Getting started using Python for algorithmic trading, market data analysis, and quant finance sucks.
500,000,000 Google results for "python tutorial" (or some weird AI generated nonsense).
Hundreds of "agent coding IDEs."
Billions of YouTube videos.
The Python ecosystem is huge (and growing):
- PyPI, the official Python package index, lists more than 900,000 projects as of this week.
- In the 2025 Stack Overflow Developer Survey, 66% of developers said AI tools give them solutions that are "almost right, but not quite."
- In the survey's AI section, 20% of developers said they've become less confident in their own problem-solving while using AI tools.
The size of the ecosystem is a strength because it's so powerful. But it's also a weakness if you're trying to get started.
The 3-step roadmap out of overwhelm
Each step takes a week or two and builds on the one before it.
Step 1: Pick one problem (week 1)
Choose one task you already do by hand. Something like computing one-day value at risk (VaR) on a 10-stock portfolio.
Write the goal as one sentence with numbers in it: "Compute 95% one-day historical VaR from five years of daily prices." If you can't write that sentence, the problem is still too big.
A clear goal tells you which 20 lines of code you need and which 900,000 projects you can ignore.
Step 2: Write minimal viable Python (weeks 2 to 3)
I call this MVP, for minimal viable Python. Write only the code the goal needs. For VaR, that's daily returns from pct_change() and the cutoff from np.percentile(returns, 5).
Use ChatGPT or Claude in a chat window (no fancy agent coding harness required) and build one function at a time. Then ask it to explain every line before you run anything.
Here's what this looks like:
Step 3: Keep building on it (week 4 onward)
Run the script on fresh prices each week and add one feature at a time. First, plot the return distribution.
Then add expected shortfall next to VaR.
Don't throw old versions away.
Save each one in a Git repo, because last month's script is next month's starting point.
That's how a 15-line script turns into a toolkit you understand.
Open your Step 2 script and change the confidence level from 95% to 99%.
If you can do it in five minutes without asking an AI, you did it!