What is a trendline?

Start with what a trend is, before any line is drawn. An uptrend is a sequence: each rally clears the previous peak and each pullback stops above the previous trough. Buyers keep regaining control before price gets back to where they last defended it. That sequence is the trend, and it exists whether or not anyone draws anything. This is the structure Charles Dow described in his editorials at the turn of the twentieth century, and which Robert Rhea set out formally in 1932. Notice that the definition contains no line and no slope. It is purely a statement about the ordering of highs and lows.

So what does a trendline add? Exactly one thing: a rate. When you lay a straight edge under two swing lows, you are asserting that the floor has been rising at a particular number of points per bar, and that this pace is a meaningful property of the move rather than an accident of two dips. Everything the line then does for you follows from that assertion. It projects a level into empty space where price has not yet traded, because a constant rate extrapolates. It gives you an alarm, because a close beneath it means the floor stopped rising at that pace. It gives you a shape you can compare across instruments and timeframes.

The trouble is that the rate is not in the data. The sequence of higher lows is in the data. The rate is something you fit to it, and the fit depends on choices you make before you see the result. A trendline is therefore closer to a regression through two hand-picked points than to a measurement. It behaves like a hypothesis: it makes a prediction about where the next low should be, later price either supports it or does not, and you are expected to abandon it when it fails. Most of the trouble people have with trendlines comes from treating the output of that fitting exercise as though it had the same standing as the high of last Tuesday.

Once you hold that idea steadily, the rest of this page is mostly consequences. If the line is a rate hypothesis, then a break is evidence about the rate. If the line depends on your anchors, then someone with different anchors is looking at a different hypothesis and will disagree with you honestly. If the vertical axis rescales the data, then the rate itself is measured in different units and the same bars support a different line. None of that makes the tool useless. It makes it a tool for organising your own view, which is a real job, rather than a detector of events that happen to the market.

The fastest way to see why this matters is to stop arguing about it and draw it. Below is one price series of twenty six bars, plotted twice. Both panels contain identical data. Both lines are anchored to the same two pivot bars, bar 5 and bar 15. Both are judged by the same rule, a decisive close beyond the line. The only difference in the entire figure is a single drawing convention: the left panel anchors the line to the lows of those two bars, and the right panel anchors it to their closes.

That one choice changes the slope, because the two pivot bars have very different lower wicks. Bar 5 is a long-tailed bar whose low is nine points beneath its close; bar 15 is a tidy bar whose low sits one point under its close. Reading the lows, the floor appears to have risen 2.05 points per bar. Reading the closes, it appears to have risen 1.25 points per bar. Neither reading is careless. The first says the wick is where buyers actually turned the market and therefore where the demand was; the second says a wick is a price that was rejected within the session and only the close is a verdict the market kept.

Sit with the arithmetic for a moment, because the mechanism generalises. Any difference in anchor height compounds with distance. Here a nine point discrepancy at one anchor and a one point discrepancy at the other produced a slope difference of 0.8 points per bar, and ten bars past the second anchor that gap has grown to seven points. The further you extend a line, the more the small decision you made at the start dominates the level you are now watching. This is why two experienced chartists rarely disagree about a line's first test and frequently disagree about its fifth.

The lesson is not that one of these lines is correct. It is that the phrase "the trendline broke" is incomplete in the same way "the temperature is 35" is incomplete without a scale. A break is always a break of a specified line, and specifying it means naming the anchors, the convention and the axis. If you cannot state those three things about a line you are watching, you do not yet have a hypothesis. You have a shape.

It is worth listing the decisions explicitly, because in practice they are made unconsciously and then forgotten. Each one is a genuine fork with a defensible answer on both sides, and each one moves the line. Together they are the reason a trendline cannot be a shared object the way a prior day's high can.

The first is wicks or closes, which the figure above has already made concrete. The second is which lows count. Every chart contains many small dips, and turning a dip into a pivot requires a threshold: how many bars either side must be higher, how deep the dip has to be, whether a two bar consolidation counts. Change the threshold and you change the candidate anchors, and therefore every line you can draw. The third is how many touches you demand before you trust the line, which sets how long you wait before acting and how often you are acting on a coincidence. The fourth is the axis, arithmetic or logarithmic, which on a long move is not a cosmetic setting but a change to the shape of the data itself. The fifth is tolerance: how far a bar may poke through before you call the line broken, and whether you measure that in points, in a percentage, or in a multiple of recent range.

