Mastering Modern
Time Series Forecasting
Trusted by data science leaders in 100+ countries.
The most comprehensive book on modern forecasting. From Slutsky and Yule to Chronos and TimesFM. ARIMA, ETS, machine learning, deep learning, transformers, foundation models. Production-grade Python. No hype, only rigour.
Six surfaces — one body of work.
The book is the spine. Around it: a primer, a reference library, a Maven cohort, the author’s GitHub, and a Substack newsletter that keeps you current as the field moves.
11 Chapters
From Slutsky and Yule to Chronos and TimesFM — eleven chapters in four parts.
478 References
Browseable bibliography — Box-Jenkins, Hyndman, Chronos, TimesFM. One-line takeaways.
Forecasting Primer
Standalone explainers on stationarity, CRPS, walk-forward CV, conformal prediction.
The course
Modern Forecasting Mastery on Maven — cohort-based, 4.8/5 from 24 reviews.
All books on Gumroad
The full catalogue — forecasting, conformal prediction, CatBoost, tabular ML, math classics.
valeriy.ai
The author hub — newsletter, courses, research, social. No hype, only rigour.
Forecasting models are 5 percent of the equation.
The other 95 percent — metrics, validation, deployment, failure modes — is where most teams quietly fail. This book starts there.
Solid foundations, not shaky assumptions
Learn how to evaluate forecasts, recognise when they’re failing, and assess the forecastability of a series before you spend three weeks building a model.
Built for understanding, not just coding
Go beyond black-box code. Grasp the mechanics and decision logic — so you can tell when a model is wrong, not just when it errors.
Production-grade Python, no throwaway scripts
Every example is fully documented and reusable. The point is not to publish notebooks — the point is to ship forecasts that hold up.
Continuously improved by readers
A living resource. Many additions came directly from reader feedback, and contributors are credited in the acknowledgments.
Comprehensive, real-world coverage
From classical models to deep learning, transformers, and foundation models. Every method tested in production or against strong academic benchmarks.
Real ROI for your work and career
Readers report immediate improvements in accuracy, interpretability, and stakeholder trust. No more silent failures or fragile production systems.
From foundations to foundation models.
Eleven chapters in four parts. Read sequentially or jump to where you’re stuck.
Take the course on Maven.
Modern Forecasting Mastery — a cohort version of the book. Build production forecasting systems with weekly assignments, live sessions, and feedback. 4.8/5 from 24 reviews.
Also running: Applied Conformal Prediction — distribution-free uncertainty quantification, 4.7/5 from 37 reviews.
5.0 / 5.0 across 34 ratings on Gumroad.
Common themes from buyer reviews:
Theory connected to practice without the academic detour.
Metrics, validation, drift, deployment — not just another model zoo.
Classical methods through transformers and foundation models, in one place.
Production-grade Python that drops into a real pipeline.
No hype. Only rigour.
Subscribe at valeriy.ai for the email list, or read the full back-catalogue on Substack and Medium.
Buyers, students, and Maven cohort attendees include data scientists and engineers from these organisations:
Also followed by academics from the University of Chicago, KTH (Sweden), UBC (Canada), and DTU (Denmark).
Valeriy Manokhin, PhD, MBA, CQF
Machine learning researcher, senior data science leader, and forecasting expert focused on probabilistic forecasting, conformal prediction, and production evaluation. PhD from Royal Holloway, University of London, supervised by Professor Vladimir Vovk — the creator of conformal prediction.
Teaches modern forecasting to data scientists, engineers, and research teams, with students and readers from organisations including Amazon, Apple, Google, Meta, BlackRock, Morgan Stanley, and Nike. His work on time series forecasting bridges academic benchmarks and the harder question practitioners face every day: which models still work when data shifts, constraints matter, and forecasts have to earn trust?
Read the full story