Settings and periods: why a default is just a default
14-period RSI. 20-period Bollinger Bands. 12/26/9 MACD. None of those numbers came from a study of your instrument.
The standard settings on most indicators were chosen decades ago by their authors, for the markets and timeframes they were working in, often with a rationale as loose as the number of trading days in a fortnight. They persist because they are printed in books, not because they were validated.
What the period controls
The lookback is a sensitivity dial. Shorter reacts sooner and more often. Longer reacts later and less often. Everything else about the indicator's behaviour follows from that one choice more than from the choice of indicator.
A practical consequence: a 9-period RSI and a 25-period RSI are, in effect, different indicators. Do not carry an interpretation from one to the other.
The timeframe interaction people miss
A 14-period indicator covers 14 bars regardless of what a bar represents. On a 5-minute chart that is a little over an hour. On a daily chart it is three weeks. Same setting, entirely different question.
If you use the same settings across timeframes, you are not applying a consistent method. You are asking a different question on each chart and reading the answers as though they were comparable.
Where tuning turns into fooling yourself
Adjusting a setting until historical results improve is not research. It is search, and given enough attempts something will always look good on the sample you searched.
Warning signs, in rough order of seriousness:
- Performance falls apart when the period moves by one. Real effects are not that fragile.
- The chosen value is oddly specific — a 37-period average is a fitted number.
- Different values are needed for each instrument, with no reason given.
- The setting was chosen by trying many and keeping the best.
This is curve fitting, and it is the most common way people convince themselves a method works.
A reasonable way to choose
Start from the question, not the number. Decide what span of market activity you want the indicator to describe — the current session, the last week, the last quarter — then set the period so it covers that span on your chart's bar size. Leave it alone.
Prefer settings that hold up across a broad range of neighbouring values. If 18, 20, and 22 all behave sensibly, 20 is a choice. If only 20 works, 20 is an artefact.