Trendlines and channels — the line that moves under you
Every trendline you have ever seen was drawn by someone who already knew how the chart turned out. That is not cheating — it is simply what a two-point line is. The trouble starts when the line gets treated as though it were a level. Levels sit still. Trendlines do not: they keep climbing at a fixed rate whether or not anything happens, which means a trendline can be broken by the calendar rather than by sellers. This lesson gives you the drawing rules, then puts a number on that drift so you can see exactly how much of a “break” was really just time passing.
KEY TAKEAWAYS
- Two anchors make a hypothesis; the third reaction is the first evidence. Any two swing lows on any chart have a line through them, so finding one proves nothing.
- Diagonal lines move, horizontal ones do not. Freeze price at $120.40 for ten bars and a $1-per-bar line goes from $0.40 above it to $9.60 below — a 7.38% break with zero price movement.
- Scale is not cosmetic. Pinned to the same two lows, a linear and a logarithmic line sit $3.10 apart by bar 34 — 2.38% of price, enough to decide whether the line is broken.
- A parallel channel keeps its dollar width and loses its percentage width. The same $8.00 gap is 6.96% of the lower rail at bar 19 and 6.15% at bar 34 — 11.6% less reward per lap, with fees unchanged.
- Before reacting to any trendline break, ask which side moved. Price falling to the line is news about sellers; the line rising to price is news about your slope.
What is a trendline, and how is it different from a horizontal level?
A trendline is a straight line drawn through two swing lows in an uptrend, or two swing highs in a downtrend, then extended forward. A horizontal level is a fixed price. That sounds like a trivial difference. It is the most important sentence in this lesson.
Write them as formulas and the gap becomes obvious. A horizontal level is f(t) = $120.00. It takes no argument. Whatever else happens, tomorrow it is still $120.00. A trendline is f(t) = anchor + slope × t. It takes time as an input, so it has a value tomorrow that it does not have today, and that value arrives whether or not a single trade prints.
Every practical consequence in this lesson falls out of that one extra term. A level can only be broken by price moving to it. A trendline can be broken by price moving to it or by price standing still while the line moves to price. Most guides to trendlines never mention the second route, which is a shame, because it is the one that catches people.
Both tools are describing the same underlying thing — areas where the balance between buyers and sellers previously changed, as covered in Lesson 11. The trendline just adds a claim about rate: not merely that buyers showed up, but that they showed up at a steadily higher price each time.
How do you draw a trendline from two points?
Pick two same-type swings, connect them, extend the line forward without touching either anchor, and then wait. The first two points make a hypothesis. The third reaction is the first piece of evidence.
Here is the exact chart used for every number in this lesson, so you can check the arithmetic yourself rather than take our word for it.
Read it as a calculation, not a picture. Point A is the swing low of bar 4 at $100.00. Point B is the swing low of bar 14 at $110.00. The slope is therefore ($110.00 − $100.00) ÷ (14 − 4) = $1.00 per bar, and the line's projected value at bar 24 is $110.00 + 10 × $1.00 = $120.00.
At bar 24 the low prints $120.40 and price turns up. That is $0.40 above the projection, or 0.33%. Whether 0.33% counts as a touch is a decision you have to make in advance, and we come back to it below — but notice that the line was never moved to make it fit. The projection was fixed at bar 14 and the market either respected it or did not.
Two habits matter more than anything else here. First, lock the point rule before you look: either wick extremes or candle-body edges, applied to every anchor in the drawing. Bodies show where trade was accepted, wicks show where it was rejected, and both are defensible — but picking a wick for A and then a convenient body for C is how you draw a line that cannot be wrong. Second, lock the scale and the timeframe, because both change the geometry. The timeframe choice is a cost decision covered in Lesson 10; the scale is dealt with two sections down.
What actually counts as a valid anchor point?
