How support and resistance levels get built
Levels are not properties of price. They are places where a lot of people are watching the same number, for reasons you can usually name.
Support and resistance describe prices where movement has repeatedly slowed or reversed. The useful question is not whether they exist — charts obviously show them — but why, since the answer tells you which levels are worth marking.
Where they come from
Prior turning points. A price where a move previously reversed is a price where a lot of transactions happened and a lot of people formed views. Someone who sold there and watched price run away wants a second chance. Someone who bought there and got hurt wants out at breakeven. Both place orders around the same number.
Reference levels everyone can see. The prior day's high and low, the session open, the previous settlement, the week's range. These are unambiguous — no settings, no interpretation, so every participant computes the same number.
Round numbers. Whole figures attract orders for no reason more sophisticated than that people think in round numbers. It is not analysis, and it is real.
Volume concentrations. Prices where unusually large volume traded represent levels where many positions were established.
Every one of these is a statement about where orders are likely to sit. That is the mechanism. A level does nothing on its own — it works when enough participants are watching the same number that their combined behaviour changes what happens there.
Zones, not lines
Drawing a level as a one-tick line implies a precision the underlying reason does not support. Orders cluster around a number, not at it. A level is a band, and how wide that band should be depends on the instrument's volatility — which is one practical use for ATR.
Why they stop working
A level holds while the orders sitting there are large enough to absorb what arrives. It breaks when they are not. Nothing mystical happens on a break — the resting supply was simply consumed.
This explains a few things that otherwise look arbitrary. Levels weaken with each test, because each test consumes some of the orders that made them. Levels break easily in fast conditions, because size arriving quickly overwhelms what is resting. And levels far in the past matter less, because the people who placed those orders have moved on.
The selection problem
Look at any chart for long enough and you can draw a level near almost every turn. The chart will then appear to confirm your levels beautifully, and it will have confirmed nothing, because you chose them after seeing the turns.
The discipline that fixes it: mark levels using a rule you could have applied yesterday — prior session extremes, a defined swing size, volume concentrations over a fixed window — and leave them where the rule puts them, including the ones that get ignored.