What auto-scaling actually does
Auto-scale sets the top of the price axis just above the highest visible price and the bottom just below the lowest, then spreads the whole frame between them. It is on by default nearly everywhere, and for good reason: without it you would spend your time dragging the axis to find the price.
The cost is that the axis is no longer a fixed ruler. It is a result — of which candles happen to be on screen, at the zoom you happen to be at, after the last scroll you happened to make.
That means the vertical size of a move on your screen carries no information about the size of the move. It tells you how that move compares with the largest and smallest prices currently visible beside it, and nothing more.
Six prices, fitted to themselves
An illustrative share's closes on one day. Switch between the axis auto-fitting them and a window a reader would recognise.
Both views plot the same six prices. The caption in the corner carries the axis range and the real move.
Four ways this misleads a careful reader
A flat day becomes a dramatic one. Seventy paise of range on a ₹500 share is a share that did nothing. Auto-fitted, it is a chart full of peaks and collapses, and it will be read as a struggle.
Scrolling changes the past. Widen the view to a year and the same candle shrinks to a hair. Nothing about that day changed; the tallest thing on screen did. Two screenshots of one candle at two zoom levels are not comparable evidence.
A high on screen may not be a high. Auto-scale fits what is visible. A price at the top of your window is the highest price in your window — which says nothing about the months sitting just off the left edge.
One wrong number flattens a year. If a single bad print shows an illustrative share at ₹900 in a year it actually traded between ₹90 and ₹130, the axis now runs to ₹900. The genuine ₹40 range is squeezed into under 5% of the height, and a year of real movement reads as a flat line with one spike. The chart is behaving correctly. The data is not.
The zero that is missing, and should be
A price axis almost never starts at zero. In most kinds of chart that would be a warning sign — a truncated axis on a bar chart of quantities exaggerates small differences, and is a standard way to mislead.
On a price chart it is normal and correct. A share that trades between ₹480 and ₹520 would occupy 8% of a zero-based axis, and the other 92% would be empty. Nobody is served by that.
But the consequence carries over. Height on a price chart is not proportional to price. A candle drawn twice as tall as another is not twice the price, and a line halfway up the frame is not half of anything. Every reading has to come from the axis labels, because the geometry deliberately does not carry it.
- First close ₹500.20, last close ₹500.80
- +₹0.60
- Highest ₹500.80, lowest ₹500.10
- range ₹0.70
- The move, as a percentage
- +0.12%
- Axis range when auto-fitted
- under ₹1
- Axis range in a ₹495 to ₹505 window
- ₹10
You compare two screenshots of the same share, taken an hour apart, and the second shows far larger swings. What is the first thing to check?
Which of these is a real defence against a misleading auto-scaled chart?
Scale type, axis range, time span — then the shape. Chapter 03.01 asked what a chart type keeps and discards. This chapter adds that even a chart keeping everything can still be drawn into any impression you like, by an axis rule, a window shape, or a range that fitted itself to whatever was on screen. Read the four settings first and the picture becomes evidence. Read the picture first and the settings become invisible.
Every distortion in this chapter was a drawing decision, made on data that was correct. The next chapter asks the harder question: whether the numbers being drawn are the right numbers at all — starting with what happens to a chart's history when the company splits its shares.