Swing highs and swing lows

A swing high is a comparison, not a judgement 

On any market and any timeframe, a swing high is a bar whose high sits above the high of a stated number of bars on each side of it. A swing low is a bar whose low sits below the low of a stated number of bars on each side. Nothing else enters the definition. The size of the bar does not matter, nor its colour, nor whether it closed above or below its open, and no reading of what the market was doing at the time is involved. One bar is compared with its neighbours, and either the comparison holds or it does not.

That flatness is worth insisting on, because the vocabulary around swings is loose almost everywhere it is used. Commentary speaks of a top, a peak, a reaction high or a pivot and means roughly this object each time without stating the test it passed. The mechanical version has a test, and on the same data it returns the same answer whether a person, a script or a charting tool applies it.

The number of bars is part of the definition 

The phrase a stated number of bars is carrying the weight in both definitions above, because nothing in the market fixes that number. It is a setting, chosen by whoever is looking. A swing found with one bar of confirmation on each side is one object, a swing found with three bars on each side is another, and a single bar can satisfy the first test and fail the second on exactly the same chart.

The trade-off is easy to state and impossible to escape. A small confirmation number finds many swings, including small ones nested inside a larger move, and marks each soon after it prints. A large number finds few swings, keeps only those that stood out from a wide neighbourhood, and cannot mark any of them until the required bars have closed on the far side. These are not a better and a worse version of one measurement. They are measurements of different things.

Two things fall out of that table. The count of swings changed without a single price changing, which is why a statement about swings that omits its confirmation setting is incomplete rather than merely imprecise. And the wider setting introduced a delay: the tallest bar in the sequence could not be labelled until three further bars had closed. Swings are identified in arrears, always.

Cases where the comparison is not clean 

Three situations make the mechanical test awkward, and each of them has more than one accepted handling, which means charting tools can disagree about the same bar.

• Equal highs. When two bars share a high to the last decimal, neither is strictly above the other, so a strict test finds no swing at all. Some tools take the earlier bar, some the later one, and some relax the test to higher than or equal to, which reports both. The difference between the three answers is a convention, not a fact about the market.
• Inside bars. A bar whose entire range sits within the previous bar's range has neither a higher high nor a lower low. Under a strict test it can never be a swing, and a run of them can also stop the bars around it from qualifying, so a quiet stretch of chart can contain no swings whatsoever.
• Gaps and session breaks. Where an instrument stops trading and reopens away from its previous close, the bars either side of the break belong to different sessions. A swing found across the break stays arithmetically valid while comparing two separate periods of trading rather than one continuous move.

There is a further split over which price a swing is measured on. The common version compares highs and lows, the extremes actually reached during a bar. Another tradition compares closing prices only, on the reasoning that a close is the one price at which trading in that period came to rest, while a high can represent a single transaction. Both are in wide use, and they mark different bars on identical data.

Timeframe changes the answer as well 

A swing is defined on the bars of the chart it was found on, and bars are a product of the timeframe. A turning point that is unmistakable on a five minute chart may sit entirely inside a single hourly bar, in which case the hourly chart records no swing there at all. A swing on a daily chart, conversely, is assembled from bars containing dozens of smaller swings the daily chart never carries. Neither chart is the more accurate view of one event, because the two are not measuring the same event.

So a swing carries two labels before the word means anything specific: the timeframe it was found on, and the confirmation setting used to find it. Stated without both, it names no particular object on any particular chart.

What the rest of the module is built on 

Swings come this early because nearly every later construction takes them as its input rather than reading prices directly. Describing an uptrend as a sequence of higher highs and higher lows, or a downtrend as lower highs and lower lows, is a statement about swing points and nothing else, so changing the confirmation setting can change the sequence itself. A trend line needs two swing points of the same kind before it can be drawn and a third before it can be tested. Levels are conventionally located where earlier swings clustered. Chart patterns are described almost entirely as arrangements of swings, a double top being two swing highs at a comparable level with a swing low between them.

That dependency runs one way only, and it is easy to lose sight of once the drawing starts. Anything built on swings inherits every ambiguity the swing definition carried into it, the retrospective delay and the sensitivity to the setting included. A trend line drawn through swings found with one bar of confirmation is a different line from the one drawn through swings found with three, on the same instrument over the same period, and no version of the drawing is free of that choice.

Where practitioners disagree 

The disagreement is not about what a swing is for. It is about what counts as one, and it has produced three families of definition that sit side by side in common use. The bar comparison described in this lesson, usually implemented as a fractal or pivot indicator, is the most literal. A second family ignores bar counts and uses a minimum move instead, registering a swing only once price has retraced by a stated percentage or a stated multiple of recent range, which is how a zigzag style tool is built. A third works from closing prices alone.

Each family has a standing complaint about the others. Bar counting is called arbitrary, because a count of neighbouring bars has no relationship to how far price travelled. Move based definitions are called equally arbitrary in their threshold and slower still, since nothing registers until the reversal has already run. Closing price definitions are said to discard prices at which the market genuinely traded. None of these objections is likely to be settled, because the three are answering slightly different questions. What follows is narrow and firm: the swings visible on a chart are a product of the tool that found them, and two people looking at one instrument can be looking at different swings without either having made an error.