The same two numbers, told two ways
Two candidates polled at 50.9% and 49.1% — a 1.8-point gap. Both charts below show exactly those numbers. Only the starting point of the vertical axis differs.
Chart 1 — axis starts at zero
"Race in dead heat"
Chart 2 — axis starts wherever you drag it
"Candidate A surges to commanding lead"
Neither chart lies about the numbers — one lies about their meaning. Truncating the axis stretches a 1.8-point difference until the bars imply a landslide. In the version starting at 48, Candidate A's bar is more than 2× the height of B's, for a difference of under two points.
This is the single most common chart manipulation in the wild — cable news graphics, earnings decks, product comparisons. The tell: look at where the axis starts before you look at the bars. Bar charts encode value as length, so a bar chart with a truncated axis is always overstating the difference; that's why the zero-baseline rule exists for bars specifically.
The polling numbers are invented for illustration; the distortion mechanics are exactly what real published charts do.