Ankur Trapasiya

Finance and machine learning, worked through in code.

Notes from a backend engineer learning both. Each series reads in order, and every number comes from a script you can run.

Latest

  1. States and rewards: what you tell the learner decides what it learnsNOV 11 2026
  2. Evaluating a strategy: where the return came fromNOV 9 2026
  3. Dyna: learn a model of the world, then practise in itNOV 7 2026

Series · read in order

No. 01 · 14 parts

Quantitative Finance 1

return, risk and what a mix buys you

  1. 01Quantitative finance in ten ideas
  2. 02What is quantitative finance?
  3. 03Probability: how often something happens
  4. 04Expected value: what a risky choice is worth on average
  5. 05Returns compound: multiply, never add
  6. 06Mean and standard deviation: return and risk in two numbers
  7. 07Sharpe, Sortino and information ratio: return per unit of risk
  8. 08Covariance and correlation: why a mix is safer than its parts
  9. 09Beta and alpha: how much of the market you carry, and what is left
  10. 10Portfolio optimisation: choosing the weights
  11. 11Value is an estimate: discounting, growth and the margin of safety
  12. 12Why prices are hard to beat: efficiency, random walks and a roulette wheel
  13. 13What a trade really costs
  14. 14Data that lies: the bugs that look like alpha

No. 02 · 14 parts

Machine Learning 1

from a straight line to a forest

  1. 01Machine learning for quantitative finance in ten ideas
  2. 02What machine learning is: rules learned from examples
  3. 03The data is part of the model: splits, gaps and outliers
  4. 04Regression vs classification: predicting a number or a label
  5. 05Linear regression: the first model, a straight line
  6. 06Every feature has a timestamp: leakage, scaling and twin features
  7. 07Decision trees: a model made of yes/no questions
  8. 08Nearest neighbours: predict what similar past cases did
  9. 09Bias and variance: where a model's errors come from
  10. 10Scoring a model: a number means nothing without a split, a metric and a baseline
  11. 11Bagging: many copies of a model, averaged
  12. 12Random forests: bagging with one extra rule
  13. 13Boosting: models that learn from the last one's mistakes
  14. 14Choosing a model is a systems decision