September 28, 2026 9 min read The Chartix Editorial Team

Every guide to support and resistance trading says the same three things: connect two swing points, more touches means stronger, use zones rather than lines. Those are reasonable habits. None of them explains why a level exists, which is the only thing that lets you tell a level you found from a level you drew because you wanted one.

Support is a price area where buying has repeatedly absorbed selling, so that declines into it have stalled and turned. Resistance is the reverse: an area where selling has repeatedly absorbed buying. Both exist because orders, positions and the memory of participants who missed the last move are concentrated at specific prices, not spread evenly across the chart.

This post covers the mechanism, a procedure for drawing a level from it, and a falsifiable test for whether the level is really there.

What is a level, actually?

StockCharts’ ChartSchool defines support as the price level at which demand is thought to be strong enough to prevent the price from declining further, and resistance as the level at which selling is thought to be strong enough to prevent the price from rising further. That is the observable half. Underneath it are three kinds of participant.

The first group has resting orders at the price. They decided in advance that they would buy at 48.00, and their orders sit there until price arrives. The second group already holds positions taken near the level: a trader who bought at 48.00, watched it rise, and now sees it back at 48.00 tends to defend that price rather than sell below it. The third group missed the last move and has been waiting for a second chance at the price they saw the first time. Bulkowski’s page on finding support and resistance describes exactly this: when price drops back to where it bottomed last time, traders buy, others do the same, and that buying is what turns price up. He calls it a self-fulfilling prophecy, and that is not an insult. It is the mechanism.

None of those three groups sits at one exact price. Resting orders cluster around a figure; positions were taken across a range of prints; memories are approximate. That is why a level is a zone. It is not a stylistic choice, it is what the mechanism produces. ChartSchool reaches the same place from the other direction, suggesting exact levels for tight, short ranges and zones for ranges that span many months.

Round numbers deserve their own line. Bulkowski notes that traders cluster orders at whole numbers, 9.00 rather than 8.93, and when Carol Osler at the New York Fed examined the levels that six currency-trading firms published for their clients, they ended in 0 or 5 far more often than chance would produce. The professionals draw where the orders are.

How do you draw one?

This procedure produces fewer levels than most charts carry. That is the point. The prices below are illustrative, chosen for arithmetic rather than taken from any real instrument.

Start from the highest timeframe you use, and work down. A level that is visible on the weekly chart has more participants behind it than one visible only on the 15-minute chart, because more sessions and more orders went into forming it. Draw the weekly levels first, then the daily, and only then look for intraday ones. If a level exists on both, it is one level, not two.
Take only swing highs and lows that produced a real reaction. A reaction is a move away from the price, not a pause. If price touched 48.00 and then rallied to 53.00, that is a swing low that says something. If it touched 48.00, drifted sideways for two days and then kept falling, the touch is not evidence of demand. It is evidence that nothing much was there.
Draw the zone, not the line. With swing lows at 47.80, 48.10 and 48.30, the level is the band from 47.80 to 48.30, not 48.07. A line at the average will be “broken” by normal noise and “held” by coincidence.
Decide wick or body by what the extreme was. If the low of 47.80 was a print that lasted minutes before the candle closed at 48.40, the closing body is the price the session actually settled on, and the zone belongs at the body. If the session closed near its low, the wick and the body agree and the question is moot. The higher the timeframe, the more a wick is a real excursion rather than a momentary print, so weekly wicks earn more weight than 5-minute ones.
Write the level down with a number and a direction. “Support 47.80 to 48.30, from the three lows in March” is a level. “Support somewhere around 48” is a feeling that will move when price does.

Two things are missing from that list on purpose. Moving averages are not on it because, on Bulkowski’s own account, moving-average support or resistance does not work well on daily charts. Gaps are not on it because the same page finds a rising gap supports price only about 20 percent of the time; the rest of the time price falls straight through it.

The test that tells a real level from a drawn one

A level that is really there produces a reaction on the next visit that a random price would not. That is a falsifiable statement, and it is the whole test. If price returns to 47.80 to 48.30 and moves away from it decisively, the level did something. If price arrives and drifts through as if the zone were not drawn, the level was a line on your chart and nothing else.

