| paper | d4-conf-atal-2016-0309 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: Polynomial-time solvable in \(m,\tau,T,L\), including the exact initial value and polynomial-size circuits for one pair of SPE behavioral strategies. · discrete side: other · adjudication: proved_with_minor_gaps
full proof & adversarial review →
Lemma 1
statement extracted from the paper’s text layer
Given a fixed horizon \(T\), finite type sets \(S_A,S_B\), positive rational masses \(\mu_t,\nu_u\) summing to one, rational match probabilities \(p_{tu}\), and terminal zero-sum utility \(U(w)\), where each supported type has at least \(T\) interchangeable representatives, compute the exact SPE value and behavioral strategies when both teams choose a type simultaneously each round, use a fresh representative, observe the result, and adapt.
A two-team game with finite supported type sets \(S_A,S_B\), population masses \(\mu,\nu\), and match probabilities \(p_{tu}\). Each round the teams choose types, consume fresh representatives, and observe the match; the objective is to compute the exact SPE value and strategies. For fixed \(T\), positive masses ensure enough representatives, so play depends on the support rather than the mass magnitudes.
With fixed \(T\), every positive-mass type has enough representatives, so the masses do not affect the value; this limits the mirror’s connection to the paper’s finite-supply effects.
fatal: False
Covers Lemma 1 and the Section 2.3 SPE computation in the abundant-type regime; leaves the results about the strategic value and optimal number of redundant or dominated players unmirrored.
The paper has no named NP-hardness, polynomial-time, FPT, or similar complexity result. The closest computation-related anchor is Lemma 1, proved here: two histories have the same value if the same players have been selected by each team and Team 1 has won the same number of rounds, regardless of the order of past matches. Section 2.3 uses this to compute subgame-perfect equilibrium values by backward induction over states \((k,X,Y,w)\), solving a matrix game at each state. I would treat that lemma and computation as a weak, computation-adjacent anchor—not as a theorem that establishes a complexity class.
The strongest continuous mirror I can defend is High-Multiplicity Team-SPE. Imagine two organizations drawing representatives from large pools for \(T\) scheduled head-to-head contests. Each representative’s type is their complete win-probability profile against the other side’s types. Representatives of the same type are interchangeable; the type supports are small, while each supported type has at least \(T\) representatives. For example, the organizations might recruit from large pools of workers in a few standardized skill or certification bands.
An instance consists of \(T\), finite type supports \(S_A,S_B\), rational population proportions \(\mu_t>0\) and \(\nu_u>0\) summing to one, rational match probabilities \(p_{tu}\), and a rational zero-sum terminal utility \(U(w)\), where \(w\) is Team 1’s number of wins. Each round, the teams simultaneously choose a type; a fresh representative of that type plays, and Team 1 wins with probability \(p_{tu}\). The teams observe the result and adapt. The question is to compute the exact SPE value and equilibrium behavioral strategies. A solution gives the value and each team’s mixed choice of types at every subgame.
This is a genuine high-multiplicity regime: if a finite roster has proportions \(\mu,\nu\) and each positive-mass type has at least \(T\) representatives, it is the paper’s game after representatives are grouped by type. The round-by-round contest, random outcomes, adaptive choices, and payoff are unchanged. The continuous population is the distribution over types; the type masses guarantee sufficient supply. In the limit, the masses themselves matter only through which types have positive mass.
For this mirror, the state is just \((k,w)\): with enough representatives of each type, previously selected individuals impose no further constraint. Set \(V_{T,w}=U(w)\). At state \((k,w)\), form the matrix
\[
M^{k,w}_{t,u}
=
p_{tu}V_{k+1,w+1}
+
(1-p_{tu})V_{k+1,w},
\]
and let \(V_{k,w}\) be its zero-sum game value. Equilibrium strategies for that matrix give the teams’ actions at the state. This is an exact polynomial-time computation in \(T\), \(|S_A|\), \(|S_B|\), and the input bit length, using rational matrix-game LPs, when \(T\) is counted explicitly. It is a derived algorithm for the mirror, not a complexity result claimed by the paper.
This mirrors Lemma 1 and the computation in Section 2.3: order and named identities can be discarded once the state records what affects future play. It does not claim a mirror for Theorem 4’s result about how many dominated players can be useful, or for the paper’s other substantive results.
The soft spot is that this limit removes much of the paper’s distinctive phenomenon. With enough copies of every supported type, a team never runs out of a type; the exact masses cease to matter, and the finite-supply effects behind the dominated-player results may disappear. So this is a plausible continuous mirror of the paper’s SPE-computation problem, but not a strong continuation of its central results about the strategic value of redundancy.
The only anchor is Lemma 1 and the backward-induction computation in Section 2.3, and neither is a named complexity result. Lemma 1 says that histories with the same played-player sets and win count have the same value; Section 2.3 uses that fact to compute SPE values. The paper does not classify the complexity of computing an SPE, give an algorithmic theorem, or identify a hard computational subproblem. So the proposed “High-Multiplicity Team-SPE” is a new computational question inspired by the paper, rather than a continuous mirror of a computational result it establishes.
The strongest version of that mirror is the proponent’s fixed-\(T\) model with positive-mass types. But as the finite populations grow, every supported type has enough representatives to cover all \(T\) rounds. Then the masses \(\mu\) and \(\nu\) no longer affect play or value; only the supports, match probabilities, and utility function matter. The model has become a finite repeated matrix game. That is a sensible game, but it does not make population mass the computational object. Nor does it preserve the state compression of Lemma 1: the lemma still tracks the sets \(X,Y\) of players already used, while this mirror makes depletion irrelevant and reduces the state to \((k,w)\). The distinctive finite-supply feature has disappeared.
A more faithful attempt would retain type inventories and ask how the SPE value changes as counts grow. With a fixed number of rounds, however, a positive type mass supplies more than enough players in the limit, and depletion again disappears. To keep depletion relevant, the number of rounds must grow with population size. Then replacing each count by a mass makes a single match consume zero mass. Capturing the effect requires a new scaled-horizon dynamic game, with choices about how to represent the evolving score, observations, and terminal payoff. That could be interesting, but it is a new limit model rather than a direct continuation of the paper’s computation.
The negative case is therefore strongest on scope: the paper offers no named computational result for the programme to mirror, and its cleanest high-multiplicity version makes the population masses irrelevant. It is weaker against the broader claim that *some* worthwhile continuous question could be built from this game. A mass-scaled sequential game may be worth studying; the paper alone does not establish that it would be.
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.