Reference

Chart Tools and What They Assume

Every measure in the indicator library is computed from the data. Every tool in this section asks the analyst to choose something first (a swing, two points, a scale), and that choice ends up inside the answer.

A platform’s drawing menu and its indicator menu look like two versions of the same thing. They are not. An indicator takes a series and a period and returns a number; a drawing tool takes a series and a decision and returns a number that depends on both.

That distinction decides what each can support. Two people who disagree about an RSI reading have an arithmetic dispute with an answer. Two people who disagree about a trendline have a dispute about which points matter, which no amount of data settles, and the tool will draw either line with equal confidence.

How much of each tool's output the analyst choosesA diverging bar chart ranking six tools by how much of their output is fixed by the data against how much is chosen by the analyst. Average true range, pivot points and moving averages sit on the data side; Fibonacci levels and regression channels are in the middle; trendlines sit furthest on the discretion side.← FIXED BY THE DATA · CHOSEN BY THE ANALYST →Average true rangedata onlyPivot pointsone sessionMoving averageone periodFibonacci levelsa chosen swingRegression channela chosen windowTrendlinestwo chosen pointsHow much of each tool's output the analyst choosesA diverging bar chart ranking six tools by how much of their output is fixed by the data against how much is chosen by the analyst. Average true range, pivot points and moving averages sit on the data side; Fibonacci levels and regression channels are in the middle; trendlines sit furthest on the discretion side.← FIXED BY THE DATA · CHOSEN BY THE ANALYST →Average true rangedata onlyPivot pointsone sessionMoving averageone periodFibonacci levelsa chosen swingRegression channela chosen windowTrendlinestwo chosen points
Fig. 1: ordinal, deliberately not numberedAn ordering rather than a measurement: a percentage here would be quoted as though it had been measured, and it has not. What the ordering says is that the left-hand tools take a series and a period, and the right-hand ones take a series and a judgement. Neither end is better, but only the left-hand end produces a figure that two people can be required to agree about, and only the right-hand end can be made to fit whatever the analyst already believes.

The self-fulfilling part is the real part

The strongest defence of a popular level is not that the arithmetic predicts anything. It is that a great many participants are watching the same number, and orders accumulate where people are watching.

That makes the level a genuine feature of the order book rather than of the price history, supply and demand really do change there, because a crowd agreed in advance that they would. It is a circular mechanism and it is not a weak one: knowing where the widely published levels sit is useful precisely because other people are using them. What it does not license is the further claim that the ratio or the geometry discovered something.

The tool this section deliberately does not cover

Chart patterns (the named shapes, with necks and shoulders and flags) are the most popular drawing tools of all, and there is no page for them here. The reason is not distaste; it is that they fail the standard the rest of this reference is held to, and failing it visibly is more useful than a page pretending otherwise.

A pattern has no formula, no stated identification rule, and no count of the occasions the same shape appeared and nothing followed. Everything on this site is either arithmetic on published data or a clearly labelled illustration, and each page states what would change its mind. A shape identified by eye after the outcome is known cannot meet that: two analysts will disagree about whether the pattern is present, which means the claim cannot be checked even in principle.

What can be salvaged is the part that is measurable. Most patterns are descriptions of range compression followed by expansion, and that is a real property with a documented mechanism, the range page covers it, including the finding that range is autocorrelated while direction is not. Read that way, "the pattern is forming" becomes "the range has been compressing", which is checkable, and the accompanying claim about which way the expansion will go turns out to be the part that was never supported.

Two habits that make a drawn level defensible

Write the rule down before you draw. "The high and low of the last sixty sessions" is a rule; "the obvious swing" is a decision you will make differently after the fact. Once the rule is mechanical, the tool can be tested like any indicator, and the honest form of the tool is the one that can be tested.

State the scale. A straight line on a linear chart is a constant number of points per session; on a log chart it is a constant percentage. Across a period where price doubled those are different claims, and a line drawn on one scale does not exist on the other. Without the scale stated, the level is not reproducible even by the person who drew it.

