The dog and the leash

André Kostolany, who traded through more currency regimes than most people read about, explained the market with a walk. A man leaves his house with his dog. He walks a mile down the road at a steady pace. The dog runs ahead two hundred yards, doubles back, stops at a tree, sprints to catch up, veers off after a squirrel. By the time they reach the end of the road, the man has walked a mile and the dog has covered four.

If you only watch the dog, its path looks like noise. If you watch the man, you know where the dog will be in ten minutes — roughly, not exactly, and not right now.

That is the whole thesis of this site. Prices are the dog. Something slower and duller — earnings power, a book value, a moving average, the historical spread between two related instruments — is the man. And the leash is whatever force keeps the two from separating permanently.

The bet, stated plainly

Mean reversion is a bet that when the distance between dog and man is unusually large, the distance shrinks. Not that price goes up. Not that the company is good. Just that the gap closes.

This is a narrower claim than most people who use the phrase realize, and the narrowness is the point. It makes the bet testable. Given a series and a definition of “the mean,” you can measure how far price is from it, in units of its own typical variation, and you can count what happened next across a few thousand historical instances. The answer is a distribution, not a promise.

That is a very different activity from looking at a chart, feeling that a stock has “fallen too far,” and buying it.

Three ways the metaphor is load-bearing

The leash has a length. A dog on a thirty-foot lead can get thirty feet away and be perfectly fine. Two standard deviations from a fifty-day mean is a normal Tuesday for a small-cap biotech and a genuine event for a utility. Every mean-reversion system needs an answer to “how far is far for this thing,” and a z-score is just the cheapest available answer.

The man moves. This is the failure mode that kills people. If you anchor on a mean that is itself drifting downward — a company whose earnings power is falling, an industry being disintermediated — then price is not stretched away from value. Price is tracking value down, and you are buying a falling knife while congratulating yourself on your discipline. Every reversion trade needs a check on whether the anchor is still where you think it is.

The leash can break. Bankruptcy, fraud, a takeover at a price below where you bought, a regime change that permanently reprices an entire sector. The distribution of outcomes has a left tail that no amount of “it’s cheap now” reasoning can argue away. This is why position sizing is not a footnote to the strategy — it is the strategy. A method with a 70% win rate and no cap on the loss size is not a method.

Why write it down

There is an enormous amount of mean-reversion content on the internet and almost none of it publishes its losers. That is not an accident. Reversion strategies have a seductive property: they win often. A system that wins 68% of the time produces long, comfortable stretches of being right, punctuated by the trade that gives back four months of gains. If you only publish during the comfortable stretches, you look like a genius indefinitely.

So the rule for this site is that a trade written up going in gets written up coming out, whatever happened. Screens get published with their construction visible so you can find the flaw in them. Backtests get published with the survivorship and look-ahead problems named, because every backtest has them and pretending otherwise is the tell.

None of this is investment advice, and I am not a registered investment adviser. I am one trader with one book, working through a loop: learn something, document it, share it. The documenting is not a side effect. It is the part that makes the learning stick, and it is the part that makes the mistakes findable later.

What comes next

The next few posts build up the toolkit from the bottom:

1. What “mean reversion” actually means statistically, and why “it went down a lot” is not it.
2. How to construct a z-score screen that does not just surface the week’s worst news.
3. The three specific ways these trades die, with the position-sizing arithmetic that survives them.

Watch the leash, not the dog.