Trendlines and Price Channels: Draw Them by Rules, Not by Feel

A trendline is a straight line through the pullback lows of an uptrend, or the bounce highs of a downtrend. It shows whether a trend is still moving at the same slope, and it gives you a place for a Stop loss. The problem is that two people looking at the same chart often draw different lines. So this article focuses on rules you can repeat and get the same line every time.

Drawing a trendline by rules

First decide what counts as a “low”. This article uses a swing low: a bar whose low is lower than the 3 bars on its left and the 3 bars on its right (similar to Bill Williams’ fractal, which uses 2 bars on each side).

1. Uptrend: join two swing lows where the second is higher than the first. Downtrend: join two swing highs where the second is lower than the first.
2. Two points can always be joined, so they confirm nothing. Wait for a third touch: the next swing low comes back to within your tolerance of the line.
3. Set the tolerance as a fraction of ATR (the average bar range, see ATR explained), for example within 0.3 ATR counts as a touch. A tolerance like this adjusts to volatility by itself.
4. Wicks or bodies: choose one. Wicks are prices that really traded. Bodies are not dragged off course by one long wick. Both work, but if you switch between them to make the touches look clean, the line stops measuring anything.

One point people forget: a swing low needs 3 bars on its right, so you only know it is a swing low 3 bars later. A line that looks like a perfect touch on a past chart was not yet known as a low when it happened.

Worked example: a trendline and channel on a sample chart

Take an illustrative 90-bar series at gold-like price levels (not real market prices). The swing lows are at bars 13, 31, 48, 67 and 84.

Step	Calculation	Result
Point ①, bar 13	Low	2,402.6
Point ②, bar 31	Low	2,418.2
Slope	(2,418.2 − 2,402.6) ÷ (31 − 13)	0.867 per bar
Line at bar 48	2,402.6 + 0.867 × 35	2,432.93
Distance to bar 48 low (2,434.2)	1.27 ÷ ATR 4.76	0.27 ATR → third touch
Distance to bar 67 low (2,450.0)	0.60 ÷ ATR 4.38	0.14 ATR → touches again
Distance to bar 84 low (2,468.1)	3.97 ÷ ATR 4.24	0.94 ATR → not a touch

ATR here is Wilder’s 14-bar ATR at that bar. Bar 15’s wick dips 0.73 below the line, which is still inside the tolerance (0.3 × 4.09 = 1.23), so it does not count as a break. This is exactly what the tolerance is for: one wick doesn’t force you to redraw the line.

A channel is the trendline shifted up in parallel so it passes through the high farthest from the line, counting only bars before your decision point. In the example, the farthest high from bar 13 to bar 51 is bar 40 (2,446.8), which sits 2,446.8 − 2,426.0 = 20.8 above the line. So the upper edge runs 20.8 above the trendline at every bar.

Using the line for stops and targets

Continuing the example, the bar 48 swing low is confirmed at bar 51 (once 3 bars on its right have closed). Suppose you decide at the close of bar 51.

1. Close = 2,439.9. Trendline at bar 51 = 2,402.6 + 0.867 × 38 = 2,435.53
2. ATR 14 = 4.54. Stop loss half an ATR beyond the line = 2,435.53 − 0.5 × 4.535 = 2,433.27
3. Risk distance = 2,439.9 − 2,433.27 = 6.63
4. Target at the channel’s upper edge = 2,435.53 + 20.8 = 2,456.33. Reward distance = 16.43
5. Risk:Reward = 16.43 ÷ 6.63 = 1 : 2.48

The Stop loss buffer (0.5 ATR) is wider than the touch tolerance (0.3 ATR), so ordinary wicks that poke the line, like bar 15, don’t hit the stop. An uptrend line rises every bar, and so does the channel’s upper edge. If you hold longer the target moves up, so using the value at the entry bar is the cautious figure.

Size the position from the risk distance. Say the account is 10,000 dollars and you risk 1% = 100 dollars. With a 100-ounce contract (common for gold, but check your broker’s contract specification), a 1-dollar move at 0.01 lot = 1 dollar. That gives 100 ÷ 6.63 = 15.1 steps of 0.01 lot, about 0.15 lot. The full method is in Size your lot from Stop loss distance.

