| paper | d4-conf-aldt-2024-0010 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The finite LP computation is reported in prose, while the numbered result the proponent cites gives bounds on incentive properties, not a computational result under the stated gate. The proposed cohort model also changes a reversal of one match between two agents into a collective reversal between types, so it does not supply a surviving mirror anchored in the paper.
fails bit a — no named computational result to mirror
No result survives as an anchored continuous computational mirror; the paper's fairness, manipulation, and rule existence results are left without a qualifying anchor.
The best positive angle is that tournament participants really can form a large population, and the paper studies what happens as their number grows. But I cannot give a fully compliant positive case under your anchor rule: the paper has no numbered result classifying a computational problem as tractable, hard, or parameterized. Its finite LP computation for \(\lambda=1\) and up to six agents is reported in prose after Corollary 1, not as a numbered result. The numbered results concern existence and incentive properties of tournament rules. [The paper’s full text](https://arxiv.org/html/2408.10346v1)
The closest quantitative result is Theorem 3.3, proved in this paper. It says: if ICR satisfies \(2\text{-}\mathrm{NM}_{\lambda}\text{-}\alpha\), then \(\lambda\geq(1-O(\alpha))\Omega(n^2)\); if RDM does, then \(\lambda\geq(1-O(n\alpha))\Omega(n)\); and if RVC, TCR, or PRSL does, then \(\lambda\geq(1-O(\alpha))\Omega(n)\). This is a lower bound on the selfishness needed to deter manipulation in those rules, not a complexity result, so it is only a nearby result—not a valid anchor under the stated criterion. [Theorem 3.3](https://arxiv.org/html/2408.10346v1#S3)
If that structural result were allowed as an anchor, the strongest concrete candidate I see is Mass-Weighted TCR Robustness. Its instance is a finite set \(T\) of competitor types, rational masses \(\mu_t>0\) summing to one, and a tournament \(H\) on the types. A type must capture everything relevant to the rule and incentives; the regime would be a large league with \(N\gg |T|\), such as repeated cohorts of near-identical entrants. The continuous rule selects winner mass \(\rho_t=\mu_t/\sum_{s\in TC(H)}\mu_s\) for each type in the top cycle. Write \(\pi_t=\rho_t/\mu_t\) for the type’s per-capita winning intensity.
The question is: over every type tournament \(H\) and every pair of types \(a,b\), what is the smallest \(\lambda\) that prevents a profitable cohort-level reversal of their match outcome? More precisely, compare \(H\) with \(H'\), obtained by reversing the outcome between the two cohorts, and compute the least \(\lambda\) satisfying
\[
\pi_a(H')+\pi_b(H')\leq \pi_a(H)+\pi_b(H)
+\lambda\max\{\pi_a(H)-\pi_a(H'),\pi_b(H)-\pi_b(H')\}.
\]
The solution would be that threshold and a tournament-and-pair witness, or a certificate that no reversal violates the bound. For this explicitly defined TCR variant, enumerating the type tournaments gives an FPT algorithm in \(|T|\). That is a property of this proposed mirror; Theorem 3.3 does not establish it.
The weakest point is also decisive: the paper studies two individuals reversing one match. In a type-based continuum, changing one individual match has zero mass and is invisible to the type-level input. The candidate therefore makes a deviation by whole cohorts, which is a materially different manipulation question. A continuous tournament kernel could retain individual match changes, but then it would need extra structure to prevent a single-edge change from disappearing in the limit.
So the population-continuum story is plausible for large, genuinely homogeneous leagues, and the paper’s asymptotic bounds give it a useful starting point. But there is no named computational anchor here, and the concrete question that survives continuization changes the paper’s pairwise action into cohort manipulation. I would present this as a tentative extension, not as a strong ChoCo mirror.
The proponent’s only named anchor, Theorem 3.3, is not a computational result in ChoCo’s sense. It bounds how large \(\lambda\) must be for several named rules to resist manipulation; it gives no complexity classification or algorithmic problem. The paper’s finite LP computation up to six agents and its open question about whether a rule exists for every \(n\) concern rule existence, not the complexity of bribery, control, or robustness on a society input. That leaves no compliant anchor in the case presented.
The deeper obstacle is that this paper’s population consists of the tournament’s vertices. Each agent’s outcomes are defined through matches against particular other agents; unlike voters with rankings over a fixed candidate set, agents do not come with an exogenous type from a fixed finite space. A type based on an agent’s full pattern of wins and losses depends on who else is in the tournament. Large sports fields are plausible, but that alone does not supply high multiplicity: a cohort of “same type” competitors must also have coherent outcomes against one another and against every other cohort. Those interactions can affect the very tournament rules under study.
Mass-Weighted TCR is a coherent *new* cohort model, but its deviation changes the paper’s problem. The paper compares tournaments differing in one match between two agents. Reversing the outcome between two cohorts changes all their cross-cohort matches. If instead only two representatives reverse their match, the result no longer has the type-level description \(H\): the deviation creates an individual exception that the mass model cannot record. A dense continuous interaction model has the related problem that changing one edge between two measure-zero agents does not change its limit object. Preserving that deviation would require extra local or identity-level structure, beyond a distribution over types.
Nor does the proposed calculation rescue the anchor. For a supplied type tournament and pair, one can compare the two top cycles and evaluate the displayed inequality directly. Searching over all type tournaments by enumeration gives a parameterized procedure in \(\tau\), but it solves the proposed cohort-deviation problem, not a computational question established by Theorem 3.3. Changing the objective to optimize a rule over a distribution of tournaments would move the distribution to the *instances*, rather than making the agents a continuous society.
So the negative case is strongest against a ChoCo mirror of this paper: its central results are about axioms and rule design over identity-indexed tournaments, and the concrete candidate becomes a collective cohort manipulation problem. I would not claim the universal no-value conclusion is airtight. A carefully specified model of homogeneous tournament cohorts could be worthwhile adjacent work. The case against it as a continuous-population computational mirror is much stronger than the case against it as research of any kind.
The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.