Levels From Gaps

There is a hole in the chart. Yesterday the instrument closed at one price, this morning it opened at another, and between the two nothing traded at all.

Everyone tells you to draw lines on that hole. The standard teaching names three of them and calls them very strong levels. I marked those three prices on four hundred and seventy six gaps and measured what price did when it came back. It is not what the lines promise. And there is something else there instead, which is more useful. By the end you'll know which price in a gap to mark, and what price tends to do there instead of stopping.

Start with what a gap is, because it is not a fast move. It is an absence. There are prices in that space at which no transaction happened — not few, none — because the market was closed while the reason to reprice arrived.

They are ordinary. On the three instruments I measured, thirty seven per cent of sessions opened outside the previous day's range. More than a third of all days. If a gap by itself meant something, then a third of all days would mean something.

Most of them are nothing. The median gap is about a fifth of a daily range, and three quarters of them are under half of one — small enough that you would not notice without measuring.

So there has to be a threshold, or you are studying noise. Mine is half of the average daily range over the previous twenty sessions: how far this instrument normally travels in a day. Above that line a gap is an event. Eight per cent of sessions — four hundred and seventy six of them in eight years.

Where does the hole come from. Overnight something arrives — earnings, a takeover, a number from a central bank — and the price people are willing to pay is no longer the price on yesterday's screen.

Orders collect while the market is closed. At the open they are matched all at once, in a single auction, and that auction clears at whichever price balances the buying against the selling. If far more of them wanted to buy, that price sits above yesterday's high. Nothing traded in between, because there was no moment in between.

The older explanation puts a person there — a specialist on the floor of the exchange, holding the imbalance and choosing where to open the stock. That was accurate once, and the accounts that describe it say plainly that the role is now essentially gone.

The mechanism did not leave with the person; it got automated. The imbalance is still real and still cleared at one price. Nobody decides it. That matters for what you take away. A gap is arithmetic performed on a pile of orders, not a message from someone who knew something.

Now the claim, and it is stated precisely, which is good — precise claims can be checked. A gap is said to leave three levels. One: the close before the gap, the last price at which business was done. Two: the open of the gap candle, the price the auction chose. Three: the extreme of that candle — its low if the gap was down, its high if the gap was up. All three are described as very strong.

And each one comes with a story, and the stories are reasonable. At the close before the gap sit people who were holding and did not get out in time. When price comes back to where they bought, they sell to get even, and that supply stops the advance.

At the open, a large seller met a large buyer, so a great deal of business happened at one price. At the extreme, the move stopped once already, so it can stop there again. That is a coherent set of mechanisms. It is also exactly the sort of thing that sounds true and can be counted.

Here is one, so the three prices are not abstract. Apple, the fifth of August, twenty twenty four. The session before closed at two hundred and nineteen dollars and eighty six cents, and its lowest trade was two hundred and seventeen seventy one. The next morning it opened at one hundred and ninety nine dollars and nine cents.

In between — eighteen dollars and sixty two cents of price — nothing traded. The average daily range going in was four dollars ninety six, so the hole alone was three and three quarters of a normal day.

Four hundred and seventy six gaps like that one. Three instruments, eight years of daily bars. On each I marked the three prices, waited for price to come back, and measured what happened when it arrived.

That needs something to compare against or the number means nothing. Not a line drawn at random — I will come to why. What I used is a day that moved just as much and did not gap: it opened inside the previous day's range and closed the same distance away. Same three prices, same measurement, four hundred and forty eight of them.

Why not a random line. Because a random line drawn on the same chart gets touched twenty-three sessions later, when nothing in particular is happening. A gap edge gets touched two sessions later, in the middle of the reaction to the biggest move in weeks.

Those are not the same test. One asks whether a price holds in a quiet market. The other asks whether it holds inside a shock. The matched day with no gap gets touched on the same schedule the gap does, one to three sessions. That is the whole point of matching it, and it is why these numbers took the time they took.

