Double Tops and Bottoms: Tolerance and Failure

Define comparable rather than perfectly equal

A double top commonly has two peaks near one another in price, separated by a lower trough. A double bottom reverses the arrangement. TradingView documents a double-top detector and its pattern rules. The permissible difference between the extremes, required spacing and intervening retracement depend on the selected convention.

Exact equality is often too restrictive, but unlimited tolerance is not useful either. State tolerance in ticks, a percentage or another declared unit. Do not increase it only when that makes a later reversal fit the desired label.

An original top candidate

Suppose a completed swing high is 150, price falls to 140, and a second confirmed swing high is 149.50. With a predeclared one-unit peak tolerance and a minimum eight-unit intervening decline, the example qualifies as a double-top candidate.

If the confirmation rule requires a completed close below the intervening low of 140, a bar with low 139.50 and close 141 has not confirmed it. A subsequent close of 139 satisfies that particular rule. A rule based on any traded price below 140 would have acted earlier and must be tested as a separate specification.

Now imagine the next closes are 143, 147 and 151. The earlier condition occurred, but the expected sustained decline did not. The test must retain this failed break and whatever entry, exit and costs its rules generated. Calling it “not a real double top” after price rises removes precisely the evidence needed to evaluate the idea.

When is a third high a triple-top candidate?

A triple top adds one comparable high and a second intervening low. In TradingView’s triple-top detector, the neckline runs through the lower intervening low and a later close below it marks the formed pattern. That is one detector’s convention, not a universal setting. Selecting another extreme, another trough, a time window and an overlap policy adds degrees of freedom; fixing them after a break can make a historical sample look cleaner. Decide whether an overlapping double candidate remains open when the third high appears, and count any shared break as one price event rather than two independent wins.

Here is a separate, invented example using completed bars of one feed and timeframe. Require each swing high to have two later bars with lower highs before it becomes knowable, all three highs to fit within a 1.00-unit band, each intervening drop from adjacent highs to reach at least 8.00 units, and the third peak to appear within twenty bars of the first. These are teaching thresholds, not tested or recommended trading parameters.

Keep the observation time separate from the charted pivot.

Bar	Selected price	Recorded interpretation
2 and 5	High 150.00; low 140.00	First high and first trough
8 and 11	High 149.50; low 141.00	A two-high candidate after the second high is confirmed
14, confirmed after 16	High 150.20	Three-high candidate if no earlier boundary break occurred
17	Low 139.60; close 140.40	Wick crosses 140.00; close does not
18	Close 139.80	First close below the 140.00 neckline

For the two-right-bar test, set the following highs in the invented sequence: bars 3–4 at 149.40 and 148.50, bars 9–10 at 148.80 and 148.20, and bars 15–16 at 149.10 and 148.70. Assume no unlisted bar creates another eligible peak and every close through bar 16 remains above 140.00. The highs at bars 2, 8 and 14 then become knowable after bars 4, 10 and 16, respectively.

The high range is 150.20 − 149.50 = 0.70, within the declared band. The first trough is at least 9.50 below its adjacent highs; the second is at least 8.50 below its adjacent highs. If the two right-side bars have not closed before a neckline break, the triple label is unavailable under this rule, even if a completed chart later draws the high at bar 14. Preserve the earlier double-candidate status and do not backdate the new label.

For this example, expire an unbroken three-high candidate five completed bars after recognition. After a qualifying break, a close at 151.00 within the next five bars would fail a predeclared downward-continuation test using a close back above 150.20. Record that failure even if another drawing could later omit the pattern. A height-based target remains geometric arithmetic, not a forecast or an executable return.

A triple bottom mirrors the bookkeeping: three comparable lows, two intermediate highs, and a close above the higher intervening high under that detector’s convention. Invented lows of 80.00, 80.50 and 79.80 with intermediate highs of 90.00 and 91.00 set an upper boundary of 91.00. A high of 91.40 with a close of 90.80 does not satisfy a strict close-above rule; a later close of 91.20 could, once the third low is confirmed.

Mirror the geometry carefully

For a hypothetical bottom, lows near 80 and 80.50 with an intervening high of 90 can form the opposite candidate under analogous tolerances. Confirmation might require a close above 90. The reference boundary is the intervening high, not an arbitrary line chosen after the breakout.

A conventional target may project the formation's height beyond that boundary. Such a target is a geometric convention, not a probability estimate. A small difference between the two extremes also means the chosen measurement reference should be disclosed.

Keep the observation available at the time

Pivot-confirmation rules can delay recognition of both extremes. Record the time a candidate became eligible, whether it expires after a fixed interval and how overlapping candidates are counted. A chart with only the final selected pattern hides these choices.

Compare three-swing head-and-shoulders geometry and rounded recovery conventions. Preserve candidate and failure records in a journal. Consistency in classification makes the pattern research interpretable; the familiar M or W shape alone does not establish a trading advantage.