The Zone That Vanished: Why Order Blocks Leave Your Chart Without Price Ever Breaking Them

Every few weeks someone asks me a version of the same question: "Your indicator removed a zone from my chart, but price never went through it. Is that a bug?"

It is a fair question, and for a long time my answer was too short. I would say something about deduplication, the person would say "okay", and we would both move on without either of us learning anything. So I went and counted. I took the detection rules out of my zone indicator, put them in a stripped-down audit build that writes one line to the log every time a zone appears, changes state, or leaves the board, and ran it through the strategy tester on EURUSD M15 over June 2026 - one full month, 2,112 bars, modelled on 1-minute OHLC. The audit produced 2,124 lifecycle events.

The breakdown surprised me enough that I think it is worth a whole article. Of the events where a zone left the chart, price breaking through it accounted for barely more than a quarter.

A zone can leave your chart in four different ways

Most order block education treats a zone as having two possible endings: it holds, or it breaks. That framing is comfortable because both endings are about price. It is also incomplete, because a zone drawn by software is not only a market object. It is also an entry in a list, and lists have their own rules about what stays in them.

Once you separate the two, there are four distinct exits:

Broken. Price traded through the zone and was beyond its far bound when the journal looked. This is a market event. The level was tested and it failed. Nothing controversial here.

First touch. Price came back into the zone for the first time. Strictly this is a state change rather than an exit - the zone stays on the chart, it just stops being fresh. I wrote about what that transition means for the zone's remaining usefulness in Order Block Mitigation: When a Zone Is Still Worth Trading, so I will not repeat it here.

Superseded. A newer zone formed in the same direction, on the same layer, overlapping the old one. The detector keeps the newer formation and drops the older. This is a bookkeeping event with a market cause: the same imbalance got re-established at a slightly different price, and the fresher footprint is the one worth watching.

Out of the scan window. The zone is still there in the sense that the candles that formed it still exist. It is simply older than the lookback the detector scans. Nothing happened to it. The window moved.

What the census actually looked like

Across that month of EURUSD M15, the 2,124 events split like this: 900 zones appeared on the board (42%), 349 were superseded (16%), 325 registered a first touch (15%), 306 fell out of the scan window (14%), and 244 were broken by price (11%).

Strip out the births and the first touches, and you are left with 899 events where a zone actually disappeared from the chart. Of those, 39% were superseded, 34% aged out of the window, and 27% were broken.

Fewer than three in ten disappearances were price doing something. The other seven were the detector managing its own list - sometimes with a market cause behind it, but not a test of the level.

Two honest caveats before anyone quotes that number back at me. First, it is one symbol, one timeframe, one month, one set of parameters. A longer lookback shifts weight away from "out of window" mechanically. A stricter impulse filter produces fewer zones overall and therefore fewer collisions. The exact percentages are not a law of nature; the ordering is what I would expect to survive a change of settings, and I would encourage you to run your own count rather than take mine. Second, when a zone qualifies for more than one exit in the same evaluation, my classifier picks one in a fixed order - superseded first, then broken, then out of window - so a zone that was both replaced and broken is logged as superseded. That is a reporting choice, not a measurement of the market, and it is the kind of thing that quietly inflates one bucket if you do not say it out loud.

Superseded is the one that looks like cheating

I want to stay on this one for a moment, because it is where an honest tool and a dishonest tool look identical from the outside.

When two same-direction zones overlap, something has to give. If you keep both, your chart turns into a stack of nested rectangles and the word "zone" stops meaning anything. If you keep the older one, you are telling the trader that a footprint from forty bars ago matters more than one from four bars ago, which is the opposite of how order flow decays. So the detector keeps the newer one.

The trouble is what a user sees: a rectangle was on the chart, and now it is not. From the screen alone, that is indistinguishable from an indicator quietly redrawing history to make its past look better. I have no way to prove the difference with a screenshot, which is exactly why I built the journal - the drop gets logged, with its price bounds and the reason, on the bar it happened, and the indicator only ever appends to it. You can go back and check that the zone which vanished was replaced rather than erased. I went through this distinction in more detail, including the test I use to catch genuine repainting, in The Order Block Skeptic's Checklist.

If you are evaluating any zone tool - mine or anyone else's - this is the question I would ask: when a zone leaves, does the tool tell you why, in writing, at the moment it happens? A tool that only shows you the current state is asking you to trust the states it chose not to show.

Out of the window is not a signal, and treating it as one will cost you

Of the four exits, this is the one traders misread most often. A zone drops off and the mind supplies a reason: something must have invalidated it. Nothing did. The detector scans a fixed number of bars back, and the formation is now older than that.

The practical consequence runs the other way from what people expect. If you are trading a higher timeframe zone from a lower timeframe chart, a short lookback means the level you care about can silently stop being drawn while price is still perfectly capable of reacting to it. The zone did not expire. Your scan did.

That is the argument for reading the higher timeframe layer separately rather than hoping your current chart happens to reach far enough back. It is also, frankly, an argument for knowing what your lookback setting is. I would guess most people using zone indicators have never opened that input.

What this changes about how you review your own trades

Here is the part that matters more than the percentages.

When a zone you were watching disappears and you did not take the trade, you will write something in your journal about it. If your only two categories are "held" and "broke", you will file it under "broke", because that is the only category available for a zone that stopped existing. Do that for three months and your journal will tell you that your zones fail far more often than they actually do - and you will start second-guessing entries that were never actually invalidated.

The fix is not complicated. Give yourself four categories instead of two, and when a zone leaves your chart, make yourself answer which one it was. Half the time the answer will be "the software reorganised its list", and that is not a data point about your edge at all.

I do not think this requires software. You can do it with a notebook and a screenshot at the end of each session, and doing it by hand for a couple of weeks will teach you more than reading about it. But it is tedious, and tedious things stop getting done in exactly the weeks when markets are interesting.

So in the free version of Order Block Zones I made the indicator keep the record instead. Once per closed bar it compares the current board against the previous snapshot and writes what changed to the Experts log, in plain text - new on the board, first touch, superseded by a newer overlapping zone, broken, out of the scan window - each with the zone's direction and price bounds. It is quiet by design - you have to go and read the log; it does not pop up, and it does not draw anything extra. The paid version adds the higher timeframe layer and alerting on top of the same journal. The audit build that produced the counts above is this same journal with everything else stripped away, which is the only reason I could write this article with numbers instead of adjectives.

Your homework

Open your zone tool, whatever it is. Find the oldest zone still drawn on your chart, and answer two questions: how far back does this thing actually scan, and what would it do if a newer zone formed on top of that one right now?

If you cannot answer either question from the tool itself, you are reading a chart that has been edited by rules you have not seen. That is survivable - most of us trade that way for years - but it is worth knowing, and it is a much smaller problem to fix than it looks.

Next in this series I want to take the same audit build and ask a harder question: of the 325 first touches in that month, how many happened at a zone that the higher timeframe also agreed with, and did the two populations behave differently afterwards. That one needs more than a month of data to say anything responsible, so it will take a while.