How RSI Actually Works: The Math Behind the Indicator - Part 6

Understanding the RSI Calculation: A Deep Dive into Smoothing and Averaging

Welcome back, traders. In this installment of our RSI Foundations module, we’re peeling back another layer of the onion, delving deeper into the mathematical underpinnings of the Relative Strength Index. While the previous sections laid the groundwork for understanding the raw components of RSI – the average gains and average losses – today we're going to dissect how those averages are actually calculated and, crucially, the impact of smoothing on the indicator's behavior. This isn't just an academic exercise; a nuanced understanding of these calculations will empower you to interpret RSI signals with greater precision and anticipate its movements more effectively in real-time trading scenarios.

As an institutional trader for over two decades, I’ve seen countless retail traders misuse indicators because they didn't fully grasp their construction. They treat them as black boxes, blindly following signals without understanding the mechanics. That's a recipe for disaster. Our goal here is to demystify RSI, turning you into an informed user, not just a consumer of signals.

The Core Components Revisited: Average Gain and Average Loss

Let’s quickly recap. The fundamental building blocks of RSI are Average Gain (AG) and Average Loss (AL). These represent the average magnitude of price increases and decreases over a specified lookback period, typically 14 periods.

For the first calculation in a new series, the Average Gain and Average Loss are simple arithmetic averages:

Initial Average Gain (AG_initial) = Sum of all positive price changes over 'N' periods / N Initial Average Loss (AL_initial) = Sum of all absolute negative price changes over 'N' periods / N

Where 'N' is the RSI lookback period (e.g., 14).

However, subsequent calculations are where the smoothing comes into play, and this is where many traders get lost.

The Wilders' Smoothing Method: A Key to RSI's Responsiveness

J. Welles Wilder Jr., the creator of RSI, didn't use a simple moving average for subsequent calculations. He employed a specific type of exponential smoothing, often referred to as "Wilders' Smoothing" or "Modified Moving Average (MMA)". This method gives more weight to recent data, making the indicator more responsive to current price action while still incorporating past data.

Here's how it works for each subsequent period after the initial calculation:

Subsequent Average Gain (AG_current) = [(Previous Average Gain * (N - 1)) + Current Gain] / N Subsequent Average Loss (AL_current) = [(Previous Average Loss * (N - 1)) + Current Loss] / N

Let's break down this formula and understand its implications.

• Previous Average Gain/Loss * (N - 1): This part of the equation gives significant weight to the existing average. It essentially "carries over" the influence of past data.
• Current Gain/Loss: This is the new piece of information from the most recent period.
• Divided by N: This normalizes the sum back into an average.*

Why is this important?

1. Responsiveness vs. Smoothness: This smoothing technique strikes a balance. A simple moving average would be too slow, and a pure exponential moving average might be too fast. Wilders' smoothing provides a good compromise, allowing RSI to react to new price information without being overly erratic.
2. Continuous Data Stream: Unlike a simple moving average that "drops off" old data completely, Wilders' smoothing continuously incorporates all past data, albeit with diminishing weight. This means that a significant price move from several periods ago still has a residual, albeit small, impact on the current RSI value.
3. Impact on Overbought/Oversold Readings: Because of this smoothing, RSI tends to hold overbought or oversold levels longer than it would with a simple moving average, but not as long as a very slow moving average. This is crucial for understanding the persistence of trends and potential for reversals.

A Step-by-Step Calculation Example: ES Futures on a 5-Minute Chart

Let’s get practical. Imagine we're looking at the E-mini S&P 500 futures (ES) on a 5-minute chart. We'll use a standard 14-period RSI.

Scenario: We have 14 periods of ES 5-min close data. Let's assume we've already calculated our initial AG and AL.

Period 14 (Initial Calculation Completed): Let's say our initial 14-period calculations yielded:

• Initial AG = 5.25 points
• Initial AL = 3.75 points

Now, let's move to Period 15.

