📈 economics

Sufficient conditions for a Heuristic Rating Estimation Method application

This paper establishes the sufficient conditions for correctly applying the Heuristic Rating Estimation method to both complete and incomplete pairwise comparisons using arithmetic and geometric algorithms, demonstrating that the arithmetic variant yields optimal inconsistency estimations.

Original authors: Jacek Szybowski, Konrad Kułakowski, Jiri Mazurek

Published 2026-05-12

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jacek Szybowski, Konrad Kułakowski, Jiri Mazurek

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to rank a group of candidates for a job, but you don't know everyone's resume yet. You do, however, know the exact scores of a few "reference" candidates who have already been hired. You also have a list of comparisons: "Candidate A is twice as good as Candidate B," or "Candidate C is better than Candidate D."

This is the world of Pairwise Comparisons. It's a way of making decisions by comparing things two at a time rather than trying to judge everyone at once (which is too hard for our brains).

This paper introduces a specific tool called Heuristic Rating Estimation (HRE). Think of HRE as a smart calculator that uses the known scores of the "reference" candidates to guess the scores of the unknown ones, based on how they compare to each other and to the known ones.

The authors of this paper asked a very practical question: "When can we trust this calculator to give us a single, correct answer, and when will it break?"

Here is the breakdown of their findings, using simple analogies:

1. The Two Ways to Calculate (Arithmetic vs. Geometric)

The paper looks at two different ways the calculator works:

The Arithmetic Method: This is like taking a simple average. If Candidate A is compared to three people, you add up those comparisons and divide by three.
The Geometric Method: This is like a "multiplicative average." It's a bit more complex mathematically, but it handles the numbers differently.

2. The "Perfect" Scenario (Complete Data)

First, the authors looked at the easy case where you have all the comparisons. You know how every unknown candidate compares to every other candidate.

The Finding: The Arithmetic calculator works perfectly fine unless the data is extremely messy (inconsistent).
The Analogy: Imagine a group of friends trying to agree on a movie. If they mostly agree, the calculator works. But if they are wildly arguing (e.g., "A is better than B," "B is better than C," but "C is way better than A"), the calculator might get stuck.
The Limit: The paper proves there is a specific "tipping point" for how much disagreement (inconsistency) is allowed before the calculator fails. They found that their formula for this limit is the best possible limit; you can't make the rule stricter without breaking the method for valid cases.
The Geometric Winner: For the Geometric method, the authors found it is always safe. No matter how messy the data is, this version of the calculator will always find a unique, correct answer. It never gets stuck.

3. The "Real World" Scenario (Incomplete Data)

In real life, you rarely have all the comparisons. Maybe Candidate A never met Candidate B, so that data point is missing (marked with a question mark). This is an incomplete matrix.

The Challenge: When data is missing, the "friends" in our analogy are missing from the room. The calculator has to work with gaps.
The Arithmetic Result: The authors found that the Arithmetic calculator still works, but the rules are stricter. It depends on:
1. How many candidates there are.
2. How many comparisons are missing.
3. How "messy" (inconsistent) the existing data is.
  They provided a specific formula to tell you if your specific set of missing data is safe to use. Again, they proved this formula is the optimal limit—you can't push the boundaries any further without risking a broken calculation.
The Geometric Result: Just like in the perfect scenario, the Geometric method is the "unbreakable" one. Even with missing data, it will always produce a unique solution.

4. The "Singular" Trap

The paper includes examples where the Arithmetic method fails.

The Metaphor: Imagine a scale that is perfectly balanced on a knife-edge. If you add even a tiny bit of weight (or in this case, a specific pattern of inconsistency), the scale tips over completely and gives no answer. The authors showed exactly what that "knife-edge" looks like mathematically. They proved that their safety rules are tight enough to keep you off that edge, but not so tight that they reject valid data.

Summary of the Paper's Claims

Goal: To define the exact conditions under which the HRE method can be used to find unknown rankings.
Main Discovery:
- The Geometric version of HRE is robust; it always works for both complete and incomplete data.
- The Arithmetic version works most of the time, but only if the data isn't too inconsistent and the missing data isn't too chaotic.
Key Contribution: The authors didn't just say "it works sometimes." They used advanced math (linear algebra and spectral theory) to draw a precise line in the sand. They proved that their "safety line" is the best possible one—you cannot make the rules any more lenient without the method failing.