The measure itself is deliberately blunt. Wait for the first session whose range touches the level. Then look at that session and the five after it, and split what price did into two parts: how far it went past the level, and how far it came back off it.

Take the share that came back. A half means the level did nothing at all — price went through it as readily as it turned away from it. More than a half means it held. Less than a half means price was more likely to keep going.

The close before the gap came out at zero point four seven. The open of the gap candle, fifty two. The extreme of the gap candle, forty five. The matched days with no gap in them gave forty nine, fifty three, and fifty. So the three very strong levels landed on a coin flip. And on all three, the ordinary day with no hole in it did slightly better than the gap did.

Six sessions is more than a week, and the fair objection is that the claim is about a local stop. So here is the tighter version. On the day price first reaches the level — did it stop there? I counted a stop as going less than a quarter of a daily range past the line and closing back on the side it came from.

The close before the gap: twenty seven per cent. The open: twenty seven. The extreme: twenty four. The matched days with no gap: thirty two, thirty two and twenty eight. The gap is behind on all three.

One more, because the obvious objection is that the wrong price was named. The true edge of the empty zone is not the close — it is yesterday's low. Two hundred and seventeen seventy one in the Apple example, the lowest price anyone traded at before the hole.

That one comes out at exactly zero point five zero, and its matched control comes out at fifty as well. So this is not a case of the usual teaching picking the wrong line out of four. There is no line of this kind, on this evidence, that holds.

Which would be the end of a fairly negative video, except that something else turned up while I was counting. Take the gaps that did not close on the first day — eighty per cent of them — and stand them next to the matched no-gap days at the end of that day. Both sit the same distance from the old close: one and two tenths of a daily range. The arithmetic of the first day is now out of the way. From there the gap gets back faster. Six sessions against nine. Within sixty sessions, seventy nine per cent against seventy one.

Now, how much weight that carries, because I would rather you trusted the small print than the headline. It is a small difference: within twenty sessions, sixty two per cent against fifty seven. That sits inside what a sample this size can actually resolve, so I would not call it established and I would not build an entry on it.

But it points the same way at every horizon I looked at, and the direction is the interesting part. It is the opposite of what is usually taught. The gap does not stop price when price arrives. If it does anything at all, it makes price arrive.

Back to Apple, because that is exactly what happened. Price came back to two hundred and nineteen eighty six on the thirteenth of August, six sessions after the gap — the level where the trapped holders were supposed to be selling to get even.

It went straight through and kept going: two hundred and twenty seven and seventeen cents within six sessions. Seven dollars and thirty one cents past the line, one and two tenths of a daily range. And that is not a cherry-picked failure. The median across the four hundred and thirty five I measured is one and four tenths. This one held up better than average.

So what do you do with a gap. Treat it as a destination, not a wall. If you are already long and an unfilled gap edge sits above you, that is a defensible place to take profit: price is likelier to reach it than to be turned by it. If you are thinking of selling into it because it will hold, four hundred and seventy six events say it will not.

And if you find a gap edge that does hold, look at what else is at that price. Usually it is a consolidation, and the consolidation is doing the work — that is the video before this one.

Two limits, and they are real. Daily bars are blind to the middle of the day. A level that stopped price for two hours and failed by the close appears nowhere in these numbers, and much of the usual teaching is aimed at intraday work. I measured whether a gap edge holds over sessions, not over minutes.

And gaps cluster around earnings. Four hundred and seventy six events on three instruments in one stretch is not the market. It is enough to stop drawing three lines and expecting them to hold. It is not enough to say nothing is ever there.

So, the thing to go and do. Take an instrument you follow and find its last gap bigger than half its daily range. Mark one price only — the close before it. Then write down two things: the date price came back, and how far past it went. Ten of those and you have your own version of these numbers, on your own instrument.

And keep the practice the usual teaching ends on, because that part is right. Draw levels on a hundred charts — just write down what you expect before you look, and check at the end of the month. An eye trained without a record does not get better. It only gets more confident. Educational content only. Nothing here is financial advice.