RSI, Reconsidered: The Directional Imbalance Ratio

2.1 Price

The term price is used in two different contexts.

In the context of the indicator, price refers to a chosen input supplied to the arithmetic process — for example, a series of closing prices. This series consists of values directly derived from transaction prices generated through market execution. The choice of input affects which data points are summarized, but it does not alter the underlying logic or the form of the indicator’s output.

Regardless of the input selected, RSI always quantifies directional imbalance between upward and downward changes in the selected series over the chosen lookback window. For this reason, the choice of input is treated as a setup detail and is not evaluated or compared further.

In the context of execution, price refers to the transaction price formed through trading at the bid and the ask. Execution mechanics describe how these transaction prices change; they are independent of how prices are later selected, labeled, or organized for the computational engine.

2.2 Parameters

Parameters define the inputs an indicator accepts as part of its definition. In RSI, these are the lookback period (n) and the price input. Users do not choose the parameters themselves; they select values for those parameters when applying the indicator.

The lookback period sets how much price history the mathematical process draws from. Changing its value alters the scope of data summarized, but it does not alter what the arithmetic represents. RSI values computed with different lookback lengths therefore describe the same underlying logic, summarized over different portions of past data.

More generally, parameter values change the input window, not the formula. The process is the same; only the lookback length changes, which changes the output values because they summarize a different historical span. Any feature that depends on a parameter value is therefore configurational, not part of RSI’s core definition.

2.3 Representation

Representation refers to how a structural output is presented for reading and use—for example, the numerical scale, sign, or reference point used to express a value.

A difference in representation does not imply a difference in the arithmetic or any new information in the resulting value. It affects how easily the reader can see what the algorithmic process summarizes.

In this article, differences in representation are treated as differences in expression, not differences in what the computational path produces. The goal of a refined representation is to align the data summary more closely with its underlying logic.

2.4 Interpretation

Interpretation refers to the meaning assigned to a quantitative result beyond what the computational logic itself specifies.

Thresholds, labels, narratives, and expectations about continuation or reversal are interpretive overlays. They are applied after the arithmetic is produced and are not treated as intrinsic properties of the indicator.

2.5 Scope

Scope defines what is examined in this article.

This article examines how RSI works mechanically and how its output is expressed to the reader. It does not evaluate trading performance, parameter optimization, or predictive use.

Where interpretation is discussed, it is treated as an external analytical choice, not something contained in the algorithmic process or added by the display conventions. The focus remains on the structural logic of the numerical summary rather than its subjective application.

4.1 Unsigned Representation

By construction, the RSI’s structural output encodes directional dominance in the imbalance between upward and downward price changes. The 0-100 scale, however, removes that direction from the representation.

Direction must therefore be inferred from the midpoint value of 50. Additional levels, such as 70, 60, or 30, further increase the interpretive burden by shifting direction from explicit encoding to user-imposed convention.

Within this numerical framework, the lack of a signed value forces the reader to perform mental arithmetic to determine orientation—a step that is not required by the computational path itself, but is necessitated by its unsigned expression.

4.2 Asymmetric Encoding

Directional imbalance is symmetric by definition: an excess of upward price changes and an excess of downward price changes are equivalent conditions in opposite directions. The 0-100 scale does not preserve this symmetry.

Values above and below the midpoint are therefore encoded asymmetrically, even when they reflect imbalances of equal size. For example, readings of 60 and 40 represent identical departures from balance in opposite directions, yet 60 is often perceived as “stronger” and 40 as “weaker” — a distinction influenced by the representation, not by the mathematical engine itself.

In this structural logic, the quantitative summary remains neutral, but the scale imposes a hierarchy. The process treats both directions as equals, but the 0-100 display creates a perceptual bias where one direction is viewed as an accumulation and the other as a depletion.

4.3 The Structural Meaning of the 0–100 Range

On the 0–100 scale, only three values possess a built-in meaning defined by the computational architecture. These points are not conventions; they are the logical boundaries of the math itself.

The most significant is the midpoint 50. An RSI reading of 50 simply means that upward and downward price changes perfectly cancel each other out over the chosen time window. This arithmetic equilibrium comes directly from the procedural rules of the indicator, not from market sentiment. It is a neutral statement of balance that does not imply hesitation or any psychological state.

Similarly, the extremes 0 and 100 represent the saturation points of the underlying logic.