Read that table as a whole and one point becomes hard to avoid: there is no column of right answers. There are two columns of defensible answers, and the practical consequence is that the set of lines a reasonable person could draw on any chart is not a single line but a family of them. Later on this page that family is drawn explicitly, and its width measured. For now it is enough to notice that a chartist who has never made these five choices consciously has still made them, and is watching a level that reflects five decisions taken by habit.

Here is the single most common misreading of the tool, and it follows directly from forgetting what the line encodes. A trendline break is routinely taught as a reversal signal. It is not. The line is a statement about rate, so breaking it is a statement about rate: the floor stopped rising at the pace you assumed. That is entirely compatible with the trend continuing, and in practice it very often does, because the pace of an advance is not constant and no market sustains its opening rate indefinitely.

The figure below is a single uninterrupted uptrend of forty four bars. A line anchored to the lows of bar 3 and bar 13 rises 2.14 points per bar and tracks the first leg cleanly. At bar 22 price closes below it. On the reversal reading, that is where the trend ended. It is not what happened. Price kept advancing for another twenty one bars, and a second line anchored to two later lows, rising 1.27 points per bar, is touched four times and is still intact at the last bar on the chart.

This behaviour is not a quirk of the illustration. It is what you should expect from any advance that begins sharply. Early in a move, positioning is light and the supply that would cap it has not yet appeared, so price can rise quickly. As the move matures, holders take profits into strength and new buyers demand a discount, so the same demand produces a smaller advance. The rate decays. A line fitted to the early, fast portion is therefore almost guaranteed to be broken eventually, no matter how healthy the trend is, and the steeper the line the sooner that happens. A near vertical line is broken by the trend simply pausing.

Which gives a practical rule of interpretation. When your line breaks, the question to ask is not "is the trend over" but "what is the new rate, and is there one". Redraw from the most recent pair of pivots and look at the slope you get. If a gentler line fits the later lows and gets touched, the trend has downshifted and is intact. If no line fits, because the lows have stopped rising at all, then the sequence itself has failed, and that is a different and far more serious event than a line being broken. The sequence is the trend; the line was only ever your estimate of its speed.

The oldest desk rule about trendlines is that one touch is an accident, two is a coincidence and three is a line. It sounds like folklore. It is actually geometry. Any two points whatsoever define exactly one straight line, so the fact that your line passes through two lows carries no information at all: it would pass through any two lows you picked. A two touch line tells you nothing except that you can use a ruler. The third touch is the first piece of evidence, because it is the first time the line made a prediction the market could have refused and did not.

The practical consequence is that at the moment you draw a line, you are usually choosing between several equally defensible ones, and there is no way to tell which is real. The figure below makes that concrete. The left panel is one price series at bar 13. A steep line drawn four bars earlier has just been broken by the close, and three new lines are now available, each running from the fresh pivot low at bar 13 back to a different earlier low. Every one of the four is anchored to genuine pivots and none of them is a bad line. The right panel carries all four forward and lets the market sort them.

Notice the third outcome carefully, because it is the one people forget. A line can fail without ever being broken. Line C is never violated by a single close, so a trader watching it would report that it is holding perfectly, while price trades fifteen points above it and every actual decision happens somewhere else entirely. A line that price has left behind is not a valid line that happens to be quiet. It is a line whose rate hypothesis was too low, and its silence is the evidence. The check is simple: if the line is nowhere near price, it is telling you nothing, and continuing to display it just makes the chart look busy.

This is also where the honest reason to prefer three touches sits. It is not that three is a magic number, and the market has no rule that respects a line the third time. It is that with only two touches you have not yet distinguished your line from the several others you could have drawn through different pairs. The third touch does that work: it is the point at which the rate you fitted has survived a test it could have failed. That is a genuinely different epistemic position, and it is worth the wait, because the price of the wait is a worse entry and the price of skipping it is acting on a ruler.

Of the five choices, the axis is the one that produces the most genuine and most invisible disagreement, because it changes the data before you ever pick up the ruler. On an arithmetic axis, equal rupee moves occupy equal vertical space. On a logarithmic axis, equal percentage moves do. A trend that compounds at a steady percentage rate therefore plots as a curve on the first and as a straight line on the second, and since a trendline is by definition straight, only one of those two pictures can put all the lows on it.

The figure below plots one series twice. It advances at a steady 2.8 percent per bar with irregular pullbacks, and in both panels the line is anchored to the lows of exactly the same two bars, bar 6 and bar 20. On the log axis, the same rate applies at every price, so the line runs through the lows at bars 6, 13, 20, 27 and 34: five touches, an unusually well confirmed line by any standard. On the arithmetic axis, that identical pair of anchors gives a straight chord that cuts through the middle of the move and is then left behind by it entirely.

