Reading the Regime: Trend, Range and Standing Aside

Two regimes, measured rather than felt

A trending market and a ranging market differ in a property you can observe: in a trend, each advance exceeds the last high and each decline stops above the last low, so the swing sequence progresses in one direction. In a range, price returns to the same levels, and the swings fail to progress — highs cluster near the same ceiling and lows near the same floor. That is not a subtle judgement: mark the principal swings on the frame you trade and ask whether they advance, retreat or repeat. Three answers, and the third is the one people miss, because a range full of legs looks like a series of trends if you only look at one leg. There are two useful measurements that make the classification less dependent on how you mark swings. The first is the relationship between price and its trend reference — the share of recent closes above the 50-day average, or simply whether the average is rising, flat or falling. A healthy trend has most closes on one side and an average with a slope; a range has closes distributed on both sides and an average with no slope. The second is the ratio of trend travel to total travel: divide the net move over a window by the sum of the absolute daily moves inside it. A market that travelled 8% net over sixty sessions while moving 40% in total is doing a lot of work to go nowhere, and that ratio is a compact description of a range. Neither measurement is a signal; both are filters, and their job is to answer one question before any setup is considered. The consequence of the answer is a frequency decision. In a trend, the continuation methods — pullback into structure, retest of a broken level — have an edge worth paying costs to exercise. In a range, the same setups form constantly and fail at a rate close to the cost-adjusted coin flip: the level is tested, the break goes ten cents beyond the zone, the stop is hit, and the market returns to the middle. A trader who cannot bring themselves to reduce activity in that regime does not have a method, they have a habit with fees attached, and the correct response is a rule written in advance — the market is classified as ranging, the trade size or the trade count falls, and the setups are logged rather than taken. Classifying before trading — Swings advance: higher highs, higher lows: Trend · continuation methods apply ← · Swings repeat: same highs, same lows: Range · the same methods become coin flips with costs ← · Net move 8% over 60 sessions, total travel 40%: Efficiency ratio 0.2 — most of the work went nowhere · Price mostly above a rising 50-day average: Trend filter agrees with the swing read The regime matters for the instrument as well as the index. A large-cap name in a range while the index trends is a common configuration, which is why the filter is applied to the instrument being traded rather than to the market headline.

The stand-aside rule, and what to do instead of trading

Standing aside has a bad reputation because it feels like inaction, so it survives only if it is written as a rule with a trigger. The version that works is short: the instrument is classified on the frame you trade, and if the classification is a range, then no continuation setup is taken. That is it — one condition, checked before the setup rather than after, and the same condition that authorises trading when the sequence resumes. Two supporting habits make it liveable. First, log the setups you declined: the log is what converts the rule from a feeling into evidence, and it is the only way to find out later whether the filter helped or merely seemed prudent. Second, have something to do that is not a trade — reviewing the journal, updating levels, checking the calendar — because a rule that leaves a person staring at a chart is a rule that gets broken. If a range must be traded, the method changes rather than disappears. The edges are the trades: buying near the established floor with a stop just beyond it and selling into the ceiling, with size reduced because the reward is bounded and the band is narrower than a trend leg. What is not defensible is taking a continuation setup in the middle of the band, which is the highest-cost, lowest-information corner of the whole configuration. And the edges are only tradeable while the band has been tested enough to be a band: a range whose floor has been touched twice and whose ceiling once is a market in the process of deciding, where a break is more likely than a bounce. Which leads to the hard part: regime changes arrive inside trades. A position taken in a trend can see the trend end while it is open, and the honest response is the structural invalidation from TR7 — the broken swing low — rather than a re-classification that keeps the position alive. The order is: structure breaks, position exits, and the regime is re-read with a flat book. Inverting it, by declaring the market to be ranging so that a losing trend position can be held in hope, is the same failure as moving a stop, described in a different vocabulary. • Classify on the frame you trade, and classify before the setup. • Trend: swings advance. Range: swings repeat. Deciding: swings have not repeated yet. • Efficiency ratio (net move ÷ total travel) compresses the range read into one number. • In a range, reduce frequency or size — and log the setups you declined. • If trading a range, trade the edges with a stop beyond the band, not the middle. • When structure breaks, exit first and re-read the regime with a flat book. Two opposite errors look like discipline and are not. The first is refusing to trade at all because the market “looks choppy”, which is a feeling without a trigger. The second is re-labelling a losing trend as a range so the position can be held. Both avoid the measurement, one by pessimism and one by hope, and the fix for both is the same written condition.

What false breakouts actually cost

The claim that a range is expensive is easy to say and worth quantifying, because the number decides whether the filter earns the trades it declines. Take a method that fires on every test of a well-defined level. In a trend, most tests either hold as support or break and continue, so the level carries information. In a range the same level is tested repeatedly, which means the method fires several times as often — and each firing pays the spread and the slippage whether or not it works. Frequency is the cost multiplier, and in chop the multiplier works against the method. The base rates in the chart-pattern literature are instructive in shape rather than in precision. Well-studied continuation patterns resolve as expected somewhat more often than not, and the figures vary widely by pattern, by market and by whether volume confirmed the move. What does not vary is that a meaningful minority of breakouts fail and reverse, and that failures cluster precisely where a level has been tested many times — which is the definition of a range. A trader who treats every break as a signal is therefore taking the pattern with the weakest average payoff in the configuration where it occurs most often. That combination is what a regime filter buys. Declining the setups in the range does not require the method to be wrong; it requires the expectancy there to be negative once costs are paid, which it usually is when the reward is bounded by the width of the band and the failure rate sits near a coin flip. The arithmetic makes it concrete. At a one-to-one reward, a 45% success rate is negative after costs; at two-to-one, the same rate is comfortably positive. The regime decides which of those two payoffs the setup is offering, which is why classification comes before evaluation rather than after it. The last piece is how much evidence a classification needs. Two failed breaks tell you almost nothing — that outcome sits inside the range of a working method. The signal is a change in the relationship between the level and the response: tests that used to lead somewhere now return to the middle, and the displacement after each attempt shrinks. That is why the efficiency ratio and the count of failed retests are more useful than a running tally of wins and losses, because they describe the structure rather than the sample. • Frequency is the multiplier: a range fires the same method several times as often as a trend. • Every firing pays the spread whether or not it works, so cost scales with count. • Failures cluster where the level has already been tested many times — which is what a range is. • Work the arithmetic per regime: 1:1 at 45% is negative after costs; 2:1 at 45% is not. • Shrinking displacement after each test is structural evidence; a two-trade sample is not. The mirror-image error is applying the filter with hindsight — declaring a range after the failures and a trend after the wins. A filter counts only if it is judged on the classification it would have produced before the outcome, which is what makes the log of declined setups a research record rather than a diary.

Sources

• Trend and range regime classificationStandard technical analysis treatments of market phases
• Trade frequency and edge decay in choppy marketsTrading-cost literature; expectancy arithmetic under higher transaction counts
• False-breakout statisticsBulkowski, “Encyclopedia of Chart Patterns”