Support and Resistance: How to Find the Levels and How Often They Hold

Support is a price level where a falling market has stopped before, and resistance a level where a rising market has stopped before. Traders buy near support, sell near resistance, and read a close through either as a change in the balance. We measured how often that expectation came true on five years of the S&P 500 ETF: a swing low held on its first test a little more often than the low of an ordinary day, and each new test made a break more likely, not less.

How to find support and resistance

Levels come from four sources; only the first is measured on this page.

• Previous swing points. A swing low, a low below the lows of the bars around it, marks where buyers stopped a decline; a swing high, where sellers stopped a rally. Here a swing point needs 5 bars on each side, so it is known 5 bars after it forms.
• Round numbers. Prices such as 60,000 on Bitcoin, drawn on the chart below; we did not test whether they hold better than other prices.
• Indicators. A moving average, Fibonacci retracement levels or pivot points give levels computed from past prices.
• Lines and zones. A trendline is a sloped support or resistance; a zone, a band of prices instead of one level, allows for price turning a little above or below the exact level.

One support from start to break: Bitcoin at 60,000

On February 6, 2026, Bitcoin made a daily low of exactly 60,000, a swing low by the 5-bar rule and a round number. Price then rose far above it, and the level was not approached again for four months. Here is what happened when it was, with the test rules given in the next section:

1. First test, June 5. The low reached 59,130.91, below the level, but the day closed above it at 61,056.47. Two days later, Bitcoin closed at 63,332.01: the support held.
2. Second test, June 9. The low came to 60,780, and on June 11 the close was back at 63,625.99: held again.
3. Third test, June 24. The low fell to 59,102.70. On June 25 Bitcoin closed at 59,794, below 60,000: the support broke.
4. After the break. The break did not last: Bitcoin closed back above 60,000 the next day, crossed it back and forth for a week, made a low of 57,800.19 on July 1, and closed at 65,255.51 on July 20, 9.1% above the close that broke the level.

The same level held twice, broke on the third test, and the break itself reversed. One level is a story, not evidence; the counts below are the evidence.

How often do levels hold? Five years of the S&P 500 ETF

We wrote the test as fixed rules so that every level is judged the same way, using the 14-day average true range (ATR), as of the previous close, as the unit of distance:

• Level: the low of each swing low, from the day after the swing is confirmed.
• Ready to test: price must first close more than 1 ATR above the level. A level broken before that is never tested.
• Test: a day whose low comes within 0.5 ATR of the level, or below it.
• Held: a close more than 1 ATR above the level comes before any close below it. The level can then be tested again.
• Broke: a close below the level comes first.
• Comparison: the same rules applied to the low of every day that is not a swing low, each also counted from the sixth day after it. If swing lows are real support, they should hold more often than these ordinary lows.
• ATR: Wilder's average, started from the simple average of the first 14 true ranges. Tests still open on October 28, 2026, neither held nor broken, are left out.

S&P 500 ETF, daily, dividend-adjusted prices, January 4, 2021 to October 28, 2026. Each test of a level is counted once; a level that held its first test can appear again in the second row.

Swing lows held their first test 62.5% of the time against 51.9% for ordinary lows. With 56 first tests, that gap is about one and a half standard errors: suggestive, not proof. From the second test on, swing lows did no better than any other low. For swing lows, the share that held fell with each test; for ordinary lows, it fell only from the third test on.

Resistance did worse in this rising market. First tests of swing highs held 15 times out of 59, 25.4%, against 33.3% for ordinary highs: in five years when the ETF rose, sellers at old highs were overrun more often than not. On Bitcoin, from March 8, 2025 to October 28, 2026, 9 of 23 first tests of resistance held, 39.1%, against 48.5% for ordinary highs; too few cases to read much into. For resistance the rules are the same, mirrored: the level is the high of a swing high, and the test fails on a close above it.

How traders use the levels

• Range trading. Buy near support and sell near resistance while price stays between them. Our counts say why the stop matters: even on a first test, a support broke more than a third of the time.
• Stops. A stop just beyond the level is hit by any brief poke through it, like Bitcoin's low on June 5, 2026, which went 869.09 below 60,000 before the day closed above it. A wider stop, for example one ATR beyond the level, survives more of those pokes, at the cost of a larger loss when it is hit.
• Breakouts. A close through resistance is a breakout; under the test rules above, 59 swing highs on the S&P 500 ETF were broken from 2021 to 2026, on a first or a later test, and 34 of those breaks, 57.6%, saw a close back below the level within the next 10 sessions. A break that reverses into a loss for buyers is a bull trap.
• Role reversal. A broken support is often said to become resistance, and the reverse. We did not test it; the 60,000 example above shows how quickly price can cross a broken level both ways.

Levels are also the building blocks of chart patterns: a double bottom is a support tested twice, a rectangle a range between support and resistance. Our bench measures those patterns with fixed rules, as described in our methodology. More guides are in technical analysis, and combining a level with other signals is covered in confluence.