Two practical points follow. The first is that the arithmetic line here is not merely less useful, it is badly drawn in a way you would reject on sight if you noticed it: a straight chord between two lows on a convex path necessarily cuts through the price in between, which it does at five separate bars. That defect is invisible in the moment because the chartist drew the line at bar 20 looking left at two lows that did lie on it. The curvature only became apparent as the move extended. Any long trendline on an arithmetic axis is quietly accumulating that error.

The second is about which axis to prefer, and the honest answer is that it depends on the length of the move rather than on doctrine. Over twenty or thirty bars in a normal instrument the difference is negligible, and the arithmetic axis has the real advantage that your risk is denominated in rupees, so the chart's geometry matches the geometry of your position size. Over a move that doubles or triples, the arithmetic axis stops describing the structure at all and the logarithmic one is the only version in which a constant rate is a straight line. What matters far more than the choice is stating it, because a line is only valid on the axis that produced its touches.

A channel is a trendline plus a parallel copy of itself, offset to sit against the opposite extreme of the same move. Draw the support line under the rising lows, then drag a duplicate up until it rests on the intervening highs, and you have bracketed the advance. What the second line adds is a rough measure of the move's normal amplitude: how far above the rising floor price has typically been willing to trade. If the channel holds, the upper line is a reasonable place to expect supply and the lower one a reasonable place to expect demand.

Everything on this page applies to a channel with the subjectivity doubled, which is the only reason it is worth mentioning here at all. The lower line inherits all five drawing choices. The upper line is then constrained to be parallel, which sounds like a discipline but is really an additional assumption: that the move's amplitude is constant, which markets do not promise. In practice the upper line touches nothing exactly, and chartists quietly shift it to whichever high makes the picture look tidiest.

One genuinely useful observation comes out of channels, and it is about breaks in the other direction. When price accelerates through the top of a rising channel, the rate has increased rather than failed, and that is frequently a sign of a move entering its final and most emotional phase rather than its healthiest one. That reading is a tendency and not a rule, and it is the sort of context that belongs alongside other evidence rather than acted on alone.

Given all of the above, it is worth being precise about the informational content of a trendline break, because the honest version is narrower than the usual version and considerably more useful. A decisive close beyond your line tells you one thing: the pace you fitted no longer holds. Everything else people read into it is inference, and each piece of inference deserves its own evidence.

The discipline that makes the reading survivable is to insist on a close rather than a touch. Intraday, a level that many people are watching attracts probing, and a wick through a trendline is the single most common way a line is briefly violated and then defended. Judging by the close does not eliminate false signals, because a close can be reversed too, but it does filter out the cheapest ones. The same trap logic that makes a horizontal breakout fail applies to a sloping line, with the added complication that your sloping line is at a slightly different price from everybody else's.

The last two rows are the ones that separate this tool from the rest of the toolkit, and they have no equivalent for a horizontal level. When a prior high fails, every chart in the country agrees that it failed. When your trendline fails, you have learned something about your line, and whether the market agrees depends on how many other participants happened to make the same five choices you did. On a heavily watched, obvious line drawn between two unambiguous major lows, that overlap can be large, and the level behaves as though it were public. On a line drawn through minor pivots that only you consider significant, the overlap is close to zero, and the break is an event inside your own analysis.

The clearest way to calibrate what a trendline is worth is to set it next to a level that is not drawn at all. A pivot point ladder is computed from the previous session's high, low and close by a fixed formula. There are no anchors to select, no wicks-or-closes question, no axis to argue about and no touch count to satisfy. Everybody who runs the arithmetic on the same session gets the same numbers to the paisa, which means that when price closes through a pivot level, every chartist watching agrees on the fact of it. Whether that fact matters is a separate question, and one that page handles; what matters here is that the fact is shared.

The same is true of the other objective references. A prior swing high is a number that occurred. A previous day's close is a number that occurred. A supply or demand zone occupies a range rather than a line, but it is a range anchored to bars that actually printed, which means two people marking it will land within a candle or so of each other. A trendline is the only common object on a chart whose position depends on decisions taken by the person drawing it.

None of this makes the trendline the weakest item on the list. It makes it a different kind of item. A computed level tells you where an unambiguous thing happened; it says nothing about pace, and a ladder of horizontal lines cannot represent a trend that is decelerating. A trendline is the only one of these that encodes a rate, which is precisely why it is worth drawing despite everything above. The mistake is not using it. The mistake is using it as though it belonged in the first two rows of that table.

