Price Action: Structure, Closes, and the Break-and-Retest
Market structure, running vs sweeping closes, and the break-and-retest — explained properly, then tested on 30 years of daily data, including what did not hold up.
What Price Action Actually Is
Price action is the study of price movement on a candlestick chart. Nothing more, nothing less. It rests on two things working together: candlestick behaviour, meaning how price reacts at a level and how momentum shows up in a single bar, and market structure, meaning the formation of highs and lows that define direction. Combined, they create context — a lens on what the market is doing and what condition it is in. A candle itself is just a unit of time showing a battle between buyers and sellers. The body is where the open and close sat, the wicks are how far price reached beyond them. That is the entire vocabulary, and it is enough, because what you are reading is not a pattern library. You are reading who was in control when time ran out.
Running Closes vs Sweeping Closes
When price trades into a key level — a prior high, a prior low, a swing point — what matters is not that it touched the level but how it closed. A running close is price pushing through the level and closing beyond it with strength; the conventional read is continuation, a breakout that means it. A sweeping close is price wicking through the level, taking the liquidity resting there, then closing back inside the prior range; the conventional read is a trap, a fake breakout rejecting the level it just broke. The principle behind the distinction is sound and worth internalising: don't just watch the level, watch the close. It is not about who tested the area — it is about who controlled it when time ran out. We were able to test that distinction directly, and the result is in the next section.
| Running close | Sweeping close | |
|---|---|---|
| What price does | Pushes through, closes beyond | Wicks through, closes back inside |
| What it takes | Territory | Liquidity |
| Conventional read | Continuation — the break is real | Trap — the break is rejected |
| Where it leaves you | Positioned with the move | Positioned against the sweep |
Free charts, alerts, and screeners for every asset discussed on this page. Used by 50M+ traders.
Testing the Distinction on Daily Bars
We defined a level as the highest high of the prior 60 sessions, then split every subsequent penetration of that level into running closes (the bar closed above the level) and sweeping closes (the bar's high exceeded the level but it closed back below), and measured the next 20 sessions against the asset's own baseline. On daily bars, the two are indistinguishable. On SPY, running closes were positive 65% of the time and sweeping closes 67%, against a 64% baseline. QQQ: 62% running, 64% sweeping, 62% baseline. GLD came in 2pp apart, TLT 6pp, BTC 1pp, and the signs do not agree across assets. These are healthy samples — 80 to 225 occurrences per bucket, not a dozen — so this is not a case of the test being too small to see anything. On this timeframe, how the candle closed did not separate outcomes.
| Asset | Running (n) | Positive | Sweeping (n) | Positive | Baseline |
|---|---|---|---|---|---|
| SPY | 225 | 65% | 221 | 67% | 64% |
| QQQ | 201 | 62% | 199 | 64% | 62% |
| GLD | 113 | 58% | 96 | 56% | 56% |
| TLT | 91 | 46% | 83 | 40% | 50% |
| BTC-USD | 88 | 67% | 81 | 68% | 56% |
Why That Is Not a Refutation
Read that result carefully, because the wrong conclusion is easy to reach. The running-versus-sweeping distinction is a lower-timeframe concept. It describes what happens inside a session — liquidity taken at a level, the reaction that follows, the close that settles it — and a daily bar is far too coarse to see any of that. A daily candle that closed back below a level may have spent six hours above it, or may never have held it for more than a minute; the daily close cannot tell you which. Our data covers daily bars only, so it can speak to the daily-timeframe version of this idea and nothing else. What the numbers do establish is narrower and still useful: if you learned this concept from daily charts and expect it to work there, it does not. The intraday version is a different claim, on data we do not have, and we make no claim about it in either direction. One genuinely interesting detail did fall out — on BTC, both running and sweeping closes ran well above baseline (67% and 68% versus 56%), which suggests that for Bitcoin what matters is that a 60-day high was challenged at all, not how the candle finished.
Structure: HH, HL, LH, LL
Market structure is the framework that tells you whether a market is trending or ranging, and it comes down to four labels. A higher high is a swing high that breaks above the previous high. A higher low is a low that forms above the last low, showing buyers stepping in earlier than before. Lower highs are failed attempts at new highs, and lower lows show sellers in control. There is a nuance most explanations skip: a higher low is only confirmed once price goes on to break the previous high. Until that happens you have a low that might be a higher low, which is a different thing from a trend continuing, and treating the two as the same is one of the more expensive habits in trading. The useful signal in structure is where it breaks. A lower low during an uptrend, or a higher high during a downtrend, says the market may be shifting gears — and that it is time to stop assuming continuation.
| Structure | Pattern | What it says |
|---|---|---|
| Uptrend | Higher highs, higher lows | Selling absorbed at rising prices |
| Downtrend | Lower highs, lower lows | Buying absorbed at falling prices |
| Range | Flat highs, flat lows | Neither side making progress |
| Possible consolidation | Lower low in an uptrend | Stop assuming continuation |
| Possible consolidation | Higher high in a downtrend | Stop assuming continuation |
Support and Resistance: Levels Are Made of People
A level matters because of what happened there, not because of the number. When price spends time at a price and then leaves quickly, it strands participants with unfinished business — buyers who wanted more and never got filled, sellers who were caught, and stops resting just beyond. When price returns, those participants act, and that is the entire reason a level produces a visible reaction. This also tells you which levels deserve a line. A high tested repeatedly by many participants across weeks has a crowd behind it. A random intraday spike touched once on thin volume does not, even though both plot identically. The question is never whether something is a level. It is who is waiting there and what they will do when price arrives.