Unlike the arbitrary 70 or 30 thresholds, these three points — 0, 50, and 100 — are intrinsic. The 0–100 representation obscures this by treating 50 as a "pass-through" point rather than the structural zero. By recognizing that the arithmetic engine reaches resolution only at the points of perfect balance or total dominance, we can see that all other values are merely degrees of directional imbalance along a continuous gradient. In this numerical summary, any value other than 50 represents a state of asymmetry, with 0 and 100 serving as the terminal poles where the computational path reaches its logical limit.

4.4 Thresholds as Conventions

Thresholds such as 30, 70, 80, or 20 do not share the structural role of the midpoint or the poles. While the midpoint corresponds to a unique condition of equilibrium, all other values represent arbitrary coordinates along a continuous data profile of imbalance.

Mathematically, imbalance admits no intrinsic breakpoints, Although the computational framework may produce these values, their treatment as “boundaries” or “states” is introduced by convention rather than defined by the algorithmic output.

In this numerical summary, 70 is no more a "limit" than 68 or 72. Elevating specific numbers to the status of thresholds is an interpretive act that happens after the calculation is complete. It effectively layers a binary meaning — such as "overbought" or "oversold" — onto a non-binary underlying logic that reflects only the magnitude of asymmetry.

4.5 A Symmetric Alternative

A signed, symmetric scale makes direction and symmetry explicit. On a normalized interval such as −1 to +1, zero corresponds to balance, while the sign and distance from zero encode the orientation and magnitude of the imbalance.

Changing the representation to this format does not alter the numerical result or introduce new assumptions. Instead, it aligns the data expression with the procedural rules of the indicator, leaving the underlying logic itself unchanged.

In this arithmetic framework, the algorithmic output becomes intuitive: a positive value reflects an upward skew, a negative value reflects a downward skew, and zero represents perfect equilibrium. The quantitative summary is simply unfolded to reveal the computational architecture that the 0–100 scale obscures.

Furthermore, the terminal poles of -1 and +1 now serve as functional anchors. They define the absolute limit of the arithmetic engine, representing the point where directional dominance is total. By centering the structural zero, the data profile allows the user to perceive the parity of opposing market states without the mental subtraction required by the traditional scale.

The name Relative Strength Index reflects descriptive convention rather than the nature of what is quantified. Its components — “relative,” “strength,” and “index” — create expectations that the indicator's structural output itself does not define.

This gap becomes clear when the expectations implied by the name are compared directly to what the mathematical framework actually specifies. The name suggests a qualitative state, while the underlying logic remains a quantitative accounting of directional imbalance.

6.1 What “Relative” Specifies — and What It Does Not

In RSI, the term “relative” refers to the ratio between upward and downward price changes within a single time series over a specified lookback period. In that limited sense, it is technically accurate, but it adds little descriptive value: any ratio is, by definition, relative to its components.

The term neither specifies the dimension being compared nor distinguishes this accounting from other ratio-based indicators. By contrast, describing the quantitative summary in terms of what is being compared— imbalance between upward and downward price changes—makes its basis explicit, without relying on a vague or generic modifier.

Within the computational architecture of the indicator, the "relativity" is simply the arithmetic relationship between two directional sums. Labeling the result as a directional imbalance removes the interpretive ambiguity that the word “relative” often introduces.

6.2 Why “Strength” Suggests More Than Is Recorded

The term “strength” carries a deeper ambiguity. In analytical usage, strength typically implies capacity, persistence, or an underlying property that endures beyond realized price changes—none of which are part of the numerical summary.

Within the arithmetic engine, "strength" has no fixed unit, baseline, or maximum. The midpoint indicates balance between opposing directional changes; it carries no implication of power or weakness, and the extremes do not correspond to any intrinsic market condition. The algorithmic pathway reports only the magnitude of imbalance.

It is common to describe markets as “stronger” when upward price changes dominate, but that description is an interpretive overlay. The structural output itself does not distinguish between dominance as strength, persistence, or durability; it simply reflects the tally observed over the chosen window.

Directional dominance is therefore not equivalent to strength. Values can register at extremes in fragile or corrective conditions because the computational architecture has limited memory that resets as new data enters. Strength, in this context, is an attribution layered onto the result, not a quantity defined by the underlying logic.

6.3 Why RSI Is Not an Index

RSI is not an index in the conventional sense. It is not a composite across constituents or a representative data profile of a system.