What follows from all of this is a short and slightly deflating set of working rules. They do not make a trendline objective, because nothing can. They make it useful in spite of not being objective, which is the achievable goal.

Before anything else, look at how wide the family of defensible lines actually is on your chart. The figure below takes one series, picks its three pivot lows, and draws every line those pivots allow: three pairs, each anchored once on the lows and once on the closes, giving six lines that are all live at the final bar. The shaded region between the highest and the lowest is not a decoration. It is the envelope of the six, computed at every bar, and it is the honest picture of where "the trendline" is.

The width of that band is the first working rule in visual form. Treat the line as a zone, not a wire. If the reasonable family spans two or three points, then a close half a point through your line is inside the noise of the drawing itself and is not evidence of anything. Demand a margin that is larger than the disagreement among lines you would have accepted. In practice this means judging breaks against a band roughly the width of a recent bar's range rather than against a coordinate.

The second rule is to draw on the closes as your default, and to know why you are doing it. The close-anchored line is the more conservative of the two: it sits lower on an uptrend, so it triggers later and less often. It also matches the rule you should be using to judge the break, which keeps one convention rather than two. The wick-anchored line is not wrong and there are good arguments for it, but if you are going to pick one and stop thinking about it, pick the one whose false alarms are rarer.

The third is to wait for the third touch before you treat the line as anything more than a sketch, and the fourth is to expect to redraw. A trendline is not a permanent feature of a chart. It is a current estimate of a rate, and rates change while trends continue. A chartist who has redrawn the same trend's line three times as it decelerated has done the job correctly; a chartist still displaying the original line from fifty bars ago is looking at a historical artefact.

Read plainly, a trendline does one job well. It converts a vague impression that something is rising into a specific, falsifiable claim about how fast, and it puts that claim on the chart where you have to look at it. That is genuinely valuable, because a rate is the one property of a trend that no horizontal level can express and no single bar can reveal. It is also the reason the tool survives despite being the least objective thing on the chart: nothing else does this job at all.

The failure mode is to forget which half of the object is the market's and which half is yours. The sequence of higher lows belongs to the market. The rate you fitted to it belongs to you, and so does the break. When you keep those separate, a trendline break becomes a prompt to re-examine the sequence rather than an instruction to act, and the question it raises is the useful one: has the pace changed, or has the structure failed? Those are different events with different consequences, and the line cannot tell them apart on its own. That kind of separation, between what the chart shows and what you have added to it, is the reading discipline that the method we teach is organised around, and it applies far beyond this one tool.

Used that way, a trendline sits comfortably alongside the rest of the toolkit rather than competing with it. It is the rate instrument. The computed ladders and prior extremes are the shared, unambiguous references. Zones describe where imbalances printed. Reading a chart well means knowing which of those you are looking at, and a good starting point for assembling the whole picture is the wider survey in technical analysis for beginners. What none of them do, including this one, is tell you what happens next.

• Robert Rhea, The Dow Theory (Barron's, 1932). The formal codification of Charles Dow's editorials from the turn of the twentieth century. It is the source of the structural definition used on this page: a trend is a sequence of higher highs and higher lows, or of lower highs and lower lows. Note that the definition contains no line and no slope, which is exactly why a trendline adds a rate estimate rather than a measurement.
• Robert D. Edwards and John Magee, Technical Analysis of Stock Trends (Stock Trends Service, 1948). The classical treatment of trendline construction that most later writing descends from: the line rests on the extremes rather than cutting through price, a third touch validates what two touches only assert, the line is extended forward into empty space where the next test will occur, and the parallel channel is built from a single line offset to the opposite extreme.
• The arithmetic of price scales. On an arithmetic axis, equal absolute moves occupy equal vertical space. On a logarithmic axis, equal proportional moves do. A series compounding at a constant rate is therefore a curve on the first and a straight line on the second, which is why the same two anchors can produce a line that is touched five times on one axis and cuts through the price on the other. The contrast in this guide is worked out on the plotted series rather than asserted.
• Exchange session mechanics. A daily bar on an NSE or BSE instrument compresses the 09:15 to 15:30 continuous session into four numbers, and the pre-open call auction sets the opening print. Which of those four numbers you anchor a trendline to is a choice, not a convention the exchange supplies, and it is the first of the five judgement calls set out on this page.
• The figures on this page. Every price series here is synthetic and illustrative, generated for this guide, and is not any real instrument. Every trendline shown is computed from that series' own pivot anchors rather than drawn by eye, and every verdict printed on a figure, the slopes, the break bars, the touch counts, the band width and the percentages, is read mechanically off the plotted bars. No outcome is scored and no success rate is implied or measured.