In short, this paper is a user manual for the limits of a decision-making tool. It tells you exactly how much "noise" or "missing info" your data can handle before the tool stops giving you a reliable answer.

Technical Summary: Sufficient Conditions for a Heuristic Rating Estimation Method Application

Problem Statement
The Heuristic Rating Estimation (HRE) method is a decision-making tool designed to evaluate a set of alternatives based on pairwise comparisons (PC) and the known weights of specific "reference" alternatives. While HRE has been established for complete datasets, practical applications often involve incomplete pairwise comparison matrices where certain comparisons are unknown. This incompleteness raises fundamental questions regarding the solvability and uniqueness of the resulting estimation problem. The paper addresses the gap in theoretical guarantees by formulating sufficient conditions under which the HRE method can be correctly applied to both complete and incomplete pairwise comparison matrices.

Methodology
The authors employ tools from linear algebra and spectral theory to analyze the arithmetic and geometric variants of the HRE method. The study distinguishes between two subsets of alternatives: those with unknown priorities ( $A_U$ ) and reference alternatives with known weights ( $A_K$ ).

Mathematical Formulation: The problem is transformed into a system of linear equations ( $A_k w = b$ ). The solvability of this system depends on the invertibility of the auxiliary matrix $A_k$ .
Spectral Analysis: The authors utilize the spectral radius ( $\rho$ ) and the Perron-Frobenius theorem to derive conditions for matrix invertibility. They relate the invertibility of $A_k$ to the Consistency Index (CI) of the submatrix containing only the unknown alternatives ( $C_k$ ).
Inconsistency Thresholds: The research derives specific thresholds for the Consistency Index. If the inconsistency of the submatrix $C_k$ remains below a calculated limit, the matrix $A_k$ is guaranteed to be invertible, ensuring a unique solution.
Geometric vs. Arithmetic: The study analyzes both the arithmetic mean approach and the geometric mean approach (logarithmic transformation) for both complete and incomplete data scenarios.

Key Contributions and Results

Arithmetic HRE (Complete Case): The paper establishes that the arithmetic HRE formulation admits a unique solution if the Consistency Index of the unknown submatrix satisfies $CI(C_k) < \frac{n-k}{k-1}$ (where $n$ is the total number of alternatives and $k$ is the number of unknown alternatives). The authors demonstrate through a counter-example that this threshold is optimal and cannot be improved.
Arithmetic HRE (Incomplete Case): For incomplete matrices, the authors derive a more complex sufficient condition: $CI(C_k) < \frac{n-k-s_{MAX}+s_{MIN}}{k-1}$ . Here, $s_{MAX}$ and $s_{MIN}$ represent the maximum and minimum number of missing entries in the rows of the submatrix $C_k$ , respectively. This result accounts for the distribution of missing data. The paper provides an example showing that this threshold is also optimal.
Geometric HRE: In contrast to the arithmetic variant, the paper proves that the geometric HRE formulation always admits a unique solution for both complete and incomplete pairwise comparison matrices, provided the matrix is irreducible. This is demonstrated by showing that the spectral radius of the relevant auxiliary matrix is always less than 1, ensuring invertibility regardless of the inconsistency level.
Optimality of Estimates: The paper includes illustrative examples (Examples 16 and 22) where the derived inconsistency thresholds are reached, resulting in singular matrices. These examples serve to prove that the derived conditions are tight; relaxing them would allow for cases where the method fails.

Significance and Claims
The paper claims to fill a theoretical gap by providing formal guarantees for the applicability of the standard HRE method. Unlike previous studies that focused on extending the method through modified aggregation rules or fuzzy logic, this work adopts a theoretical perspective to determine under what conditions the standard arithmetic and geometric HRE methods are well-defined.

The authors emphasize that their contribution addresses a distinct problem: establishing the invertibility and uniqueness of solutions for the standard formulations. They assert that the derived inconsistency thresholds are optimal, meaning they represent the precise boundary between solvable and unsolvable instances for the arithmetic variant. For the geometric variant, the paper provides a robust guarantee of solvability independent of inconsistency levels. The research does not propose new applications or future experimental directions but rather solidifies the mathematical foundation required to apply existing HRE methods to real-world, incomplete decision-making scenarios.