Higher Highs and the Shape of a Trend

Structure first, and how to count it

Trend is a sequence of swing points, and the discipline is to mark them before forming a view. An uptrend is a series of higher highs and higher lows; a downtrend is lower highs and lower lows; anything else is a range, and ranges are where most chart-reading mistakes happen because a learner sees a trend in every leg. Marking swings is mechanical: on your chosen execution frame, label each significant high and low, then compare each to the last. Two advances and two declines is enough to state a direction, though a trend with one leg is indistinguishable from the first move of a range, and the honest answer at that point is “not established yet.” The most useful single price in a trending market is the last higher low, because it is both the definition of the trend and the invalidation. While that low holds, the sequence stands and any decline into it is a pullback; when it breaks, the trend as defined has ended and the burden of proof has flipped. That gives the plan its stop for free, in the same way a level did in TR13 — and it gives a rule that resists the most expensive version of the same mistake, which is buying a bounce after the trend has already been broken and calling it a pullback. If the low broke, the structure is gone; the bounce is a different trade with a different thesis. Moving averages are the crude summary of all this, and knowing what they are is worth more than using them blindly. A 50-day average is the mean of the last fifty closes, so it rises when the recent closes are higher than the ones it drops off the back — which is why it lags a turn by weeks. Its use is as a filter and a reference rather than a signal: price above a rising average is the arithmetic version of an uptrend, and the average is where pullbacks in healthy trends repeatedly end, because it is the price most participants are using. A 200-day average is the same object on a slower clock and is watched by institutions, which is why it can be self-fulfilling. What an average cannot do is tell you a trend is about to change; it can only confirm that one has, later. Reading the same chart as structure — Higher highs and higher lows, last low intact: Uptrend · declines into the last low are pullbacks ← · Higher high, then a lower low: Sequence broken · the trend as defined has ended · Price above a rising 50-day, pullbacks ending at it: Trend filter agrees with structure · Chop between two levels for months: Range · the swing sequence is absent, so trend tools do not apply ← Not every wiggle is a swing high. The practical threshold is relative to the frame’s own ATR: a swing needs to be a move you would notice — several times the average bar — otherwise the chart is a forest of swings and the sequence you “find” depends on where you place the cursor.

Age, and the signs that a trend is old

A trend that has run for two months is not the same object as one that has run for two years, and the difference is measurable. Age is the calendar; extension is how far price has travelled from its own reference, usually the 50-day average in percentage terms — a name 40% above its 50-day average is in a different regime from one 5% above it, and the first is far more sensitive to disappointment. Width is the third marker: when daily ranges and volume both expand as price rises, the trend is being driven by more participants disagreeing rather than agreeing, and blow-off tops usually carry the largest volume of the whole move. None of these predict the end, and that is the point — they are risk markers. A trend that is old, extended and wide is one where the position size should be smaller and the stop no longer, because the ordinary pullback of such a regime is large. The counterpart is the pullback that is healthy. In a sound uptrend, declines typically come on lower volume than advances, and the recovery comes on higher volume — participation leaving as price dips and returning as it turns. When that pattern inverts, with heavy volume on a decline and light volume on the bounce, the market is telling you the character changed even if the structure has not yet broken on the frame you are watching. Volume is the next lesson precisely because it is the qualifier that distinguishes an orderly pullback from distribution, and the two are indistinguishable on price alone. Which leaves the practical question: what do you do with a trend? Two approaches are defensible. The first is continuation on a pullback — wait for the decline into a level or the last swing low, enter with structure intact, and let the last swing low be the stop, which is what makes the geometry good. The second is a breakout continuation after consolidation, buying the resolution of a flag rather than the pullback. What is not defensible is counter-trend trading on the same frame: it is not that reversals never happen, it is that the ones you can act on announce themselves on the frame that defines the trend — a broken swing low, a failed recovery, structure gone — rather than on a bounce that appears while the structure is intact. • Age: how long the sequence has held, in calendar terms. • Extension: distance from the trend reference, in percent — the pullback cushion. • Width: expanding ranges and volume mean more disagreement and larger ordinary pullbacks. • Volume character: declines on light volume and recoveries on heavy is healthy; the inverse is not. • Position response: old, extended and wide means smaller size, not a tighter stop. Trend lines are drawings rather than measurements. Two people draw the same line differently, and a line that has been redrawn three times to avoid being broken is a story about the analyst. Use structure — the last higher low — as the objective version of the same idea, and treat any line that disagrees with structure as decoration.

From one chart to five hundred: screening for trends

Reading structure by eye works on the chart you chose to open. It does not scale, and the practical problem is the other direction: out of the thousands of instruments available, which handful deserve the careful read? A screen is the numerical stand-in for the structure the lesson just taught, and knowing what each stand-in misses is the skill. The simplest proxy is **price relative to a long average** — above the two-hundred-day, say. It catches sustained advances but cannot tell a smooth trend from a stock oscillating around the line, so it produces a universe full of range-bound names. The second is **the slope of a regression over a window**, normalised by price or by ATR so that a fast stock and a slow one can be compared; it measures direction without saying anything about the shape of the path. The third is **ADX above a threshold**: a measure of strength that is deliberately blind to direction, which is why it is used as a filter on a directional screen rather than as the screen itself. The fourth is the one closest to the lesson — **counting higher highs and higher lows in the last few swings** — and it is the most faithful, the most dependent on the swing definition, and the hardest to compute reliably across hundreds of names. Each proxy fails in a recognisable way, and the failures are worth naming because they are the reason a screen is a funnel rather than an answer. A moving-average rule mostly lifts names that have already gone nowhere for months. A regression slope lags a fresh turn by construction, because it needs several bars to lean. ADX is high in a violent range as well as a trend when the range is wide enough in both directions. And proximity to a fifty-two-week high — the classic momentum proxy — catches both the resilient leader and the blow-off top, because it measures distance from a price rather than the path that reached it. The remedy is to use the cheap numeric filters to shrink the list and the structural read to make the decision. A screen that reduces three thousand instruments to forty candidates has done its job; the forty are then read the way the earlier screens in this lesson taught, one chart at a time, with structure, volume and the shape of the pullbacks examined properly. Automating the last step is where learners get into trouble, because the thing they were actually good at was the reading. One caution applies to the screen as much as to any other rule. It is a parameterised rule whose thresholds decide its output, so it inherits the discipline discussed elsewhere in this subject: fix the definition in a sentence, test it on data you did not choose it from, and accept that tightening it trades missed trends for fewer false positives while loosening it does the reverse. There is no threshold that catches every trend and rejects every fake, and a screen that appears to is one that was tuned until it did. • Price versus a long average: catches sustained advances, and a lot of nothing. • Normalised regression slope: gives direction without shape. • ADX: strength without direction, and it is wide across violent ranges too. • Swing counting: closest to the lesson, and as stable as the swing definition you fixed. • Use numbers to shrink the list and the structural read to make the call. A screen written as a sentence — “above the two-hundred-day average, ADX above twenty-five, higher lows in the last four swings” — can be tested, revised and compared across weeks. A screen that lives in the software’s menu cannot be argued with, which is exactly why it is dangerous.

Sources

• Market structure and swing definitionDow theory as summarised in standard technical analysis texts
• Moving averages as trend filtersFaber, “A Quantitative Approach to Tactical Asset Allocation” (2007)
• Trend persistence and regime evidenceMomentum factor literature; time-series momentum studies