A swing low is a bar whose low is lower than the bar on each side; a swing high is the reverse. Beyond that minimum, the useful test is not "does it look right" but "would I have selected this point before I knew what happened next".
| Criterion | Accept | Reject |
|---|---|---|
| Point type | Low to low, or high to high | A low connected to a high because the angle looked better |
| Separation | Two distinct swings with a counter-swing between them | Two adjacent bars from one turn — that is one reaction seen twice |
| Geometry | The line does not slice through the bodies in between | The line cuts through several accepted closes on the way |
| Selection order | The earliest qualifying swing, then the next qualifying swing | Skipping a nearer valid low because it spoils the slope |
| Consistency | The same wick-or-body rule at A, B and C | Rule chosen after seeing which one produces a touch |
| Confirmation | A third reaction near the fixed projection | The line rotated after the third bar so that it fits |
| Every row in the reject column produces a line that has never been wrong. That is the problem with it, not the appeal. | ||
The reject column has a name: curve fitting. It is rotating a line until recent bars respect it, then presenting the result as analysis. A line discovered after every reaction describes history perfectly and forecasts nothing, and it is very hard to catch yourself doing it, because at every individual moment the adjustment feels like a small correction rather than a rewrite.
One practical guard: write down the projected value before the bar arrives. If the line says $120.00 at bar 24, note "$120.00" somewhere you cannot edit. Then a low of $120.40 is a 0.33% miss you recorded, not a touch you remembered.
Why can a trendline break while price does not move?
Because the line keeps climbing after you draw it. If the slope is $1.00 per bar, the line is $1.00 higher every bar regardless of what price does — so a trade that sits perfectly still slowly ends up underneath a line it started above. This is the mechanic almost no trendline guide states plainly, and it is the reason so many "breaks" happen on quiet days.
Isolate it by freezing one variable. Take the third touch at bar 24, hold price motionless at $120.40, and let only the clock run.
| Bars after the third touch | Trendline value | Price (frozen) | Price vs the line | Price vs a $120.00 level |
|---|---|---|---|---|
| 0 — bar 24 | $120.00 | $120.40 | +$0.40 (+0.33%) | +$0.40, intact |
| 2 — bar 26 | $122.00 | $120.40 | −$1.60 (−1.31%) | +$0.40, intact |
| 5 — bar 29 | $125.00 | $120.40 | −$4.60 (−3.68%) | +$0.40, intact |
| 10 — bar 34 | $130.00 | $120.40 | −$9.60 (−7.38%) | +$0.40, intact |
| Our own model, run on the chart above. Price is held constant on purpose — the whole point is that it contributes nothing to the result. | ||||
Ten bars of absolutely nothing produce a 7.38% breach of the trendline and leave the horizontal level untouched. Not one seller was required. The instrument did not fall; the line rose past it.
Two things follow, and both are worth more than the drawing rules above.
A trendline break is not automatically bearish. Before you react to one, ask which of the two routes produced it: did price come down to the line, or did the line come up to price? A break made of falling candles is evidence that sellers acted. A break made of sideways candles is evidence that the rate of ascent you assumed was too steep — which is information about your line, not about the market's intent. They deserve different responses, and they look identical if you only glance at whether price is above or below.
A trendline stop gets tighter every bar you hold. Suppose you enter at $120.40 with a stop just under the line. At bar 24 the line is $120.00, so the stop sits roughly $0.40–$0.60 away. Do nothing for five bars and the line is at $125.00: the same rule now places your stop above your entry price. A stop that ratchets is not automatically wrong — it is a trailing stop, and some people want one — but you should choose it deliberately rather than inherit it because you attached your risk to a moving object. If you want a fixed invalidation, anchor the stop to a horizontal price such as the bar-24 low of $120.40 or the swing structure beneath it, and let the diagonal line be commentary rather than the trigger. Work the size out from that fixed distance with the position size calculator.
Does the chart scale change where the line is?
Yes, and by more than most people expect. A linear chart spaces equal dollar moves equally; a logarithmic chart spaces equal percentage moves equally. Anchor a line to the same two swing lows on each and you get two different lines that diverge further the longer you extend them.
