RSI compresses recent gains and losses into a bounded reading. Its neat scale encourages a rigid habit: above 70 means sell, below 30 means buy. Strong trends expose the weakness of that shortcut. Momentum can remain elevated while price advances far beyond the first “overbought” print.
Find the neighbourhood
Hide the threshold labels for a moment and observe the oscillator’s working range. During a sustained advance, pullbacks may repeatedly hold near 40 or 50 while rallies reach 70 and above. In a sustained decline, rebounds may fail near 50 or 60 while sell-offs revisit 30.
This regime is descriptive, not permanent. It tells us how momentum has behaved under the current structure. A shift becomes interesting when price and RSI both stop behaving in their established neighbourhoods.
For example, suppose an advance has held RSI above 42 for several months. A later pullback breaks a meaningful price swing and carries RSI to 34. That combination deserves more attention than an isolated dip below a standard line, because both structure and momentum character changed.
Separate condition from trigger
An elevated RSI reports strong recent gains relative to losses. It does not identify the bar on which demand must disappear. Likewise, a depressed reading reports forceful recent losses; it does not establish that selling is complete.
Before acting on a threshold, a careful reader asks:
- Is price trending, ranging or breaking from balance?
- Has the oscillator remained in a recognisable range?
- Did the latest price swing break structure or merely test it?
- Does volume show expanding opposition or ordinary contraction?
- What price event would confirm that momentum loss now matters?
Failure swings and divergence
A lower RSI peak while price makes a higher high can warn that upward pace has faded. But comparable price swings must be selected, and trend structure may remain intact. A subsequent failure to regain the prior RSI zone can strengthen the warning; a price break can supply confirmation.
The order matters. Momentum disagreement is observed first. Price evidence follows or does not. If price resumes with participation, the warning may be cancelled.
Record the regime, the change and the required confirmation in separate sentences. This keeps a useful measurement from becoming a command it was never designed to give.
The Divergence Practice Lab applies this method to hidden and regular divergence.