Instead, it is a bounded ratio derived from a single price series, scaled for convenience rather than to represent a system-level quantitative summary. While an index typically tracks the aggregate behavior of a group, the procedural rules of RSI focus exclusively on the internal arithmetic relationship of a single instrument's history.

Within this numerical framework, the structural output serves as a standalone tally. Calling it an index implies a scope of observation that the underlying logic does not support.

This article introduces the Directional Imbalance Ratio (DIR) as an indicator that expresses the same underlying arithmetic engine as RSI, but through a signed, symmetric scale.

By shifting the representation, the DIR provides a quantitative profile that makes the orientation of price movement explicit.

The relationship between the two indicators is defined by the following transformation:

This structural conversion ensures that the numerical summary remains mathematically identical to the original logic while removing the interpretive burden of the 0-100 scale.

A reference implementation of DIR for MetaTrader 4 is provided in Appendix B

7.1 Signed, Normalized Imbalance

Its natural range is −1 to +1: the endpoints denote extreme imbalance, and the midpoint denotes equilibrium. This quantitative signature is inherent rather than cosmetic; it follows directly from the procedural logic of the computational path.

By adopting this numerical framework, the arithmetic results are rendered intuitively. The user no longer needs to mentally subtract 50 to find the zero-point; the structure of the scale now matches the underlying mechanics of the execution process.

7.2 Mapping RSI → DIR

DIR introduces no new arithmetic; it is a linear rescaling of the computational path’s output. This structural translation preserves all information present in RSI while making direction, symmetry, and the meaningful midpoint explicit.

Conventional RSI thresholds map directly into DIR under the same procedural rules. For example:

These levels remain interpretive conventions rather than structural boundaries. In the DIR data profile, their parity and distance from balance are immediately visible. This quantitative summary allows the user to see the degree of departure from zero without requiring the categorical "overbought/oversold" narrative.

7.3 Comparison of Encodings: RSI vs. DIR

Defined independently, DIR is the normalized difference between cumulative upward and downward price changes over a sampling window. It serves as a data profile of the arithmetic process without the interpretive friction of the traditional index format.

The full mathematical derivation of the relationship between RSI and DIR is provided in Appendix A

The figure above shows RSI(14) and DIR(14) derived from the same price series over the same lookback period. As the plots make clear, the shapes of the RSI and DIR curves are identical: peaks, troughs, and transitions occur at the same points in time, and the two series evolve in exact correspondence. Selected points are annotated to show the numerical results produced by each indicator at those moments.

This is expected. DIR is a linear rescaling of RSI, not a different underlying logic. No information is added, removed, or altered; only the representation changes.

What differs is how that information is encoded. In RSI, direction must be inferred relative to a midpoint on an unsigned scale. In DIR, direction is encoded directly by the sign, and equilibrium is explicitly anchored at zero. Equal magnitudes of imbalance in opposite directions therefore appear as symmetric values rather than as qualitatively distinct “high” and “low” readings.

The chart is not presented as an example of signal behavior or interpretation. It serves only to illustrate how the same arithmetic appears when direction, symmetry, and balance are made explicit in the data summary.

A reference implementation of DIR for MetaTrader 4 is provided in Appendix B

9.1 Signed Outputs as a Baseline Expectation

Analytical practice generally expects quantities that encode opposition to do so through sign rather than inference. Across statistics, signal processing, and quantitative analysis, sign conveys direction and magnitude conveys degree, without reliance on interpretive cutoffs or midpoint-based inference.

DIR is consistent with this expectation. Positive and negative values correspond to opposing imbalances, while zero corresponds to balance. Orientation is available from the value itself. This does not change the underlying logic; it changes how directly the calculated result can be read.

9.2 Symmetry and Portability Across Analysis and Automation

Symmetry allows quantities to move easily across analytical contexts. Symmetric representations integrate naturally into aggregation, transformation, modeling, and comparison because opposing conditions are encoded as numerical inverses rather than as offsets from an arbitrary center.

This matters in automated and multi-instrument settings. When quantities are centered at zero and symmetric by construction, they can be combined or transformed without conditional logic or special handling. DIR preserves the built-in symmetry of directional imbalance, making it structurally portable across charting, analysis, and automation.

In contrast, RSI’s asymmetric encoding requires explicit handling of balance, direction, and equivalence. DIR’s advantage is structural, not informational.

