Chart Patterns

What chart patterns are

Price moves up and down. The turning points are called swing highs (a high with lower prices on both sides) and swing lows (a low with higher prices on both sides). A chart pattern is a name for a particular arrangement of those turning points. Two swing highs at about the same level make a double top. A high with a higher high in the middle and a lower high on each side makes a head and shoulders. Highs that fall while lows rise make a triangle.

Each pattern also comes with a story about what the crowd was doing. In a double top, buyers pushed price up to a level twice and failed twice. The story is the reason traders watch the shape. It is not a measurement. The same shape can be followed by a move in either direction, and how often each happens depends on the market, the period and the exact definition. No count of that is given on this page.

Continuation versus reversal

Patterns are normally split into two families.

• Continuation patterns are read as a rest inside a trend. Flags, pennants and rectangles are examples. The usual reading is that price pauses, then breaks out of the shape in the direction it was already going.
• Reversal patterns are read as an attempt to turn a trend around. A double top, a double bottom and a head and shoulders are the standard examples. The usual reading is that price breaks the other way.

Some shapes fit both families. A triangle can break either way, and which way it breaks is what decides the label, afterwards. That is the first hint of the main problem with patterns: the name often arrives after the result. A shape that is called a reversal pattern when it works is quietly called something else when it does not.

Common patterns in one line each

Read the table as vocabulary. For deeper pages on two of them see head and shoulders and double bottom. Both pages show the arithmetic of one pattern, with made-up prices.

Why drawing a pattern is subjective

A human draws a pattern by eye. The eye is flexible, and that is the trouble. Look at the points where you place a line, how many bars you allow between two highs, and how close two highs must be to be "about equal". Change any of those and the chart gives a different answer. A trader who knows the shape they are hoping to find tends to find it.

The same flexibility works backwards. After a move has happened, it is easy to look back and see the pattern that led to it. Those cases get remembered and shown in books and videos. The cases where the shape appeared and nothing followed, or where a shape was drawn only because the line could be placed that way, are not shown. This is a form of survivorship bias, and it is why a pattern that looks obvious on a finished chart may be hard to use on a live one.

Support and resistance lines have the same issue. A level is a price where turns have happened before, but which turns count, and how wide the level is, is a choice. See support and resistance.

How a bot could detect a pattern

A bot cannot use a feeling. It needs a procedure with numbers in it. A common one has three steps.

• Find the swing points. Pick a window size K. A bar is a swing high if its close is above the K closes before it and the K closes after it. A swing low is the same, reversed. Note that the rule needs K bars after the point, so a swing high is only known K bars late. A bot that treats it as known at the time of the high is using future data. See look-ahead bias.
• Compare the swing points with a tolerance. Two highs count as equal when their gap, as a percent of the first, is no more than a set tolerance.
• Define the trigger. For a double top, that could be a close below the lowest low between the two highs. That is a breakout rule, and it meets the same traps as any other. See breakout trading and false breakout.

A worked example with made-up numbers

Take the 22 made-up closes in the picture at the top. With K = 2, the swing highs are at bars 3, 8, 14, 20, at prices 105.0, 106.0, 105.5, 104.5. That gives 4 swing highs and 6 possible pairs. For each pair the gap is the difference in price divided by the first price. The table shows the gap and whether it passes at tolerances of 0.25, 0.5, 1, 1.5 percent, in that order.

At 0.25 percent no pair passes, so a double top rule finds nothing. At 0.5 percent 3 pairs pass. At 1 percent 5 pass, and at 1.5 percent all 6 pass. Nothing about the prices changed. Only one setting did, and the number of patterns went from 0 to 6. All numbers here are made up and were calculated by us on 31 Oct 2026, exact. This is the same kind of problem described in overfitting: a setting that can be tuned until a backtest looks good.

How pattern rules break in a bot

• Too many settings. Window size, tolerance, minimum bars between points and the trigger are all numbers. Choosing them by looking at past results fits the past. The overfitting page shows the effect.
• Late detection. A swing point needs bars after it to be confirmed. By then price may have moved a long way from the high.
• Costs at the trigger. The trigger is a break of a level, so the fill is often worse than the level. See slippage.
• Too few examples. A rare pattern gives a handful of trades, and a handful proves nothing.
• Indicators layered on top. Adding a filter such as RSI, Bollinger Bands or an EMA crossover adds more settings and more ways to fit noise.

How to check a pattern rule

Write the rule down in full before you test it: window size, tolerance, trigger, stop and exit. Run it on several years of real data and several markets, with spread, fees and slippage included. Count the trades, the share that lose, and the average result of each, as in expectancy. Hold back a stretch of data that you do not touch until the end, then follow the method in how to backtest. Test how the result changes when the tolerance moves a little either way. If a small change flips the answer, the rule was fitted to noise. Size any trade by the loss you can accept, using the position size calculator.

What this does not tell you

It does not say that any pattern predicts anything, how often a pattern is followed by a given move, or which pattern is best. The shapes and names are a way to describe a chart. The prices are made up and show only how to count. This page is education only.