WhitmanTrading

How to Draw Support and Resistance

To draw support and resistance, mark the prices where the market actually turned, draw each as a band rather than a single line, and require two touches before treating it as a level. Start on the slower timeframe, and delete anything that fails the two-touch test.

Drawing levels is easy and drawing few levels is hard. Almost every problem people have with support and resistance is a volume problem — too many lines, not the wrong lines.

Before you start

A bare chart with every indicator removed. Levels are read from price. Anything else on the screen competes for the same attention and adds nothing to this task.

Two timeframes of the same instrument. A slow one to find the levels and a faster one to trade them. Drawing only on the fast chart produces levels nobody else can see.

A rule for deleting lines you will actually apply. Decided before you start, because deletion never feels urgent in the moment and the chart silently fills up.

The steps

1. Mark where price actually turned

A candlestick chart with turning points marked.
Mark where price turned, not where you wish it had. Illustrative chart - not real market data.

Swing highs and swing lows. The place the market reversed, not the place a trade would have been convenient. Work left to right so the outcome cannot influence the marking.

2. Draw a band instead of a line

Price bars with a level drawn as a band.
Draw a band, because price does not respect a line. Illustrative chart - not real market data.

Cover the wicks and the closes in that area. Price reacts to a region, and a single line is a precision the market has never once honoured.

3. Require two touches before it counts as a level

A candlestick chart of the site's shared price history.
A level needs two touches before it is a level. Illustrative chart - not real market data.

One turn is an event. Two turns at the same area is a level. Delete every line with a single touch, however dramatic that single turn looked.

4. Start on the slower timeframe

A long-horizon candlestick view of the same price series.
Draw them on the slow chart first. Illustrative chart - not real market data.

A weekly level is watched by more participants than a five-minute one. Draw those first, then drop down and keep them on the chart.

5. Add the levels nobody had to draw

A candlestick chart with a volume histogram beneath it.
The busiest price is a level nobody had to draw. Illustrative chart - not real market data.

The price with the most volume traded at it, the prior day’s high and low, and the session open. These exist without anyone’s opinion.

6. Mark the round numbers separately

A candlestick chart with round-number levels marked.
Round numbers act as levels because everyone sees them. Illustrative chart - not real market data.

Whole figures attract orders because they are easy to name. Mark them in a different colour so you never confuse a psychological level with a structural one.

7. Watch the ones that break and then hold from the other side

Price bars with a broken level being retested.
And watch the ones that break and hold from the other side. Illustrative chart - not real market data.

Old resistance holding as support is the strongest confirmation a level was real. It survived a test from both directions, which most lines never get.

8. Delete until every line has earned its place

The second half of the long price series.
If every bar touches a line, delete half the lines. Illustrative chart - not real market data.

Apply the rule you wrote before step one. A chart where price is always near something has stopped being able to tell you anything.

How to tell it worked

Count the lines on your chart after 20 minutes of work. Six or fewer on a single timeframe is a chart that can be read. More than 12 and the levels have stopped carrying information, because something is always nearby and nothing can therefore be wrong.

A candlestick chart with session boundaries marked.
Session highs and lows are levels with a timestamp. Illustrative chart - not real market data.

Then check the two-touch rule across all of them. Every surviving line should have at least 2 touches, and any line with 1 is a candidate for deletion regardless of how convincing it looks.

Finally, check your band widths against the bars. On the shared series, bar ranges run from 0.17 at the tenth percentile to 1.101 at the ninetieth — a 6.5-fold spread. A band narrower than a typical bar will be crossed by ordinary movement, which means it cannot be tested.

Why the band width is the hard part

A candlestick chart annotated with the round-trip cost.
Every level you trade costs 2% of a typical bar. Illustrative chart - not real market data.

A fixed band width is wrong in both directions. The same number is enormous during a quiet stretch and invisible during a volatile one, and the 6.5-fold range between the tenth and ninetieth percentile bar is the measurement of exactly that problem.

Size the band to recent bars rather than to a constant. A band roughly the width of a typical recent bar is defensible, and it changes as conditions change — which is the point.

And remember every level you act on costs. A round trip on the shared series is 2% of a median bar’s range, so a chart producing many marginal signals is expensive by construction.

The original data

Of the 24,971 unique videos in research/search-study-corpus.jsonl, 41 have an instruction-shaped title mentioning support and resistance, at a median of 51,611 views across 38 channels, with a maximum of 1,975,589. That is the fourth-highest instructional median measured. The counts come from site/rank_howto.py.

A rising stretch of the price series cut short at the decision bar.
Price is sitting on the line. Is it holding? Illustrative chart - not real market data.

Thirty-eight separate channels made those 41 videos, which is unusually distributed — most subjects in this corpus are dominated by a handful of channels. A wide spread of creators at a high median means the demand is broad rather than driven by one popular upload.

The answer to the question on that chart is that the line cannot tell you. On the shared series, measured breakouts followed through in 85% of 39 twenty-bar events and in all 11 fifty-five-bar events, so a level being touched says very little on its own. What decides it is whether the band was drawn before price arrived — a line adjusted while you watch is describing the past.

When it fails

A sideways, range-bound candlestick series.
Too many lines and every bar is at one. Illustrative chart - not real market data.

The dominant failure is volume of lines rather than accuracy of them. Once a chart carries twenty levels, price is permanently near one, every move can be explained afterwards, and nothing can be disproved. That is not a chart with more information on it; it is a chart with none, and it feels more thorough than the correct version, which is why it persists.

The second failure is a band narrower than a typical bar. Ordinary noise crosses it constantly.

A third is drawing only on the fast timeframe. Those levels are visible to almost nobody else.

A fourth is moving a line after price approaches it. The line then records what happened rather than predicting anything.

A fifth is treating a single dramatic turn as a level. One touch is an event, and the two-touch rule exists precisely because dramatic single turns are the most tempting to keep.

And a sixth is never deleting. Levels from six months ago on a five-minute chart are archaeology.

Support and resistance explains why levels hold at all and separates the mechanism from the memory. Supply and demand is the related idea built on departure rather than on repeated touches. And breakout covers what happens when one of these levels stops holding, which is the other half of drawing them.

What I actually do

The single change that improved this for me was deleting more than I drew. A chart with thirty lines on it means every bar is near one, which makes the levels unfalsifiable — whatever happens, something was nearby. Cutting to the five or six that had genuinely turned price twice made the chart less comfortable and much more useful.

— Michael Whitman

This page is educational, not financial advice. Test every idea on your own charts before risking money.