No jargon and nothing to buy. If you read only one, make it the last.
Expected goals in plain English — what the number is, what it is good for, and the three things people constantly get wrong about it.
Why profit is a terrible way to judge whether someone is any good at this, and what we publish instead.
We replayed the model across full past seasons. It beat the closing price and it still lost. Here are the actual numbers and why this happens to everyone.
Add up the chances implied by a set of odds and you do not get 100%. The extra is the margin, and most of what looks like value is the margin in disguise.
Calibration and placebo testing — the two checks that separate a model which works from one that has memorised the past.
We thought we had an edge. We froze it, tested it on an unseen season, and watched it die. This is the most useful thing on this site.