The test also settles what “more touches means stronger” actually means. Each touch of a support zone consumes some of what made it support: resting buy orders get filled, holders who defended the price the first time have less appetite the third time, and the buyers who missed the move have now had their second chance. If the reactions shrink with each visit, a rally of five points, then three, then one, the level is being consumed, not confirmed. The common teaching, which Wikipedia records as the more often a level is tested the more significance it is given, is true of the level’s history and silent about its future. The honest reading is that a level tested repeatedly with strong reactions each time is being defended, and a level tested repeatedly with shrinking reactions is being eroded. The chart tells you which one you have.

The evidence for either version is anecdotal. No study in the sources here counts touches and measures what happens on the next one. What has been measured is narrower, and worth knowing exactly.

Osler’s study is the rigorous one. She took the support and resistance levels six firms published daily from January 1996 to March 1998 for the dollar against the mark, the yen and the pound, and checked them against quotes sampled every minute of the New York day. Rates bounced at the published levels 60.8 percent of the time, against 56.2 percent at arbitrary levels chosen at random. The edge was real, statistically significant for most firm-currency pairs, and it lasted: five business days after publication the levels still beat arbitrary ones, at a bounce rate 1.7 percentage points lower than on day one. The part that matters for “stronger”: three of the firms rated each level’s strength, and those ratings had no consistent correspondence with how often price actually bounced. The professionals could find levels. They could not rank them.

Bulkowski’s tall-candle tests point the same way. On 453 stocks over three years, candles with bodies more than twice the prior 22-day average later acted as support or resistance 39 percent of the time; a second test over 15 years of data put it at 66 percent for bodies over four times the average, with his own mundane explanation attached: a taller candle covers more price, so more future turns land inside it by arithmetic alone. That is the state of the measured evidence: a modest, real effect at the level of “price reacts here more than at a random price”, and nothing that grades one level against another.

When does support become resistance?

The most repeated claim in the subject is that broken support becomes resistance, and broken resistance becomes support. ChartSchool states it as a principle: once price breaks below a support level, that former support can turn into resistance. The trapped positions are the engine. Everyone who bought at 48.00 expecting it to hold, and then watched price fall to 44.00, now sees 48.00 not as a bargain but as the price at which they can get out even. Their selling is what makes the old floor into a ceiling.

It is a common observation, not a law. The flip does not happen every time, because some broken levels are simply forgotten: the trapped holders capitulated on the way down and nobody is left to sell the return. Role reversal is worth expecting on the first return to a broken level and worth abandoning if that first return passes through cleanly.

What do candles add at a level?

A level tells you where. The candles that form when price arrives tell you what happened there, and only some of them carry information.

The tweezer bottom is two adjacent candles reaching almost the same low after a decline.

Tweezer bottom, 2 candles, after a decline. Drawn to the real proportions the formation rules describe. Full tweezer bottom guide

At a support zone this is a level being tested twice in consecutive sessions and holding both times, a small, specific version of the test above. It needs the decline before it; it is confirmed by a close above the high of the second candle; it is invalidated by a close below the shared low. Bulkowski is blunt that on its own the tweezers bottom tests near random, continuing lower 52 percent of the time in his data. At a level it is location evidence, and that is the honest amount to ask of it. Its mirror, the tweezer top, is the same two-touch structure at resistance.

A long lower wick at support, the hammer shape, records a session where price went into the zone and was pushed back out by the close. ChartSchool lists the hammer, the engulfing patterns and the morning star among the candles that can confirm or mark a support level, and it treats a doji or spinning top at support as a sudden letup in selling pressure, a stand-off rather than a turn. That distinction is the useful one: a doji at a level says the sellers stopped, not that the buyers started.

What carries no information is a candle’s colour on its own. A green candle that closed in the bottom quarter of its range was sold into all session; a red one that closed in its top quarter was bought. Where the body sits inside the range matters, and how to read a candlestick covers that measurement.

The chart-level version of a tested level is the triple top: resistance tested three times, with the reaction after each test being the whole question. The triple top pattern post covers what confirms it and what breaks it.

What this post is not

It is not a promise that a well-drawn level will hold. Levels are where reactions are more likely than at a random price, and the measured edge for even that is a few percentage points. What the procedure gives you is a level with a number attached, so that the next visit can prove it right or wrong instead of leaving you to decide after the fact whether the line counted.

For the trend and timeframe context every level sits inside, start with how to read a stock chart. For the candles that form at the levels, the candlestick pattern library draws each one to proportion. And for the swing points every level is drawn from, higher highs and higher lows pins down what counts as one.

Sources

Every figure above was checked against these before publishing.