Sitemap - 2025 - Valeriy’s Substack
Why Big Funds Lose Their Edge: The Math Behind the Claim
The KPI Machine: How Revolut Tries to Kill Corporate Politics With Metrics
Why Big Funds Almost Never Beat the Market (It’s Not Incompetence — It’s Math)
The 100-Year-Old Statistical Trick Powering Modern AI and Finance
Internal Venture Capital at Revolut: Ship 50 New Products, Kill the Bottom 25%, Scale the Winners
The Quiet Model Behind Modern Finance and Tech: Counting Frequencies, Not Sequences
The 100-Door Version of Monty Hall: Where Intuition Finally Breaks
The Monty Hall Problem, Explained the Way Probability Actually Works
What You Didn’t Study in Statistics Class
Compliance as Software: Revolut’s Most Underrated Product Might Be Its Regulation Engine
The Anti-Startup Strategy That Wins: License First, Then Launch
Before Tesla, There Was Romanov: A Russian Nobleman’s Electric Car That Could Go 60 km in 1899
From the Gaussian Integral to Black–Scholes
Revolut Killed “Move Fast and Break Things”
The STEM Race Is Won on Saturday Mornings
The Integral That Invented the Bell Curve
Why “Mean = Variance” Quietly Explains Scale in Finance and Tech
Revolut isn’t “move fast and break things.” It’s move fast and don’t break the bank. (MIPT DNA)
The MIPT Export: Why Mexico’s Next Banking Leap Is Being Built by “Foreign” Engineers
Time Is a Rounding Error We Agree To Ignore
Why the Square Root of n Refuses to Go Away
The Problem of Points: A Coin-Flip Puzzle That Explains Modern Finance
The Most Important Random Variable You’ve Probably Been Ignoring
Volatility Doesn’t Scale with Time. It Scales with Correlation.
🚀 The Forgotten Soviet AI That Kept Cosmonauts Alive
The Oxford Physics Test Is Trivial — and Unfit for Purpose
Measuring Market Roughness: A Practical Guide to the Higuchi Fractal Dimension
Same RMSE, Different Risk: Why Forecast Evaluation Keeps Failing Decisions
he Arrow of Time: Why Your Headphones Always Tangle 🎧
Why Student’s t Is Wider Than the Bell Curve
AIC vs. AICc: Understanding the Nuances in Model Selection for Time Series Forecasting
Order from Chaos: How Adding Up Enough Garbage Creates Perfection 🔔
The Rational Traitor: Why Game Theory Says You Should Betray Your Friend 🔪
🧨 The TabPFNv2 Myth Just Exploded
The IRS’s Favorite Magic Trick: How Benford’s Law Catches Cheats 🕵️♂️
The 80/20 Rule of the Universe: Why You Can’t Escape Zipf’s Law 📉
The Goldfish Memory of the Universe: Why ChatGPT and Google Don’t Care About the Past 🐠
Lyapunov’s Fourier Trick: How a 1901 idea became a superpower in tech and finance
The Casino at the End of the World: How Solitaire Built the Atomic Bomb 🎲
The Embedding Model That Thought Like a Human, Not a GPU
Byte-Pair Encoding Is Compression in Disguise
⚙️ The First Reinforcement Learners
The Law of Rare Events: Why Bad Luck Comes in Clusters
The Calibration Fallacy of Logistic Regression
Throwing Away the Truth: Why JPEG and Hedge Funds Love the Frequency Domain 🌊
Why We Add One: Laplace, Probability, and the Moon That Kept Rising
No, Schmidhuber, Linnainmaa Didn't Invent "Learning" Either
What Have You Stolen, Mr. Kalman?
The Math That Built the Iron Curtain: How Soviet AI Shielded the East Decades Before Silicon Valley
The Lost Decades of British AI — Part IV
Why Sharpe Ratio Optimization Is Mathematically Flawed in Non-Ergodic Systems
The God Fraction: How a Bell Labs Physicist Solved Greed 🎲
The Lost Decades of British AI — Part III
The Ergodicity Problem: Why the "Average" Person is Dead 💀
Khinchin’s Law: The 1928 Theorem That Makes Deep Learning Possible
The Lost Decades of British AI — Part II
The God Vector: How Eigenvalues Rank the Web and Crash the Market 🕸️
Why Nature Loves e but Finance Loves log
Boris Delone: The Mathematician Who Built Mountains, Models, and a Department
Weaponizing the Recursion: Why Wall Street & AI Obsess Over Moments 📉
The Lost Decades of British AI — Part I
The Hidden Origin of MOOCs: How MIPT Invented Online Learning in the 1970s
The "Cheat Code" Hidden Inside the Bell Curve 🔔
How Vapnik, Solomonoff, and a Handful of Visionaries Resurrected British AI After Three Lost Decades
A Gentle but Surprising Proof of Markov’s Inequality
The "Darth Vader Rule" of Probability: A Darker Path to the Mean
🎯 A Simple Trick for Computing Hit Probabilities
The Geometric Alpha: How "Path Signatures" Unlock Hidden Patterns in Time Series
The Principle of Practical Certainty
How Jacob Bernoulli Proved That Frequencies Stabilize
Move Over Damodaran: Michael Burry Wants to Teach You Valuation Properly
How Byron (1978) Turned Stone’s Wartime Accounting Into Generalized Least Squares Reconciliation
The Economist Who Invented Coherence: Richard Stone and the Birth of Forecast Reconciliation
When Charts Come Alive: The AI Revolution That's Redefining Trading
The Enron Ghost: How Lou Pai Walked Away Clean, Married a Stripper, and Became a Land Baron
The Most Important Lemma in Conformal Prediction That Almost Nobody Talks About
The Boosting Paper the World Forgot (and Why It Matters for CatBoost)
Spectral Leakage: The Hidden Distortion Inside Your FFT (and How to Fix It)
🎲 The Bernoulli Secret Behind Market Moves
Stop Hauling Around Bloated DataFrames
The “Russian” Long-Division Method for Square Roots
Bayesianism Strikes Again: How America’s Energy Forecaster Got It So Wrong
Public Time-Series Benchmarks Are Mostly Useless
🔍 Don’t Trust a Correlation Coefficient Until You Test It
📐 How the USSR Invented Regularization
🌀 The Hidden Link Between the Poisson and Exponential Distributions
🧠 The Forgotten Origin of Kernel Methods
The Statistician Who Tamed Chaos (with a Horse Kick)
Breaking Down Forecasting Ensembles: Bias, Variance, and Covariance
How Statisticians Got Prediction Wrong
The Student Who Found a Faster Way to Multiply
How One GitHub Repo Broke the Stagnation of Conformal Prediction
Would You Trust This Calculator With Your Life?
