| paper | d4-conf-wine-2022-0019 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The paper’s named theorems establish strategic properties of Randomized Death Match, but none states a runtime, computational hardness, approximation guarantee, or a formulation used to compute. The proponent also acknowledges that no anchor meets the required computational criterion, so no continuous mirror can pass bit (a).
fails bit a — no named computational result to mirror
No ChoCo-style continuous computational mirror is supported by the named results; Theorems 4.1, 5.2, and 5.3 remain strategic results about finite tournaments.
The paper has no named computational-complexity result of the kind required for an anchor: it proves no problem is in P, NP-hard, FPT, or otherwise classified by computational complexity. Its named theorems concern strategic properties of a tournament rule.
The closest population-related result is Theorem 5.3, proved here: for any finite tournament \(T\) and team \(u_1\) that is not a Condorcet winner, the maximum collective probability that \(u_1\) or its \(m-1\) Sybils wins tends to zero as \(m\to\infty\). The theorem proves the stronger claim that the probability that a Sybil or a team \(u_1\) beats tends to zero. Theorem 5.2, also proved here, says that the Sybils cannot increase their collective winning probability by changing matches among themselves. Theorem 4.1, proved here, gives the exact bound \(\alpha^{\mathrm{RDM}}_3=31/60\) for three-team manipulation. These are meaningful strategic results, but none is a computational-complexity anchor under the stated criterion.
So the strongest honest positive foothold is the growing cohort of identical Sybil teams in Theorem 5.3. Turning it into a continuous computational question would require defining a population version of Randomized Death Match and studying a coalition with positive mass. That would change the theorem’s setting: its focal coalition grows while the rest of the base tournament stays finite, and a fixed three-team coalition would have zero mass in a nonatomic limit. I therefore can’t give an anchor-based continuous problem or predict its complexity class without presenting a new, unsupported model as the paper’s mirror.
The strongest negative case is that the paper has no computational problem whose complexity it classifies. Theorem 4.1 gives a tight strategic guarantee for RDM; Theorem 5.2 proves an invariance under Sybil-internal match changes; Theorem 5.3 gives an asymptotic strategic property. None gives a runtime, hardness, or algorithmic result for an input problem. The proofs’ calculations do not supply that missing anchor.
Theorem 5.3 is the closest fit to a population story, but its limit exposes a problem with the mirror. It adds Sybils while keeping the base tournament finite, so the Sybil fraction tends to one and the base teams’ fraction tends to zero. Yet the theorem says the Sybils’ collective winning probability tends to zero when the original team is not a Condorcet winner: the vanishingly small base tournament determines who wins. The same limiting mass concentrated on the Sybil type can arise with no such opponent, when the Sybils win. A distribution over types alone therefore loses information that controls the result. Keeping the exceptional teams as named individuals gives a finite-core-plus-continuum model; giving them positive mass changes the theorem’s regime. Either could motivate new work, but neither is a continuous mirror of the stated result.
Theorem 5.2 has the same mismatch in a sharper form: it concerns whether identities can alter pairwise outcomes within a finite clone set. A continuous type mass does not specify those within-type match outcomes. Adding a type-to-type outcome kernel or a graphon could specify them, but that supplies a new model and a new question; the theorem itself classifies no computation. For Theorem 4.1, keeping its three manipulators makes them a zero-mass coalition in a continuum. Giving the coalition positive mass requires scaling its size with the population, or treating three bulk classes as manipulators. Both change the object from the fixed three-team deviation in the theorem, and require rules for matches within and between large classes.
A graphon version of RDM, with an explicit account of how its finite random elimination process behaves near the last survivor, is the best escape hatch. It may be an interesting adjacent project. But it would be a new stochastic population model, not a computational result from this paper translated to continuous society. I would not claim that no worthwhile extension is conceivable; I would claim that none of these three results provides a ChoCo-style continuous computational mirror.
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.