Period 15:

• ES closes at 4520.00.
• Previous close (Period 14) was 4518.50.
• Price Change = +1.50 points.
• Current Gain = 1.50 (since it's a gain)
• Current Loss = 0 (since it's not a loss)

Now, we apply Wilders' Smoothing:

New AG (Period 15) = [(Previous AG * (14 - 1)) + Current Gain] / 14 New AG = [(5.25 * 13) + 1.50] / 14 New AG = [68.25 + 1.50] / 14 New AG = 69.75 / 14 New AG = 4.982 (rounded)

New AL (Period 15) = [(Previous AL * (14 - 1)) + Current Loss] / 14 New AL = [(3.75 * 13) + 0] / 14 New AL = [48.75 + 0] / 14 New AL = 48.75 / 14 New AL = 3.482 (rounded)

Now that we have our New AG and New AL, we can calculate the Relative Strength (RS) and finally the RSI for Period 15:

RS = New AG / New AL RS = 4.982 / 3.482 RS = 1.431 (rounded)

RSI = 100 - (100 / (1 + RS)) RSI = 100 - (100 / (1 + 1.431)) RSI = 100 - (100 / 2.431) RSI = 100 - 41.135 RSI = 58.865 (rounded)

So, for Period 15, our RSI is approximately 58.87.

Let's do one more period to solidify the concept.

Period 16:

• ES closes at 4517.00.
• Previous close (Period 15) was 4520.00.
• Price Change = -3.00 points.
• Current Gain = 0
• Current Loss = 3.00 (absolute value)

Using our New AG (4.982) and New AL (3.482) from Period 15:

New AG (Period 16) = [(Previous AG * (14 - 1)) + Current Gain] / 14 New AG = [(4.982 * 13) + 0] / 14 New AG = [64.766 + 0] / 14 New AG = 64.766 / 14 New AG = 4.626 (rounded)

New AL (Period 16) = [(Previous AL * (14 - 1)) + Current Loss] / 14 New AL = [(3.482 * 13) + 3.00] / 14 New AL = [45.266 + 3.00] / 14 New AL = 48.266 / 14 New AL = 3.448 (rounded)

Now for RS and RSI for Period 16:

RS = New AG / New AL RS = 4.626 / 3.448 RS = 1.342 (rounded)

RSI = 100 - (100 / (1 + RS)) RSI = 100 - (100 / (1 + 1.342)) RSI = 100 - (100 / 2.342) RSI = 100 - 42.698 RSI = 57.302 (rounded)

Notice how the RSI has moved from 58.87 down to 57.30, reflecting the recent negative price change, but not drastically, thanks to the smoothing.

Practical Implications for Day Traders

Understanding this calculation isn't just for trivia night. It has direct practical implications for your day trading:

1. Lag is Inherent, But Manageable: Wilders' smoothing, while responsive, introduces a degree of lag. RSI will always be a lagging indicator. This is why combining it with price action and other leading indicators (like order flow or market structure) is crucial. Don't expect RSI to call the absolute top or bottom on every single swing.
2. Persistence of Trend: When RSI is trending strongly (e.g., staying above 70 in an uptrend or below 30 in a downtrend for an extended period), it’s because the average gains are consistently outweighing average losses (or vice versa) and the smoothing mechanism is reinforcing that trend. This tells you that the momentum is persistent, not just a fleeting spike.
   • Actionable Tip: In a strong uptrend on NQ futures on a 1-minute chart, if RSI hits 70 and stays there for several bars, it suggests sustained buying pressure, not necessarily an immediate reversal. Instead of shorting an "overbought" RSI, look for signs of momentum weakening in conjunction with price action (e.g., lower highs, increased selling volume).
3. The "N" Period Impact: The choice of 'N' (the lookback period) directly influences the sensitivity of the smoothing.
   • Shorter 'N' (e.g., 6 or 9 periods): The indicator will be more sensitive and volatile. It will reach overbought/oversold levels faster and reverse more quickly. This might be suitable for very fast-paced scalping on 1-minute charts, but it also generates more false signals.
   • Longer 'N' (e.g., 20 or 21 periods): The indicator will be smoother and less volatile. It will take longer to reach overbought/oversold levels and will stay there for longer. This is more suited for swing trading or identifying longer-term trends on higher timeframes (e.g., 15-min or 30-min charts).
   • Actionable Tip: For day trading ES or NQ on 5-minute charts, the standard 14-period is a good starting point. If you find it too laggy for your style, try 12. If it's too choppy, try 16. Experiment on your demo account to find what suits your trading style and the asset's volatility.
4. Divergence Validation: Understanding the smoothing helps validate divergence. A strong price move without a corresponding RSI move (or an RSI move in the opposite direction) becomes more meaningful when you know how RSI is constructed. It indicates that the average momentum is not confirming the latest price extreme, suggesting a potential shift in the underlying strength of the move.

