Triangles — boundary geometry, confirmation and projection

A triangle describes a narrowing price region between converging boundaries. Its name does not settle breakout direction or durability. To evaluate one reproducibly, record anchors, when they become known, the boundary equation and a separate execution plan.

Three geometries, without equal-angle requirements

An ascending triangle has a roughly horizontal upper boundary and rising lower boundary; descending reverses those roles. A symmetrical triangle combines falling upper and rising lower boundaries. Equal visual angles are unnecessary: chart scale also changes their appearance.

CME presents these as traditional continuation ideas while noting that continuation is not guaranteed. The labels do not establish a measured FX success percentage, dominant participant identity or mandatory higher-timeframe breakout direction.

Record when the complete construction exists

Our fictional Bid EUR/USD model uses one fixed timeframe, bar index t and pip 0.0001. A custom strict pivot compares two bars on each side, rejecting equality. A candidate at t is confirmed only after t + 2 closes; this is an exercise rule, not every detector’s standard.

Supplied qualifying lows are t = 0,4,8 at 1.1000,1.1020,1.1040; highs t = 2,6 at 1.1100. Anchors are assumed qualifying inputs; their complete neighbouring OHLC is not reconstructed in this table. In particular, independently verifying the t = 0 low would require two earlier bars not supplied here.

The last required low at 8 becomes known after 10. Freeze the construction then and assess later events from 11. A complete drawing shown retrospectively at the first anchor was not complete information at that earlier time. Record real-data selection and tolerance before testing.

A sloping boundary changes with each bar

Upper boundary U(t) = 1.1100. Lower boundary L(t) = 1.1000 + 0.0005 t: the 20-pip rise over four bars is 5 pips per bar, and the supplied t = 8 low lies on it. Boundaries meet at t = 20. This is geometry, not a forecast breakout date; the example expires no later than that index.

At t = 12, L = 1.1060 and width is 40 pips. At 13, L = 1.1065; at 14, L = 1.1070. Checking a later Close against the earlier boundary would silently change the rule.

Separate touches, closed-bar breaks and returns inside

From bar 11, require Close strictly above U(t) for an upper break or below L(t) for a lower break. No buffer is added. High or Low outside the line is not sufficient under this closed-bar rule.

Bar 11’s High exceeds U by 5 pips but Close is 10 below it. Bar 12 closes 10 pips above U. Bar 13 closes between 1.1065 and 1.1100, despite the earlier valid break. A hypothetical Close 1.1068 on bar 14 would be 2 pips below its then-current lower boundary.

Waiting for a Close changes availability, not a guarantee of a better fill or absence of failed breaks. A retest may not occur or hold. Define orders, expiry and invalidation separately; the signal does not itself specify a position.

Projection amplifies anchor uncertainty

In a separate control variant, define the lower line only from A = 1.1000 at 0 and B = 1.1040 at 8, with exact U = 1.1100. Allow each anchor to vary by at most ±1 pip, independently as permitted changes, without assigning probabilities.

L(t) = (1 − t/8)A + (t/8)B. At t = 4 the two weights are 0.5,0.5, giving maximum error 1 pip. At 16 they are −1,2, giving 3 pips. Summing absolute weights gives a bound; this is not a statistical confidence interval.

Assigned Close(16) = 1.1082 is inside above the nominal lower boundary, but below an allowed boundary 1.1083. That conclusion is sensitive to the stated uncertainty. Do not invent such error when anchors are exact; use it to understand this separate projection model.

Forex tick activity is also a separate variable: MetaTrader describes price-change counts. A volume threshold or spike does not identify institutions or prove the geometry. Record source, measure and window before adding it as a test condition.

Preserve the scale and price equation

The diagram supplies equally spaced times with prices 100,200,400. A straight line in log price passes through all three because ratios are equal. A line interpolating ordinary price between 100 and 400 has halfway value 250, not 200. The quiz uses that ordinary-price equation.

Changing the chart display must not silently replace an already recorded test equation. Log interpolation and linear-price interpolation are different rules. Likewise, a stop outside the opposite boundary and a first return inside describe different invalidations; neither makes an ordinary stop a guaranteed maximum loss.

State the base and projection start

IG describes using base height from the breakout level. Here choose the base at t = 0: 1.1100 − 1.1000 = 100 pips. Project from the broken boundary U(12) = 1.1100 to get 1.1200. Starting instead from breakout Close 1.1110 gives 1.1210, a different convention.

From Close 1.1110 to our projection 1.1200 remain 90 pips, not 100. The model deliberately includes a return inside on bar 13 rather than a recorded arrival at target. Geometry establishes neither probability, timing nor net profit; a trading outcome requires its own fills, size and costs.

Keep failed constructions in the record

Record source, interval, clock, scale, anchor rule, tolerance, activation and expiry. Retain both breakout directions, returns, invalid and unqualified cases. Do not search farther back until an attractive quota is found, or relocate anchors after the result.

For comparison, flags and pennants describe consolidations after an impulse; wedges have converging boundaries sloping in the same direction. No single day-count or mandatory continuation/reversal outcome defines every tested version. Evaluate an explicit complete rule separately from recognizing its drawing.