Log scale on long charts

On a normal (linear) scale, 100 → 200 and 200 → 300 are the same height. On a log scale, equal heights mean equal percentages: 100 → 200 and 200 → 400 are the same height, because both are +100%.

A simple example: price goes from 100 to 200 over one period. Over an equal second period, a straight line on a linear scale points to 300, but on a log scale it points to 400. That’s a 100-point difference.

A simple rule: if price on the chart you are reading has changed by several tens of percent or more, such as a weekly chart covering years, or a crypto coin that has moved several-fold, switch to a log scale (some platforms, such as TradingView, have a log toggle; some don’t). On an hourly chart covering a few days, price changes only a few percent and the two scales look almost the same.

Breaks and retests

Define a break as a rule too, for example a close below the line by more than the tolerance, not just a wick. After a break, price often comes back to test the line from the other side, called a retest. Sometimes the old line acts as resistance. Sometimes price simply climbs back above it.

Why a broken trendline is not automatically a reversal

A trendline measures the slope of the move. A break only tells you price is no longer rising at that speed. Three things can follow:

• A reversal into a downtrend: a lower high, then a break of the latest swing low
• A range: price swings sideways
• A slower uptrend: a new, shallower trendline

A reversal in terms of price structure needs a lower high and a lower low. A trendline break alone is not there yet. For more on markets changing character, see Why trading systems stop working.

A broken trendline means the trend is no longer moving at the same slope, not that the trend is over.

Steepness and redrawing traps

• Too steep. A line drawn from two lows during a sharp acceleration is so steep that price can’t keep up for long. It breaks fast, and most of those breaks just mean price went back to a normal pace.
• Redrawing after every break. If you keep moving the line so it is “not broken yet”, it can never be wrong, which means it measures nothing. Set rules in advance, for example the line ends on a close beyond the tolerance, and a new line may only use swing lows that formed after the break.
• Several timeframes. An hourly line can break while the daily line is still fine. Decide in advance which timeframe makes the decision. See Multi-timeframe analysis.

Three-touch lines happen by chance, often

We simulated pure random-walk prices (no trend) of 200 bars, 10,000 times, using every rule from this article: swing lows with 3 bars each side, join two consecutive rising swing lows, then check whether the next swing low touches the line within 0.3 ATR with no bar breaking below the line by more than the tolerance in between.

• 59.0% of runs contained at least one uptrend line with three touches
• With only 100 bars, it was still 30.4% of runs
• Of all rising pairs of lows, 12.4% got a third touch

The exact percentages depend on small details (how each bar is generated, how a break is counted): a second run with a slightly different reading of the same rules gave about 40% for 200 bars instead of 59%. The point stays the same.

Nobody placed orders at any line in random prices, yet three-touch lines still showed up in a large share of runs. And this is a strict rule that only uses consecutive lows. If you can pick any pair of lows you like, you’ll find them even more often. It’s the same story as support and resistance that price touches by chance; see Support, resistance and supply/demand zones and Why past charts always look easy.

A third touch makes a line more believable, but it doesn’t prove the line has power. Use it to place stops and measure slope rather than as a prediction.

This is a simulation of random prices, not real market data. It doesn’t say trendlines are useless. It only says that a neat-looking touch on a past chart is not evidence yet. To find out whether your own rules have an edge, write signals down before price gets there and count both hits and misses. Tests of other chart patterns are in Chart patterns, tested.

Common mistakes

• Treating a line through two lows as important right away, when two points can always be joined
• Switching between wicks and bodies, or skipping lows that don’t fit, to make the touches look perfect
• Redrawing after every break until the line is never wrong
• Putting the Stop loss exactly on the line and getting taken out by ordinary wicks
• Opening the opposite side as soon as the line breaks, before any lower high and lower low
• Using a linear scale on a multi-year chart where price has moved several-fold
• Forgetting that a swing low is confirmed 3 bars later than it appears on a past chart