Trends, trendlines and market structure

The answer

A trend is a countable sequence: higher highs with higher lows, or lower highs with lower lows. A trendline is a straight line you draw on top of that sequence, and the line is a much weaker thing than the sequence it sits on.

Why this costs you money

Here is the mistake, and almost every chart reader makes it for years.

You open a chart. You see a rising market. You place a straight line under the lows, and the line touches 4 times. It looks exact. You decide the line is support, you buy near it, and you place your stop just under it.

Then the price cuts through the line and you sell. Two weeks later the market is higher than where you sold.

The line was never the thing. You drew it after all 4 touches had already happened. At the moment of the second touch — the earliest moment you could have drawn any line at all — there were at least 5 different lines you could have drawn through the same 2 points and the neighbouring bars, and you have no way of knowing which one you would have chosen at the time. You chose the one that fits the last 4 touches because you can see all 4 touches now.

This is a specific, measurable error, and it has a name in every other field: fitting a rule to data you have already seen. In a chart it is almost invisible, because the chart shows you the past and the future on the same screen.

The cost is not one bad trade. The cost is that you build a whole method on lines that only exist in hindsight, and then you cannot understand why the method does not work in real time. It works perfectly on old charts. It has to. You drew it on old charts.

How it works

Split the subject into 2 parts. One part is arithmetic. The other part is drawing. Keep them separate for the rest of your life.

Part 1: market structure, which is arithmetic

A swing high is a bar whose high is higher than the bars on both sides of it. A swing low is a bar whose low is lower than the bars on both sides. You usually require a set number of bars on each side — 3 bars either side, or 5 — so that small wobbles do not count. Once you fix that number, finding swing points is a calculation. Two people using the same setting get the same points.

Now write the swing points down in order.

That is the whole definition. It is not a matter of taste. Either the last swing high was above the one before it, or it was not.

The important event in this system is not a line breaking. It is a break of structure: the first time an uptrend makes a lower low than its previous swing low, or the first time a downtrend makes a higher high than its previous swing high. That event has a date and a price. You can write it on a calendar. Two analysts using the same swing setting will agree on the date.

Part 2: trendlines, which are drawings

A trendline needs 2 points. Two points always make a line, so 2 points prove nothing at all. Every chart has thousands of pairs of points, and therefore thousands of possible trendlines.

Every trendline also requires 4 choices, and each choice changes the line:

1. Wicks or bodies. Do you connect the lowest price of the day, or the lowest close?
2. Linear or logarithmic scale. On a log scale, a line that fits a long rise is a completely different line.
3. Which swing lows count. Skip 1 small low and the angle changes.
4. Where you started. A line from the 2020 low and a line from the 2022 low through the same recent points are not the same line.

None of those 4 choices has a correct answer. That is what makes a trendline a drawing rather than a measurement.

This does not make trendlines useless. It makes them a summary. A trendline is a quick picture of the rate at which the lows have been rising. If the line is steep, the market has been rising fast, and a fast rise is harder to keep up than a slow one. That is real information. It is just much less precise than the thin line on your screen suggests.

Trend strength, measured properly

Do not judge trend strength by the angle of a line. Angle depends on the scale of your chart, and your chart's scale depends on the size of your window.

Judge it by 2 numbers you can count:

• The depth of each pullback, as a percentage from the swing high to the swing low. In a strong uptrend, pullbacks stay shallow — often under 5% on an index. When pullbacks start getting deeper each time, the buyers are getting slower, whatever the line says.
• The time each pullback takes. A strong trend recovers to a new high quickly. When each recovery takes longer than the last one, that is a change even if no low has been broken.

Both of these are numbers. Both can be written down. Neither depends on how you hold your mouse.

Channels and timeframes

A channel is 2 parallel lines, one along the lows and one along the highs. It inherits every problem of a single trendline, and adds one more: the second line is placed by copying the first, so it is not independent evidence of anything.

Timeframe alignment is more useful and much less discussed. A market can be in an uptrend on the weekly chart and a downtrend on the hourly chart at the same time, with no contradiction, because the swing points are different at each setting. When somebody says "the trend is up", the honest question is: measured on which chart? A daily uptrend inside a weekly downtrend is a normal, common state, and it is the state where most people lose money, because 2 true statements point in opposite directions.

What it tells you, and what it does not

Trends persist more often than chance. This is one of the very few claims in technical analysis with serious evidence behind it. Jegadeesh and Titman documented in 1993 that stocks which had gone up over the previous 3 to 12 months tended to keep going up over the following months, and the effect has since been found across countries, decades and asset classes. That is real, and it is the honest foundation under the word "trend".

Note carefully what that evidence supports. It supports the idea that direction over months tends to continue. It does not support any claim about a particular line on a particular chart. The academic work measures a stock's return over the last 6 or 12 months. It never draws a line.

So a trend tells you the current direction of the sequence, and that direction has a mild tendency to continue. It does not tell you:

• How much further. Nothing in a trend contains a target.
• When it ends. By definition, you only know a trend ended after it ended.
• That the price is reasonable. A trend is a fact about price history and contains no information about the business.
• Anything at all about a single day. Momentum evidence is about months. Nothing about it survives being applied to the next 30 minutes.

The decision rule

Use structure to decide whether. Use a trendline only to decide where to look.

