Settings and Thresholds: Trading Off Sensitivity

Shorter period, more sensitive

The shorter the lookback period, the more the indicator reacts to each candle and the more often it hits overbought/oversold: more signals, more noise. Longer periods are smoother, with fewer and later signals.

The more extreme the threshold, the higher the bar

Loosen the signal threshold from 70 / 30 to 80 / 20, so only more extreme readings count. Signals usually get fewer, but that doesn't guarantee they're more accurate.

Don't worship the "best settings"

With only one or two signals, a "100% success rate" means nothing. Picking the setting that did best in history is very likely overfitting. Start with common defaults, then check whether they still hold up in different kinds of markets.

Key takeaways

• The shorter the period, the more sensitive the indicator: more signals, but more noise. The longer the period, the smoother: fewer, later signals.
• The more extreme the threshold (e.g. 80 / 20), the fewer signals you usually get, but that doesn't guarantee a higher success rate.
• With only a handful of signals, the success rate means almost nothing, and picking the "best settings" from history easily leads to overfitting.
• Common defaults (like RSI 14, 70 / 30) are a sensible starting point. What matters more is whether the settings hold up across different market conditions.

Try it yourself

You can do these exercises in this lesson's interactive simulation:

1. In "Choppy," click the "70 / 30" cells in the RSI(7), RSI(14) and RSI(21) rows in turn: how does the signal count change?
2. Click RSI(14) · 80 / 20: how many signals appear in this ranging stretch? Read the explanation on the right.
3. Switch to "Strong trend" and compare RSI(7) at "70 / 30" and "80 / 20": how do the signal count and success rate change? Are stricter thresholds always more accurate?
4. Switch back to "Choppy" and find a cell marked "small sample": why shouldn't you trust its success rate?

Quiz and explanations

1. If you change the RSI period from 14 to 7, what's most likely to happen?

Answer: More signals, more sensitive, and more false signals.The shorter the period, the more each candle affects the reading.

2. A set of settings produced only 2 signals over the past year, and both worked (100%). What's the right view?

Answer: The sample is too small for the success rate to mean anything.The result of 2 signals is very likely luck; a different stretch of market could look completely different.

3. Changing the thresholds from 70 / 30 to 80 / 20 usually…

Answer: Gives fewer signals, but they're not guaranteed to be more accurate.A higher bar means fewer triggers; whether they're more accurate depends on the market.

Terms in this lesson

• lookback period (Lookback Period): How many recent candles an indicator uses; shorter is more sensitive but noisier, longer is smoother but slower.
• overfitting (Overfitting): Tuning parameters so tightly to past data that the strategy does badly on new data.
• signal threshold (Signal Threshold): The line an indicator must cross to count as a signal, such as RSI's 70/30; the more extreme the threshold, the fewer the signals.