Run it on our anchors. On a linear chart the line rises $1.00 per bar. On a log chart it rises at a constant rate instead: $100.00 to $110.00 over ten bars is 10%, so the log line multiplies by 1.10 every ten bars.
| Bar | Linear trendline | Log trendline | Gap | Gap as % of price |
|---|---|---|---|---|
| 14 (anchor B) | $110.00 | $110.00 | $0.00 | 0.00% |
| 24 | $120.00 | $121.00 | $1.00 | 0.83% |
| 29 | $125.00 | $126.91 | $1.91 | 1.52% |
| 34 | $130.00 | $133.10 | $3.10 | 2.38% |
| Both lines pinned to the identical swing lows at bar 4 ($100.00) and bar 14 ($110.00). Log values are 100 × 1.10 raised to the power of (bar − 4) ÷ 10. | ||||
By bar 34 the two honest versions of the same trendline are $3.10 apart — 2.38% of price, which is roughly a full day's range on a lot of crypto pairs. A close at $131.50 has broken the log line and is comfortably holding the linear one. Nothing about the market decided that. Your chart settings did.
There is a defensible default. Percentage moves are what actually compound an account, so on anything that has travelled a long way in percentage terms — and most crypto has — the log line is the more faithful description of a constant-rate trend. The linear line quietly demands a decelerating percentage growth rate: $1.00 is 1.00% when price is $100 and only 0.77% when price is $130, so a linear trend that "holds" is a trend that is slowing down. Whichever you pick, pick it once and label the analysis with it, because a line without a stated scale is not reproducible — not even by you next week.
How do you build a parallel price channel?
Copy the trendline without changing its slope and place the copy through a meaningful swing on the opposite side. In an uptrend the lower rail runs under the swing lows and the upper rail is that same slope passing through a swing high. The space between them is the channel.
On our chart the swing high at bar 19 is $123.00, and the lower rail at bar 19 is worth $100.00 + 15 × $1.00 = $115.00. The vertical width is therefore $8.00. Because the rails are parallel, that $8.00 is preserved forever: at bar 34 the lower rail projects to $130.00 and the upper rail to $138.00.
Note what you have just committed to. A parallel channel asserts that the market's volatility is constant in dollars at the same time as its trend is constant in dollars. That is two assumptions stacked, and the second one is much weaker than the first. If you want a midline, put it exactly halfway — $134.00 at bar 34 — and treat it as geometry, not as a magnet.
What does a channel tell you, and what does it quietly hide?
It tells you whether price has been oscillating inside a stable slope and a stable width. It hides the fact that a constant dollar width is a shrinking percentage width, which matters because your costs are charged in percent.
| Bar | Lower rail | Upper rail | Channel width | Width as % of the lower rail |
|---|---|---|---|---|
| 19 | $115.00 | $123.00 | $8.00 | 6.96% |
| 24 | $120.00 | $128.00 | $8.00 | 6.67% |
| 29 | $125.00 | $133.00 | $8.00 | 6.40% |
| 34 | $130.00 | $138.00 | $8.00 | 6.15% |
| Same $8.00 gap in every row. Our own arithmetic on the chart above. | ||||
A rail-to-rail move is worth 6.96% gross at bar 19 and 6.15% at bar 34 — the same trip pays 11.6% less after fifteen bars of trend. Your round-trip fee, meanwhile, is a flat percentage of position value and does not shrink with it, so the ratio of reward to cost erodes quietly the whole way up. Nobody notices, because the channel on screen looks exactly as wide as it did at the start.
Three things a channel does not do. It does not make every rail touch a trade — price can crawl along a rail for ten bars without turning. It does not promise a measured move: breaking above an $8.00-wide channel does not owe you another $8.00, it just means the old geometry no longer contains price. And it does not tell you whether the touch had any conviction behind it; for that you need participation and the state of the book, which is Lesson 8 territory. Rail-to-rail trades also assume you get filled at the rail, which in fast conditions you will not — see slippage.
When is the line invalidated rather than just pierced?
Judge it by closes, not by wicks, and decide the rule before the bar that tests it. A wick through the line means price visited; a close beyond it means price was still there when the bar ended.
Here is one transparent rule you can adopt or replace, offered as a starting point rather than a standard: treat an uptrend line as invalidated after two consecutive closes more than 0.5% below it. On our line at bar 24, 0.5% of $120.00 is $0.60, so the threshold is $119.40. Closes at $119.20 and $118.90 satisfy it. A wick to $119.10 that closes back at $120.30 does not.
Then apply the drift lesson to the rule itself, because the threshold moves too. At bar 29 the line is $125.00, so 0.5% is $0.63 and the threshold has climbed to $124.38. Recompute it every bar, or you will be testing today's price against last week's line without noticing.