The Role of Relative Strength (RS) in the Formula

Before RSI is calculated, we determine Relative Strength (RS).

RS = Average Gain / Average Loss

This ratio is critical.

• If Average Gain > Average Loss, then RS > 1. This means upward price movements have been stronger on average than downward movements. The higher the RS, the stronger the upward momentum.
• If Average Gain < Average Loss, then RS < 1. This means downward price movements have been stronger on average than upward movements. The lower the RS, the stronger the downward momentum.
• If Average Gain = Average Loss, then RS = 1. This means price is in equilibrium, or sideways consolidation.

Why not just use RS? Why convert it to RSI?

The conversion to RSI (100 - (100 / (1 + RS))) normalizes the indicator into a fixed range of 0 to 100. This normalization is incredibly powerful:

1. Universal Comparison: It allows for easy comparison across different assets, timeframes, and even different types of indicators. An RSI of 70 on SPY 15-min chart means the same thing qualitatively as an RSI of 70 on a GOOGL 5-min chart – it's in the overbought territory.
2. Clear Overbought/Oversold Levels: The 0-100 scale provides clear, universally accepted thresholds (70/30 or 80/20) for identifying potential overbought and oversold conditions, which would be much harder to define on an unbounded RS ratio.
3. Visual Interpretation: The bounded nature makes it visually intuitive to spot extremes and trend changes on a chart.

Advanced Insight: How "Zero" in the Denominator is Avoided

A common concern with ratios is division by zero. What happens if Average Loss is zero for a period? If there are no down periods within the 'N' lookback period, Average Loss would mathematically be zero.

Wilder's original calculation handles this implicitly. If there are no losses over the 'N' period, the Average Loss will remain a small non-zero number from prior calculations, even if it's trending towards zero. It won't instantaneously hit zero unless the market has been only moving up for an extremely extended period, which is rare in liquid markets. Even then, most charting platforms have internal safeguards to prevent division by zero errors, often by assigning a very small positive number if AL truly becomes 0.

However, if Average Gain is zero (i.e., only losses or flat periods), then RS becomes 0 / AL = 0. RSI = 100 - (100 / (1 + 0)) = 100 - 100 = 0. This means RSI hits 0, indicating extreme oversold conditions. Conversely, if Average Loss is zero (only gains or flat periods), then RS becomes AG / 0 (effectively infinity). RSI = 100 - (100 / (1 + infinity)) = 100 - 0 = 100. This means RSI hits 100, indicating extreme overbought conditions.

These extreme values (0 and 100) are rare but signify absolute dominance of one direction, which is a powerful signal in itself.

Trade Example: Using RSI's Smoothing for Confirmation on NQ

Let's consider a practical trade setup on the Nasdaq 100 futures (NQ) on a 1-minute chart. We're looking for a short-term reversal after a strong move.

Scenario: NQ has been rallying hard for the past 20 minutes. Price is pushing up, but you notice that the rate of ascent is slowing down. You're using a 14-period RSI.

Observation:

1. NQ makes a new swing high at 18,250.
2. RSI (14-period) reaches 78, indicating deeply overbought conditions.
3. NQ pulls back slightly to 18,235. RSI drops to 72.
4. NQ then attempts another push higher, making a marginal new high at 18,255.
5. Crucially, RSI on this new high only reaches 75.

Interpretation (applying our understanding of smoothing): The fact that RSI made a lower high (75 vs. 78) despite price making a higher high (18,255 vs. 18,250) is a classic bearish divergence. Because of the smoothing, for RSI to make a lower high, it implies that the average gains over the last 14 periods are no longer as strong as they were during the previous peak. The recent upward price action, while pushing price slightly higher, is not accompanied by the same underlying momentum as before. The new gain was not significant enough to outweigh the influence of the previous period's average gains, leading to a weaker RSI reading. This is a powerful signal of weakening momentum.

Trade Strategy (Short Entry):

• Entry: After the divergence is confirmed and you see a negative candle close below the prior bar's low (e.g., NQ closes at 18,245 after the 18,255 high). This confirms sellers are stepping in.
• Stop Loss: Place your stop loss just above the recent swing high (e.g., 18,260, allowing for a few points of buffer above 18,255).
• Target: Look for NQ to retrace to a previous support level or a key moving average (e.g., the 20-period EMA on the 1-minute chart, or a prior low around 18,200).
  • Let's say your target is 18,200.
  • Risk: 18,260 - 18,245 =