Lesson 1 / 6 · 10 min · take your time
The output is allergic to a 1% lie
Classroom slides
- →Next
- ←Previous
- HomeSlide one
- EndLast slide
- 1–6Jump
- Swipe leftNext
- Swipe rightPrevious
- Hold PrevSlide one
1 / 6
Board 01
The output is allergic to a 1% lie
Mean-variance is greedy about expected return. Nudge one name’s guessed average up a little and the solver may dump everything into it. Nudge it down and the name can vanish, or flip to a short. That is not robustness. That is a machine taking fiction seriously.
Next time you open this lesson
We’ll open this slide when you come back. This device remembers it.
Keys move the deck. N next, P previous. Ours. Not PowerPoint.
Swipe the deck. Hold Previous for slide one. A vertical swipe still scrolls the page.
Mean-variance is greedy about expected return. Nudge one name’s guessed average up a little and the solver may dump everything into it. Nudge it down and the name can vanish, or flip to a short. That is not robustness. That is a machine taking fiction seriously.
Variances and covariances estimated on a short window are noisy. Estimated on a long window they mix regimes that no longer rhyme. Either way the table is a record plus a hope. The hope is the part a sales page skips.
In-sample is the window you fit. Out-of-sample is the next window, the one that pays rent. A mix that looks brilliant on the fit window and ordinary later is the usual classroom tragedy. Overfitting is the word: the drawing memorized a souvenir.
We will not fit your names. A is already late. A fitted weight would be later than that, and still not yours. Look. Do not buy. Not a broker. Not a planner. Not a solver. A famous paper is a memory, not a coupon. We will not size your .
Say it back below. Then the next lesson opens. If you leave the page, we will ask to save your spot.
Ask the Professor