning This Approach

We're running a structured program starting September 2025 where we teach these methods step-by-step. You'll work with real data, build actual models, and understand both what works and what doesn't.

It's not quick—takes about eight months to really grasp everything. But by the end, you'll have practical skills you can apply immediately. No promises about making you rich, just teaching you a valuable analytical approach.

View Program Details