Line Chart Maker

Paste your X values on the left and your Y values on the right — commas, spaces, newlines, or a pasted spreadsheet column all work — and the chart draws instantly in your browser. Nothing is uploaded, and the camera icon in the chart toolbar saves a PNG.

What the Line Between Two Points Actually Claims

A line chart makes a statement that a table of the same numbers does not. Every drawn segment asserts that the two points it joins are connected: that the quantity travelled continuously from one to the other, and that a value existed at every moment in between. That single implication decides where a line chart belongs.

It belongs on anything measured repeatedly along an ordered, evenly spaced dimension. Monthly revenue, daily temperature, weekly active users, dose against response, distance against fuel used. In every case the reader can point at a spot between two markers and ask "what was it here?", and the question has an answer.

It does not belong on unordered categories. A line joining sales figures for France, Japan, Brazil and Canada implies a path from France to Japan, and there is no such path — the slope of that segment means nothing, and reordering the countries would produce a completely different picture from identical data. Categories need a bar chart, where the bars have no relationship to each other except height.

The same test rules out one subtler case: ordered but unevenly spaced readings. If you measure on days 1, 2, 3 and then day 60, plotting them as four evenly spaced points compresses two months into the same width as one day. Use the real X values (1, 2, 3, 60) so the horizontal distance stays proportional to elapsed time.

Getting Your Data Into Shape

Both inputs read numbers. Tokens that are not numbers are skipped and reported as "values skipped" under the box, which is your signal that something was lost. Three practical consequences:

• Dates must become numbers. Years work as-is. For months, use 1–12 and put the year in the chart title.
• The two lists must be the same length, because each X is paired with the Y at the same position. A mismatch shows an error rather than guessing.
• Order is preserved, not sorted. The line is drawn in the sequence you supply, so unsorted X values make it double back.

If you paste two columns from a spreadsheet, the tool detects the table, offers a column picker, and lets you choose the header row — paste the same block into both boxes and select the X column in one and the Y column in the other.

Should the Y Axis Start at Zero?

Bar charts must start at zero, because a bar encodes its value as a length: half the bar means half the value, and cropping the baseline breaks that arithmetic. A line chart encodes value as vertical position instead, and reads as a shape. Cropping the axis to the data range is therefore normal practice, not a trick — a chart of body temperature or a stock price starting from zero would be a flat, useless band across the top.

Truncation becomes deceptive at the point where the picture invites a conclusion the numbers do not support:

• When the reader is meant to judge a ratio. If a chart makes 51% look like triple 49%, the axis has done the arguing.
• When the change is inside the noise. Plotting a 99.2%-to-99.1% uptime dip on an axis running 99.0 to 99.5 turns ordinary variation into a cliff.
• When you switch on Fill area. The fill runs down to zero, so the shaded region is only honest if zero is a real floor and the whole distance from it is meaningful.

The workable rule: crop freely, but let the axis labels be legible enough that a reader can see the range for themselves, and don't crop a chart whose whole point is how big something is relative to nothing.

Aspect Ratio Is Part of the Argument

The same series looks urgent in a tall narrow frame and placid in a wide flat one, because stretching the vertical axis multiplies every slope on the page. Nothing about the data changed; the drawing did.

William Cleveland's rule for choosing deliberately is banking to 45 degrees: pick the width-to-height ratio that puts the typical line segment near a 45-degree slope. Slopes far shallower than that read as flat and hide real turns; slopes far steeper compress the horizontal detail into a wall. Applied to the sunspot cycle, banking revealed that the cycles rise faster than they fall — a feature invisible in the tall version everyone had been printing.

Charts exported from this page are written at 1000 × 600 pixels, a 5:3 frame that suits a series with a moderate overall trend. If your exported chart looks like a spike or a flatline, that is worth a second look before you draw a conclusion from it.

Gaps, and Why the Line Lies About Them

A sensor drops out for a week. A shop closes for a public holiday. There are three things a chart can do with the hole, and only two of them are defensible:

1. Leave a gap. The most honest: absence of data is shown as absence of line.
2. Bridge it. Acceptable when the quantity genuinely continued and you say so.
3. Plot zero. Almost always wrong. It claims a measured value of nothing, and a single false zero drags the line to the floor and distorts every trend read off the chart.

This tool has no gap mode — it connects consecutive points — so a missing period is bridged silently. Because non-numeric cells are skipped during parsing, the practical failure is worse than a bridge: a blank in your Y column shortens that list, and every later Y shifts up to pair with the wrong X. Delete the matching X value, confirm both boxes report the same count, and note in the title that a period was interpolated.

When One Line Becomes Five

A chart with two or three lines is a comparison. A chart with eight is a plate of spaghetti: lines cross, colours become impossible to hold in memory, and the reader spends their attention on the legend rather than on the data. The usable ceiling is around four or five series, and it drops fast if the lines overlap.

Past that, three alternatives work better:

• Small multiples — the same chart repeated once per series, in a grid, sharing one Y range. Twenty series stay readable this way; twenty lines do not.
• Highlight one, grey the rest. Draw the series you're discussing in colour and the others in light grey as context. Repeat with a different highlight for each point you want to make.
• A slope chart, if only two time points matter: plot start and end, join each pair, and let the direction of the lines carry the story.

Because this page draws one series at a time, small multiples are the natural workflow — generate each chart with the same options and put the PNGs side by side.

Twelve Months of Orders

Monthly orders for a small online shop, X as month index 1–12:

X: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12
Y: 412, 388, 455, 501, 478, 522, 495, 540, 610, 688, 902, 1145

The headline is a rise from 412 to 1,145 orders, about 2.8× or +178% across the year. The chart says something more specific. The first eight months average 474 orders and never leave the 388–540 band — that stretch is noise, not growth, however much a mental straight line through the whole series suggests otherwise. The real movement is in the last quarter: November alone jumps 27% over October, and November and December together account for 457 of the year's 733 additional orders, roughly 62% of all the growth.

That shape is a seasonal spike sitting on a flat baseline, and it changes what you would do next. A straight trend extrapolated into January predicts around 1,200 orders; the seasonal reading predicts a fall back toward 500. To tell which, plot the previous year on the same scale as a second chart and look for the same November step. Switching on Show markers helps here — it makes the twelve actual readings visible so nobody mistakes the smooth stretch between month 9 and month 12 for more data than there is.

For a chart of how the same values are distributed rather than how they moved, use the histogram maker; to check whether two measured quantities move together at all, the scatter plot maker is the right starting point.