Confessions of a Predict Addict

A personal introduction — and what’s coming next

Confessions of a Predict Addict

A personal introduction — and what’s coming next

🧠 “We don’t just need better predictions — we need to know when to trust them.”

Hi, I’m Valeriy Manokhin, and I have a confession:

I’m obsessed with predictions.
Not just making them — but understanding when to trust them, how to measure their uncertainty, and what to do when they inevitably fail.

This obsession has taken me from finance to achine learning research, through PhD and MBA work, into classrooms, codebases, and client projects.

I’ve written books. I’ve published machine learning papers. Built models. Taught courses. And the more I learn, the more one thing becomes clear:

Most people don’t talk about prediction the way real practitioners experience it.

They focus on accuracy, not calibration.
They build models, not systems.
They chase hype, not reliability.

That’s why I’m starting this newsletter.

---

🧪 Why This Substack?

I created Confessions of a Predict Addict to bring serious, applied, code-backed thinking to an area that needs it:
Prediction — from theory to uncertainty to real-world deployment.

This Substack is where I’ll share:

• How to quantify uncertainty in ML (e.g., conformal prediction, calibration)

• How to forecast — markets, time series, risk, behavior — across domains

• How to apply ML responsibly in complex systems, especially in finance and decision-critical industries

• How to bridge the gap between academic research and high-stakes implementation

This isn’t hype, and it isn’t marketing fluff. It’s practical, rigorous, and sometimes uncomfortable truth-telling — from someone who’s been in the room where the predictions go wrong.

---

📬 What You’ll Get (and How Often)

Posting cadence: 2–4 posts per month

🔍 Expect content like:

• 🔬 In-depth tutorials with clean, executable Python code

• 📈 Real-world case studies from finance, forecasting, and beyond

• 🧠 Practical commentary on new research and methods

• 🧩 Essays and lessons learned from deploying predictive models

• 🧰 Code snippets and tools you can use right away

Some content will be free, including essays, ideas, and overviews.
Other content — including full tutorials, case studies, walkthroughs, and tools — will be for paid subscribers.

👉 Subscribe now — free or paid — and never miss an issue
👉 Become a founding member to unlock exclusive benefits and support this work from day one

---

💡 Who This Is For

This newsletter is built for people who care about getting predictive modeling right:

• ML engineers & data scientists tired of shallow advice

• Researchers & PhD students trying to bridge theory and industry

• Finance & risk professionals who deal with model failure and uncertainty

• Forecasting practitioners who need interpretable, robust models

• Anyone building systems that rely on predictive insight

If you’ve ever had a model break at the worst moment… if you’ve been asked to “just make it more accurate”… or if you’ve thought “what do we really know about our model’s confidence?” — this is for you.

---

🚀 Why Now?

We live in a world increasingly built on predictions — from market forecasts and demand modeling to algorithmic risk decisions and AI-generated insights.

But:

• Most orgs don’t understand prediction error

• Most of data scientists don’t understand time series and forecasting

• Most models don’t expose uncertainty

• Most teams don’t know when to trust their outputs

This newsletter aims to help fix that.

One post at a time.

---

💬 How You Can Help

If this resonates with you:

• 👉 Subscribe for free to get posts in your inbox

• 👉 Upgrade to paid or founding member to support deep, independent content — and get access to premium tutorials, tools, and more

• 🔄 Share this with a colleague, team, or friend who works with predictive systems

• 💬 Comment or reply — tell me what you’re working on, what you’d like to see covered, or just say hello

---

👋 Let’s Build Something Smarter

This isn’t about chasing trends. It’s about building predictive systems we can understand, stress test, and trust — because the world increasingly depends on them.

Thanks for being here at the start of Confessions of a Predict Addict.
Let’s explore this space together.

– Valeriy

---

P.S. First deep-dive post coming soon. If you're curious about conformal prediction, forecasting methods that actually work, and how to code them — stay tuned.

👉 Subscribe now
👉 Become a founding member

---