If the sequence of swing lows is still rising, treat the market as an uptrend even when a drawn line has broken. A broken line with the structure intact is a change of speed, not a change of direction.

If a swing low breaks below the previous swing low, treat the uptrend as over for that timeframe, even if no line has broken. The structure break is the event. The line is decoration.

Unless the break of structure happens on a single day driven by a one-off event — a policy announcement, an index rebalancing, a single large forced seller — in which case wait for the next swing point before you conclude anything, because 1 bar is not a sequence.

Try this now

This takes 5 minutes and it is the single most useful thing in this article. Almost nobody does it, and everybody who does it changes how they draw lines.

1. Open any chart you already look at. Set it to daily bars, 3 years.
2. Use your platform's drawing tool to draw the trendline you think is obviously correct. Take 20 seconds. Do not be careful about it. Be as careful as you normally are.
3. Now hide the last 6 months. Most platforms let you drag the chart to the left so the recent bars move off screen, or set the date range to end 6 months ago. If yours cannot do this, cover the right side of the screen with a piece of paper. Paper is fine.
4. Delete your line. Looking only at what is on screen, draw the trendline you would have drawn on that day, with no knowledge of what came next.
5. Bring back the last 6 months and put both lines on the chart at the same time.

What you should see. For most people, most of the time, the 2 lines are not the same line. They start at different points, they have different angles, and they give different answers about whether the trend is still intact today. On a market that has had a sharp pullback in the last 6 months, the difference is usually large.

If your 2 lines happen to match, do it again on a different chart, and then on a market that fell. It is easy to draw the same line twice on something that went straight up.

The lesson is not that trendlines are worthless. The lesson is that the confidence you feel when you look at a completed chart is not available to you at the right edge, and the right edge is the only place you ever trade.

Then do this second step, which takes 1 minute. Look at your last 5 trades. For each one, ask whether the line or level you used was visible on the chart before you entered, or whether you can only see it now. Write down the count.

Three real cases

1. NIFTY 50, March 2020 to September 2024 (India) — the line broke, the trend did not The Indian index bottomed around 7,511 intraday on 24 March 2020 and reached about 26,277 intraday on 27 September 2024. Over those 4 and a half years, the sequence of higher swing lows held. Any trendline drawn along the lows in 2020 was broken well before the end — the pullbacks of early 2021, of October 2021 to June 2022, and of early 2023 all cut through lines that had looked exact. In every one of those cases the structure of higher lows on the weekly chart survived, and the index went on to a new high. A reader who sold on the line break sold 3 times inside one long rise. the exact intraday extremes and the dates of each intermediate low against NSE historical data.

2. S&P 500, 3 January 2022 to 12 October 2022 (United States) — structure told you, and it told you early The index closed at a record 4,796.56 on 3 January 2022 and closed at 3,577.03 on 12 October 2022. Inside those 9 months there were at least 4 rallies of more than 6%, and every one of them stopped below the previous rally's peak. That is the definition of a downtrend, and you could confirm it with 2 numbers each time, without any line. The largest of those rallies peaked in mid-August 2022 around 4,325 and was widely described at the time as a new bull market. The sequence of lower highs said otherwise, and the sequence was right. the exact intraday peak of the August 2022 rally.

3. NIFTY 50, 27 September 2024 to early 2025 (India) — the break of structure had a date After the September 2024 high, the index fell through the following months and reached roughly 21,743 intraday in early April 2025. What is instructive is the order of events. The first lower low below the prior swing low happened months before most commentary accepted that the uptrend had ended, and it happened on a specific dateable day that anybody using a fixed swing setting could identify. The people still arguing about whether a trendline had broken were arguing about a drawing. The people counting swing points had a date. the exact April 2025 low and the date of the first break of structure.

The question that resolves it

A novice looks at a chart and asks: is the trendline holding?

An expert asks: would I have drawn this line 6 months ago?

The second question cannot be answered by looking at the chart in front of you, which is exactly why it is the useful one. It forces you to reconstruct the information you actually had at the time, and that reconstruction is the whole skill.

What would make this wrong

If trendlines carried real information beyond the structure underneath them, then a line drawn using only past data would predict future touches better than a random line of similar angle. This has been tested many times and the results are weak. If somebody shows you a study where lines drawn in advance, by a fixed rule, on a large sample, beat the simple structure rule, this article should change.

Three honest limits on what is written above.

First, structure has its own hindsight problem, though a smaller one. A swing high is only confirmed once the bars after it have printed. With a 3-bar setting, you know a swing high existed 3 days after the fact. That delay is real. It is fixed and knowable, which is the difference from a trendline, but it is not zero.

Second, the momentum evidence is about portfolios of many stocks over months. It does not promise that any single trend continues. A rule with a positive average across 500 stocks can lose money on the 1 stock you own.

Third, on a market that is genuinely moving sideways, all of this produces noise. Swing points alternate without direction, structure breaks in both directions within weeks, and every method described here gives false signals. Ranges are covered in the next article, and the honest answer for a range is to stop applying trend tools to it.

Carry this

• A trend is a sequence you can count. A trendline is a drawing you chose.
• Structure breaking is an event with a date. A line breaking is an opinion.
• Before you trust any line, ask whether you would have drawn it 6 months ago.

Knowledge check

Related

• Support, resistance and role reversal
• Fibonacci retracements and extensions
• Reversal chart patterns
• Continuation patterns and gaps
• ← Trend, support and market structure