9.3 Clarity Aligned With Analytical Requirements

Analytical workflows benefit when structure is visible and interpretation is deliberate. When direction, symmetry, and balance are explicit in the representation, the boundary between the underlying logic and interpretation is easier to maintain.

DIR reflects this preference by making directional imbalance evident in the value itself, while keeping interpretive classifications clearly identifiable as user-imposed. This supports analysis, automation, and modeling by ensuring the arithmetic remains distinct from the narrative assumptions layered upon it.

10.1 Categorization Remains Possible

DIR makes direction, symmetry, and balance visible, but it does not encode categorization. Thresholds, labels, or narratives — such as “overbought,” “oversold,” “extreme,” “overextended,” or “reversal-prone” — remain deliberate, user-imposed modeling decisions, not attributes of the calculation or its representation.

10.2 Lookback Selection Remains External

Lookback selection remains external. DIR, like RSI, summarizes directional imbalance over a chosen window, but the choice of lookback period reflects analytical intent rather than a property of the underlying logic itself.

No representation can infer time relevance on the user’s behalf. The arithmetic provides a summary of the selected window, but the determination of whether that window is relevant remains a user-imposed parameter.

10.3 Context Remains Non-Negotiable

DIR does not supply context; it summarizes directional imbalance in price changes but does not encode regime, volatility, market structure, or information outside the price series itself. Expressing the underlying logic more clearly does not eliminate the need for judgment; it clarifies what the numerical result contributes and what it does not.

The arithmetic remains a closed system. While the representation is improved, the requirement for contextual analysis remains external to the indicator itself.

10.4 Expectation Management

Expectations about continuation, reversal, acceleration, or decay are inferences layered onto the arithmetic, not properties of the underlying logic itself.

Reframing RSI as DIR narrows the scope of what the indicator output is asked to do: it makes explicit the boundary between the calculation and inference without crossing that boundary on the user’s behalf.

11.1 RSI Reconsidered, Not Rejected

This article does not argue against RSI as an arithmetic process. Its computation is coherent, internally consistent, and structurally sound. What has been reconsidered is not the calculation itself, but how it is presented and named.

The issues addressed here arise from representation, not from the indicator’s underlying logic. DIR re-expresses that same numerical process in a form that makes its structure visible without interpretive mediation.

11.2 DIR as a Linguistic and Representational Correction

DIR is a linguistic and representational correction: a signed, symmetric expression of an existing computation whose structure is not made explicit in RSI’s conventional naming and scale.

DIR uses a name and scale that reflect the arithmetic, removing interpretive prompts that encourage premature inference and restoring symmetry, direction, and balance to the representation. By aligning the data summary with its underlying logic, the indicator is transformed from a source of mythology into a transparent analytical tool.

11.3 What Is Computed Is a Starting Point, Not an Answer

Neither RSI nor DIR answers questions about intent, regime, durability, or what comes next. They describe what has occurred within a defined window—nothing more. Any interpretation beyond that description is an analytical choice layered on top of the underlying logic.

Terms such as “overbought” and “oversold” reflect common interpretive conventions, not properties defined by the computation itself.

Clearer representation does not eliminate interpretation; it relocates it. When structure is explicit, interpretation becomes deliberate rather than implicit. The arithmetic remains the starting point of analysis, not its conclusion. The distinction between the quantitative result and inference is preserved.

A.1 RSI Construction (Summary)

RSI is derived from two accumulated quantities over a lookback period n:

• U: cumulative upward price changes
• D: cumulative downward price changes (expressed as positive values)

The relative strength is defined as:

The procedural framework then rescales this ratio to a bounded interval:

This transformation produces values in the range 0 to 100, with 50 corresponding to the condition U = D.

A.2 DIR Definition

DIR is defined as a signed, normalized expression of the same imbalance:

A.3 Linear Mapping Between RSI and DIR

RSI and DIR are related by a simple linear transformation. To derive the DIR from an existing RSI value, the numerical process is:

Conversely, to project a DIR back into the traditional 0–100 arithmetic summary, the algorithmic blueprint requires explicit grouping to maintain the computational path:

This mapping preserves all information present in RSI while making direction, symmetry, and balance explicit. No interpretive thresholds are introduced or removed by this procedural logic. Any thresholds applied to RSI map directly to corresponding DIR values; they remain interpretive conventions rather than structural features of the algorithmic pathway.