Once the line is invalidated, mark it invalidated and leave it on the chart. Do not rotate it. If you want a new candidate, build it from newly qualified swings, and let the failed line stand as a record. Keeping the wrong lines visible is the only thing that makes the right ones mean anything — a chart where every line eventually worked is a chart that has been edited.
Which mistakes turn a trendline into an illusion?
Most bad trendline work is one of five habits, and all five share a structure: they remove the line's ability to be wrong.
- Rotating the line after a breach. The single most expensive habit. It converts a falsifiable claim into a description of whatever already happened.
- Mixing wicks and bodies. Choosing the rule that produces a touch, one anchor at a time, guarantees a line that has always held.
- Switching scale mid-analysis. Worth $3.10 by bar 34 on our chart — enough to flip the verdict on a break. Linear and log are two different hypotheses, not two views of one.
- Reacting to a break without asking what caused it. A break made of sideways bars is news about your slope. A break made of falling bars is news about supply.
- Attaching a stop to the line without meaning to trail. The stop tightens by the slope every bar. Choose that or anchor risk to a fixed price, but do not let it happen by default.
A sixth is subtler and mostly invisible: drawing too many lines. Any chart with enough diagonals on it will have something being touched at all times, at which point the tool can no longer disagree with you. If your chart cannot produce the answer "there is no trendline here", it is decoration.
When is this whole framework the wrong tool?
Trendlines assume price advances at a roughly steady rate. Everything above stops working when that assumption fails, and it fails in four recognisable ways.
When the move is exponential rather than linear. Parabolic advances leave a linear line so far behind that it never gets tested; the log line is the only version with a chance, and even it will be broken while price is still going up.
When the range is genuinely sideways. A market with no rate of ascent has no trendline — only two horizontal boundaries, which is the Lesson 11 toolkit. Fitting a gentle diagonal to a range is how you manufacture a slope that does not exist.
When the timeframe is too fast for the structure. On a 1-minute chart a $1.00-per-bar slope is a demand for 60 dollars an hour, forever. Trendlines drawn on very fast charts are usually measuring the last twenty minutes of noise, and paying full costs to do it — the arithmetic is in Lesson 10.
When there is a scheduled discontinuity. A funding reset, a listing, an unlock, a macro print — these move price by mechanism rather than by trend, and a geometric line has nothing to say about them.
The honest summary: a trendline is a hypothesis about rate, drawn from two points, testable at the third, and invalidated on your own written terms. Treated that way it is a genuinely useful organiser of a chart. Treated as a prediction, it is the most reliably self-confirming tool in technical analysis, because you are holding the pencil.
Frequently asked questions
Are two points enough to draw a trendline?
Two points define a line mathematically, and any two points will. That is exactly why two points are not confirmation of anything: through any pair of swing lows on any chart there is a line, so finding one tells you nothing about the market. The third reaction near the unmoved projection is the first result the line could have failed to produce. Until it arrives you are holding a hypothesis, and it should be labelled that way in your notes.
Should trendlines use wicks or candle bodies?
Either works as long as the choice is made before you look and applied to every anchor. Wicks capture the extremes of where price was rejected; bodies capture where trade was accepted for a whole bar, which is why body-based lines tend to be cleaner on volatile crypto charts with long tails. The failure mode is not choosing wrongly, it is choosing per-point: a wick at A and a body at C produces a line that was always going to fit, and a line that was always going to fit has told you nothing.
Can a trendline be redrawn after it breaks?
You can draw a new candidate from newly qualified swings, but the broken line should stay on the chart marked as invalidated rather than being rotated into a shape that works. The distinction matters because it is the difference between a record and a story. If you keep the failures visible, you can eventually count how often your lines hold, which is the only way to know whether the technique earns its place in your process. If you rotate them, your hit rate is 100% by construction and completely uninformative.
Do channel boundaries predict reversals?
No. A channel identifies two geometric locations that have coincided with turns in the recent past, which is a reason to pay attention, not a reason to expect a turn. Price routinely rides a rail without reversing, exits the channel and keeps going, or drifts sideways until the rails pass it — the same time-drift effect described earlier applies to both rails at once. Treat a rail touch as a place to look for other evidence, and note that a break above an $8.00-wide channel does not owe you another $8.00 of movement.