The Break-and-Retest, Properly Defined and Tested
The break-and-retest is the most widely taught structure in price action, and its logic is sound: price breaks a level that has held, extends away, returns to that level, and holds — the participants who were selling there are gone or have changed their minds, so old resistance becomes support. The entry comes on the return rather than the break, which gives tighter and better-defined risk. We tested it strictly: a close above the prior 60-day high, a required extension of at least 2% before any retest counted, a retest at least three sessions later coming within 1% of the level, and a fixed 20-session forward horizon set before we looked. The result was mixed and small. SPY held retests were positive 76% of the time against a 64% baseline and QQQ 67% against 62%, but GLD came in at 47% against 56%, BTC at 53% against 56%, and TLT matched its baseline exactly. Three modestly better, two worse, largest sample 30 observations. That is what no edge looks like.
| Asset | Retests held (n) | Positive after 20d | Baseline | Difference |
|---|---|---|---|---|
| SPY | 21 | 76% | 64% | +12pp |
| QQQ | 30 | 67% | 62% | +5pp |
| TLT | 14 | 50% | 50% | 0pp |
| BTC-USD | 15 | 53% | 56% | −3pp |
| GLD | 17 | 47% | 56% | −9pp |
The Result That Should Change Your Mind
The most instructive number is not in that table. Conventional teaching says a failed retest — price returning to the broken level and closing back underneath — is bearish and a reason to exit or reverse. In our data it was not. On SPY, the four failed retests were positive over the following 20 sessions 75% of the time, essentially identical to the 76% for held retests. On GLD, failed retests were positive 80% of the time against 47% for the ones that held, which is the opposite of the folklore. The samples are far too small to claim that failure is bullish, and we are not claiming it. The honest conclusion is narrower and more useful: on daily bars, whether a retest held or failed carried no reliable information about what came next. If you trade this pattern because you believe the hold-versus-fail distinction is predictive, the daily data does not back you.
Why the Pattern Still Earns Its Place
None of this means abandon break-and-retest. It means being honest about what it does for you. Its value is structural rather than predictive. A retest entry gives a specific, visible, non-arbitrary level at which the idea is wrong, which means you can size the position from the distance to that level instead of guessing. It gets you into a move at a better price than chasing the break, which improves the ratio of reward to risk even when the hit rate is unchanged. And it imposes patience, which for most traders is worth more than any signal. What it does not do, on the evidence we have, is raise the probability that price goes up. Use it to structure a trade you wanted to make for another reason and it is genuinely useful. Use it as the reason for the trade and you are trading noise with extra steps.
The Three Questions
The method used on this page is more valuable than any of its conclusions, and it is portable. Whenever someone presents you with a pattern — in a course, a thread, a video, or a book — three questions settle almost everything. How many occurrences are in the sample? What exactly counts as the pattern, written down before the testing started? And what would a random window in the same asset have returned over the same horizon? Most of the time at least one of the three is missing, and it is usually the third, because a pattern that is positive 65% of the time sounds compelling right up until you learn the asset was positive 64% of the time regardless. Chart examples are not evidence. Anyone can find five charts where a pattern worked; the question is what happened across all the times it appeared.
Combining Price Action With Regime
Price action and the Alphameter work at different scales and answer different questions, which is what lets them be used together. Structure tells you where this specific instrument sits in its own sequence of highs and lows, and where you would be proven wrong. The regime score tells you whether cross-asset risk appetite is expanding or contracting behind it. The sensible division of labour is to let the regime decide whether you want exposure at all and roughly how much, and let price action decide where you enter and where you get out. A break-and-retest during a risk-on regime and the same setup during a deteriorating one look identical on the chart — the difference is what the rest of the market is doing while it happens. Neither replaces the other, and neither is a complete method alone.
Free charts, alerts, and screeners for every asset discussed on this page. Used by 50M+ traders.
Frequently Asked Questions
What is the difference between a running close and a sweeping close?▼
A running close is price pushing through a key level and closing beyond it with strength, conventionally read as genuine continuation. A sweeping close is price wicking through the level, taking the liquidity resting there, then closing back inside the prior range — conventionally read as a trap. The distinction is an intraday concept; on daily bars we could not find any difference in what followed.
What is a break and retest?▼
Price breaks a level that has held for some time, extends away from it, then returns to that level and holds above it, on the premise that former resistance now acts as support. Traders enter on the return rather than the break because it offers a tighter, better-defined invalidation level.
Does the break-and-retest pattern actually work?▼
Tested on daily bars across SPY, QQQ, GLD, TLT and BTC over 30 years, it produced no consistent edge: SPY and QQQ came in modestly above baseline, GLD and BTC below, TLT exactly level, with samples of 14 to 30 each. Its real value is defining risk and improving entry price, not raising the probability of a favourable move.
Is a failed retest bearish?▼
Our daily data does not support that. On SPY, failed retests were positive over the next 20 sessions 75% of the time versus 76% for held retests, and on GLD failed retests outperformed. Samples are small, so the honest conclusion is that on daily bars the hold-versus-fail distinction carried no reliable information.
When is a higher low confirmed?▼
Only once price goes on to break the previous high. Until that happens you have a low that might turn out to be a higher low, which is not the same as a confirmed trend continuation. Treating an unconfirmed low as confirmed structure is one of the more expensive habits in trading.
Related Topics
Track These Indicators Live
The Alphameter synthesizes six macro indicators into a single regime score — updated daily. See the current reading and full indicator breakdown on the dashboard.
Get notified when the market regime changes
Regime alerts + weekly macro brief. Unsubscribe anytime.