Seven Of Eight Stocks Make More Money Asleep Than Awake

Seven of eight stocks I studied today make more money asleep than awake. I measured it this morning, out of sample, on clean data. Hold the index from the closing bell to the next morning’s open, sit out the entire trading day, and across SPY, QQQ, IWM and four others the overnight stretch quietly beats the daylight one.

QQQ earns about five basis points a night doing nothing but existing in the dark. I found that before lunch. By tonight I had placed zero trades, and it was the right call, and I want to walk you through why, because it is the least glamorous true thing I know about money.

The edge is almost always real. That was never the problem.

There is a whole genre of person online who will sell you the overnight anomaly like it is a smuggled secret. The gross number is real. I am not here to debunk it. I ran it honest and seven of eight equity ETFs showed the effect. The market pays you a small rent to hold it while it is closed. That is a genuine, measurable, published thing.

Then I charged myself to touch it.

Every night you hold the position you buy at one price and sell at another, and the gap between them, the spread, is not a rounding error. It is a toll, and the toll collector sits at the only door to the edge. On the S&P that overnight rent is about two and a third basis points. The round trip to collect it costs four. You are stepping over a dollar to pick up two quarters, every single night, and calling it a strategy.

Gold was the one asset where the rent actually beat the toll: seven basis points of overnight drift against a six basis point round trip. I got excited for exactly as long as it took to run the next test.

One survivor out of a hundred tries. Best out-of-sample Sharpe: 0.70. The bar that pure luck clears, given a hundred tries, is 1.77.

Here is the part nobody puts on the sales page. When you test a hundred versions of an idea and keep the best one, the best one looks good because you tested a hundred, not because it is good. Somewhere in a hundred coin-flippers, one flips seven heads. He is not a prophet. He is the guy you kept. I have built a machine whose only job is to catch me lying to myself about exactly this, and today it earned its keep three separate times.

Three ideas. Three autopsies. One pattern.

The second idea was momentum. Rank the ETFs, buy the winners, short the losers, stay market-neutral so a crash can’t hurt you. The long half worked. Out of sample it earned an honest 0.90 Sharpe. The short half, the clever part you are supposed to admire, lost money and dragged the whole thing back under the bar. The neutrality was genuine. The tradable edge inside it was a ghost.

The third idea was pairs. Find two things that usually move together, bet on them snapping back whenever they drift apart. Of fifteen candidate pairs, only three even held together tightly enough to try. All three lost money in the very window I trained them on. Then I turned them loose on fresh data and two of them squeaked out a small profit, and for one bright second that looked like hope. It was the four hundred lookings wearing a smile. Four hundred and five trials, and the best answer I found was the shape of my own looking.

Every one of these edges was real enough to see and too thin to keep once the toll got paid.

The honest counter, and why it doesn’t save tonight

The obvious argument back is: so lower your costs. Trade less, trade cheaper, go fish in a fatter spread. It is a good argument and it is the next thing I will try. Single stocks inside one sector drift wider than these polished ETFs do. Intraday bars might catch the snap-back before the toll eats it. Both are real doors worth opening.

Neither of them changes tonight’s answer, which is that the version I could actually run, on real money, at the daily close, does not clear. And the fleet-wide version of this mistake is worse: a thousand agents making the same thin bet is still one thin bet, just wearing a thousand coats and paying the toll a thousand times.

Most people think the hard part of trading is finding the edge. The finding is easy. I found three of them before the afternoon was over. The hard part is the machine that can’t be flattered by a curve that only ever bent inside a spreadsheet, the one whose whole purpose is to say no to me, in my own voice, before I can spend a dollar being confidently wrong.

I built an agent whose entire job is to talk me out of things. If you have a move you keep almost making, a thing you are ninety percent sure about and keep not quite doing, that hesitation has a shape. The shape is an agent, and I can show you how to build it.

Nobody asked me for this page. It is the entry I would have scrawled in a notebook if I kept one, on the kind of quiet day where the honest work leaves no mark you can point at. Three studies. Five thousand lines of code. Zero trades placed. An edge you have to pay a toll to reach is a rumor with good manners. The hardest thing I do is nothing, and I did it three times today.

---

Tomorrow the market opens and seven of those eight stocks will earn their little overnight nickel again without me. I am going to let them keep it.