⛏️ When Soviet AI Struck Gold (Literally)
🚀 The Forgotten Soviet AI That Kept Cosmonauts Alive
The Statistician Who Hunted Red Submarines
🪙 The Forgotten Soviet AI That Actually Found Gold
🚩 12 Red Flags in Forecasting Hires (2025 Edition)
📊 Why Your Variance Estimates Are Probably Wrong (Thanks to Dependence)
The Calibration Trap: Why Your Perfect Model Fails in the Real World
Taming Chaos: The Statistical Safety Net You Need to Know
Why the Gaussian Distribution Naturally Emerges from the Maximum Entropy Principle
Don’t Assume What You Don’t Know: The Genius of the Maximum Entropy Principle
The Forgotten Precursor to Shannon: Hartley Entropy
ROC AUC Is Overrated—Here’s Why You Should Think Twice
The Calibration Fallacy of Logistic Regression
🎲 From Lotteries to Information Theory: The Hidden Origins of Conformal Prediction
A Mathematician, a Poet, and the First Language Model
The Pólya Urn and Path Dependence in Asset Prices
CatBoost vs XGBoost: Busting the Kaggle Fairytale
🍵 From Tea Parties to Permutation Inference: How Fisher Turned Afternoon Tea into Statistics History
🕰️ Time-Traveling Through AI’s History: The 1928 Idea That Predicted Our AI Future
📉 Why Z-Score Normalization Is the Unsung Hero of Time Series Transformers
From Shannon to Deep Learning: Why Fewer Neurons Can Be Smarter
📉 Tired of Whiplash from Constant Forecast Changes?
🔄 The Natural Connection Between Complexity, Information Theory, and Conformal Prediction
🎲 Estimating Binomial Coefficients with Entropy: A Beautiful Shortcut
How My Neural Network Learned to Predict Equipment Failures — One Shapelet at a Time
Forecasting in an Uncertain World — A Realist’s Guide
The Practical Playbook — Forecasting in a Non-Stationary World
📉 Why Transformers Still Struggle with Time Series Forecasting
🧠 Why Are Extrema Associated With Fast Oscillations?
Transformers vs. Time — Deep Learning’s New Battle with Non-Stationarity
Sequence-Savvy Models — How RNNs and LSTMs Learn Without i.i.d.
Kolmogorov's Quiet Revolution: The Mathematical Roots of Long Memory in Time Series
Can Decision Trees Forecast the Future?
When Theory Bends — Learning Without i.i.d.
Post 1: The Myth of i.i.d. — Why Time Series Break the Rules
📈 Understanding Market Memory: The Hurst Exponent and S&P 500 Volatility
If You Think AutoARIMA Is the Same as AutoML…
Bernoulli’s Law, Leibniz’s Warning
🧠 Why No Program Can Fully Measure Complexity
📊 Are You Misreading Your ACF Plot?
STEM BSc in the West? Think Again. The AI Frontier Doesn’t Wait.
🧠 Can ChatGPT Really Forecast? The Illusion of AI Time Series Modeling
🧠 LLMs Are Stochastic—Even When You Think They're Not
🥇 Why CatBoost Wins (and the No Free Lunch Theorem Doesn’t Mind)
Forecasting, VC Dimension, and the Hidden Bias That Makes Models Work
Boosted Trees: The Workhorse of Forecasting—Until They Fail
📈 From Obscurity to Adoption: The Journey of Conformal Prediction in Forecasting
📚 The True “Founding Fathers” of Machine Learning
📐 “Is Markov in the Room With Us?” — A Friendly Guide to Markov’s Inequality
📉 The Coefficient of Variation: A Misleading Measure of Forecastability
🎯 The Litmus Test for Hiring Data Scientists in Forecasting
⚽ Brazil Scores Against the Cargo Cult of Data Science
🔎 Making CLIP Trustworthy: Conformal Prediction Meets Zero-Shot Vision Models