Levels drawn from a chosen swing

Both of these project support and resistance from a move the analyst selects. The arithmetic is trivial and reproducible; the selection is neither, and it is where two people looking at one chart end up with different levels.

  • Fibonacci retracementsWhere the ratios come from, why 50 per cent is not one of them, and what happens to a level that a great many people are watching.
  • Fibonacci calculatorThe retracements and extensions computed from any swing, in either direction, with every step shown.

Lines fitted to price

A line through a chart is a claim about a relationship. One of these requires you to pick the two points it passes through; the other fits itself to a window you pick instead.

  • TrendlinesFixed by a choice of two points and a choice of scale, and both choices change what the line says.
  • Displaced moving averageA line that needs no points chosen: the average shifted forward, which is what a trend channel approximates.

Levels computed from the data alone

For contrast, the tools that leave the analyst nothing to choose but a period. If two people disagree about one of these, one of them has made an arithmetic error, which is not true of anything in the first two groups.

  • Pivot pointsSupport and resistance from one completed session’s high, low and close. Four published variants, all reproducible.
  • Average True RangeHow far the instrument has been moving in its own units, the measure that makes every other distance on the chart comparable.
  • Bollinger BandsBands at a stated number of standard deviations, which is a computed envelope rather than a drawn one.

Frequently asked questions

What separates these from the indicator library?

Discretion. Every measure in the indicator library is computed from the price and volume series, feed two people the same data and the same period and they get the same number. The tools here require a choice: which swing, which two points, which window, which scale. That choice is part of the output, and it is why two analysts can draw incompatible conclusions from one chart without either making a mistake.

Does that make them useless?

No, and it is worth being precise instead of dismissive. A drawn level is a way of recording a decision about where you think a market changed character, that is a legitimate use, and having the decision written down beforehand is better than making it in the moment. What discretion rules out is a claim that the tool discovered something in the data, because the analyst put a good deal of it in.

Which of these has the strongest case?

The computed ones, and among the drawn ones the honest answer is the horizontal level. A price at which a market has repeatedly turned is a fact about the chart that two people will usually identify similarly, and there is a mechanism behind it: orders accumulate at round numbers and at previous extremes. Sloping lines and ratio projections require more choices and have thinner evidence.

Why do levels sometimes seem to work?

Partly because enough participants watch the same ones. If a large number of people place orders at a widely published level, the level becomes a place where supply and demand genuinely change, not because the arithmetic predicted anything, but because the arithmetic told everyone where to stand. That is a real mechanism, it is self-fulfilling rather than predictive, and it is the strongest argument for knowing where the popular levels are.

Should I use a log or a linear scale?

Log for anything spanning a large price change, and the choice is not cosmetic: a straight line on a linear chart is a constant number of points per session, while on a log chart it is a constant percentage. Over a period where price doubled, those are entirely different claims, and a trendline drawn on one scale will not exist on the other. State which scale you used, or the line is not reproducible.

How would a drawn level be tested?

By fixing the drawing rule in advance, "the highest high and the lowest low of the last sixty sessions", not "the obvious swing", and then measuring what happened at the projected levels across a long sample, counting the failures alongside the successes. Once the rule is mechanical the tool has become an indicator and can be tested like one. Most of the published evidence for these tools skips that step, which is why it consists of examples rather than results.

Do these tools work on volume?

Some of them, and one is genuinely useful. A horizontal level drawn at a price where a great deal of volume traded is the same idea as a volume-by-price profile, which computes it from the data rather than asking you to eyeball it. Sloping lines on a volume series are hard to justify, because volume has no trend to speak of. It oscillates around a level, so a line implying growth is describing the instrument’s liquidity changing rather than anything about the market.

What does this site recommend?

Prefer the computed tools where one exists, because they are reproducible and can be argued about productively. Where you use a drawn one, write down the rule you used to draw it — which swing, which points, which scale — so that the level can be checked later by you or by anyone else. That single habit is the difference between a drawing tool and a decoration.

Where to go next