RSI Reversal: what it means
RSI Reversal is a practical way to describe a potential “turn” in market momentum using the Relative Strength Index (RSI). The key idea is not the RSI itself, but the change in RSI behavior that you treat as a reversal.
In education terms, RSI has a continuous history value (one RSI number per bar). A “reversal” rule then converts that history into a yes/no decision or a labeled event, based on conditions such as crossing a level or forming a turning point.
Because there is no single universal standard definition of “RSI Reversal” in all charting tools and blogs, you should always state the exact reversal rule you are using. The RSI calculation is standard; the reversal interpretation is the part that can vary.
The RSI calculation (the standard input)
Before applying any reversal logic, you calculate RSI from price.
Step 1: Choose the price series
Pick what “price” means for your RSI. Common choices are:
- Close-to-close price changes (most common): use the bar-to-bar close changes.
- Other sources (less common): some tools allow different price inputs.
To keep results consistent, use the same price type and the same bar frequency (e.g., 1H candles or daily candles) for all calculations.
Step 2: Use a lookback length
Select an RSI lookback length L (often 14 in common examples, but it is not required to be 14). This length controls how much history RSI uses.
Step 3: Compute gains and losses
For each bar t, compute the change from the previous bar:
- Change(t) = Price(t) − Price(t−1)
- Gain(t) = max(Change(t), 0)
- Loss(t) = max(−Change(t), 0)
Step 4: Smooth gains and losses
RSI relies on smoothed averages of gains and losses over the lookback. Many standard implementations use an exponential moving average style smoothing (often described like Wilder’s smoothing).
In a generic form (using smoothing averages), you compute:
- AvgGain(t): smoothed average of Gain over the last L period
- AvgLoss(t): smoothed average of Loss over the last L period
(Exact smoothing details can differ by platform, so the most important verification step is to match your platform’s RSI method.)
Step 5: Convert to RSI
Compute the relative strength and then RSI:
- RS(t) = AvgGain(t) / AvgLoss(t)
- RSI(t) = 100 − [100 / (1 + RS(t))]
Edge cases matter:
- If AvgLoss(t) = 0, then RS becomes very large, and RSI approaches 100.
- If AvgGain(t) = 0, RS becomes 0, and RSI becomes 0.
“RSI Reversal” calculation: define and apply the reversal rule
Now you have RSI values: RSI(t), RSI(t−1), RSI(t−2), … The reversal part uses a rule on these values.
Below are several example reversal rules. Pick one, then calculate it exactly. If your source uses a different rule, your results will differ even with identical RSI values.
Option A: Reversal as crossing a threshold
A common approach defines reversal relative to a threshold level T (for example, an RSI “overbought” threshold or “oversold” threshold).
Example reversal logic (generic):
- A downward reversal event occurs when RSI crosses from above T to at or below T.
- An upward reversal event occurs when RSI crosses from below T to at or above T.
Parameters you must state:
- threshold level T
- whether you look for upward crossing, downward crossing, or both
- whether equality (exactly T) counts as crossing
Option B: Reversal as a turning point (local extremum)
Another definition uses RSI shape.
Example reversal logic (generic):
- Identify a local maximum when RSI(t) is greater than RSI(t−1) and RSI(t+1).
- Identify a local minimum when RSI(t) is less than RSI(t−1) and RSI(t+1).
Parameters you must state:
- the window size (using immediate neighbors is one simple version)
- whether you require the extremum to also be above/below certain levels
Option C: Reversal as divergence between price and RSI (requires extra data)
Some RSI reversal concepts involve divergence: the price makes a higher high (or lower low) while RSI makes the opposite.
A generic divergence check typically needs:
- identification of two swing highs/lows in price
- identification of corresponding swing highs/lows in RSI
Because swing detection depends on a window and a method (how you decide “a swing high”), the divergence-based approach is more sensitive to parameter choices and to how your platform detects turning points.
Evidence or example (manual verification with assumptions)
Here is a simple, verifiable example using Option A (crossing a threshold). This is not a forecast—just an illustration of how to compute the event.
Assumptions for the example:
- Use RSI length L (any fixed integer; the RSI calculation must match your tool)
- Use threshold T = 50 (you can choose a different T, but then use that T consistently)
- Use Close-to-close price changes
Example event:
- You compute RSI at two consecutive bars: RSI(t−1) = 48 and RSI(t) = 52.
- Under upward-crossing logic with threshold T = 50, you flag a reversal upward crossing at bar t because RSI moved from below 50 to above 50.
If instead RSI(t−1) = 52 and RSI(t) = 48, you flag a reversal downward crossing.
What you can independently verify:
- Recompute RSI(t−1) and RSI(t) from the same price series and the same RSI method.
- Check whether the crossing condition you chose is satisfied.
This separation—compute RSI with a standard method, then apply a clearly stated reversal rule—is the core of “RSI Reversal calculation.”
Material limitations and failure modes
RSI-based reversal ideas can fail for multiple reasons. These are not edge cases; they are common.
1) Strong trends can keep RSI “reversed” repeatedly
In trending conditions, momentum often stays elevated or depressed for long stretches. If you use a simple crossing rule, you may see frequent events that do not correspond to a sustained change in direction.
2) Platform differences in RSI smoothing change the numbers
Even though the RSI formula shape is standard, implementations can differ in the smoothing method and the handling of initial periods (startup values). If you compare results from two charting tools, you may see different RSI values and therefore different reversal events.