Battlefield Transfers in Coalitional Blotto Games

The paper’s only named result is an unnumbered measure-zero statement, with no finite computational problem or complexity claim, so bit (a) fails. The proposed population mirror is also degenerate: divisible battlefield mass affects payoffs only through \(\phi_1(\mu)\) and \(\phi_2(\mu)\), making \(\mu\) a verbose encoding of the original four-parameter game. Thus no anchor supplies both a qualifying computational result and a population-continuum question.

The objection that survived

The opponent’s decisive objection is that the type distribution \(\mu\) is quotiented out by divisibility and aggregation; retaining type-sensitive information would require constraints or technologies absent from the paper and would change its theorem.

What the mirror covers

The proposed mirror reaches only the main joint-transfer existence theorem; it leaves the budget-only comparison, numerical illustrations, case analysis, and external results from [13] untouched.

Open questions for a prover

The case FOR (proponent)

There is a real, but qualified, positive case here. The qualification is important: the paper contains no named computational-complexity result. Its only named result is the unnumbered Theorem in Section 3, proved in this paper:

Let \(G^{b,v}_0\) be the set of coalitional Blotto games with no mutually beneficial joint transfer. Then \(G^{b,v}_0\) has measure zero.

The paper has no numbered theorem, lemma, corollary, or proposition asserting membership in \(P\), NP-hardness, fixed-parameter tractability, or a related complexity classification. I therefore use this unnumbered Theorem as the sole anchor, and do not count the results cited from [13] as results of this paper.

My lead mirror is Continuous Joint-Transfer Existence.

Take a large population of campaign districts, electoral micro-contests, or comparable battlefield units. There are finitely many complete district types \(T=T_1\cup T_2\). A type \(t\) records its current player \(i\in\{1,2\}\), its battlefield-value density \(v_t>0\), and all other public characteristics relevant to the contest. A society is a distribution \(\mu\in\Delta(T)\), where \(\mu_t\) is the fraction of districts of type \(t\). Thus a discrete election with \(N\) districts and \(n_t\) districts of each type gives \(\mu_t=n_t/N\). The intended regime is \(N\gg |T|\): millions of interchangeable local contests, but only a moderate number of district types.

The aggregate battlefield valuations are

\[ \phi_i(\mu)=\sum_{t\in T_i}\mu_t v_t. \]

A joint transfer consists of a budget transfer \(\tau_b\in(-X_1,X_2)\), positive when budget moves from Player 1 to Player 2, together with fractions \(z_t\) of districts transferred between the players. If \(z_t^+\) is mass transferred from Player 1 to Player 2 and \(z_t^-\) is mass transferred in the opposite direction, the induced battlefield-value transfer is

\[ \tau_v = \sum_{t\in T_1}v_tz_t^+ - \sum_{t\in T_2}v_tz_t^-. \]

The post-transfer parameters are

\[ X'_1=X_1-\tau_b,\qquad X'_2=X_2+\tau_b, \]

and

\[ \phi'_1=\phi_1-\tau_v,\qquad \phi'_2=\phi_2+\tau_v. \]

The common adversary then chooses \(a_1,a_2\ge 0\), with \(a_1+a_2\le 1\), to maximize its equilibrium payoff. Player \(i\)'s payoff is the paper's Blotto payoff

\[ p_i(\phi_i',X_i',a_i)= \begin{cases} \phi_i'\dfrac{X_i'}{2a_i},&X_i'\le a_i,\[6pt] \phi_i'\left(1-\dfrac{a_i}{2X_i'}\right),&X_i'>a_i. \end{cases} \]

Let \(U_i(\mu,\tau_b,z)\) denote the resulting payoff under the paper's equilibrium and a fixed tie-breaking convention on the measure-zero boundary cases. Let \(U_i^0\) be the payoff with no transfer. The continuous problem is

\[ \Delta^*(\mu,X_1,X_2) = \sup_{\tau_b,z} \min_{i\in\{1,2\}} \bigl(U_i(\mu,\tau_b,z)-U_i^0\bigr). \]

The task is to decide whether \(\Delta^*>0\), and, if so, output a feasible \((\tau_b,z)\) witnessing strictly higher payoff for both players. This is exactly the paper's Stage 1 question, with battlefield mass replacing a finite collection of named districts.

The mirror is plausible because it uses the paper's own political-campaign interpretation rather than importing an unrelated story. The paper already treats the number of battlefields as arbitrarily large, treats battlefield valuation as divisible, and explicitly discusses transferring voting districts between parties. The continuous population simply makes the implicit high-multiplicity regime explicit: many districts share the same value and contest characteristics, and a transfer acts on a fraction of that mass.

I would expect this problem to be Class A. Given \(\mu\), one computes \(\phi_1(\mu)\) and \(\phi_2(\mu)\) in time polynomial in \(|T|\). The paper's proof then identifies one of four adversarial-response cases, computes the two payoff gradients, and chooses a common ascent direction. A sufficiently small step gives a mutually beneficial transfer. Since the strategic variable is only two-dimensional, and the payoff formulas are explicit piecewise algebraic expressions, turning this into an exact polynomial-time algorithm looks credible, although the paper itself does not prove a bit-complexity bound.

The population distribution is not entirely redundant: it is the representation from which aggregate battlefield values and realizable transfer mass are computed, and it provides the high-multiplicity bridge back to finite district systems. If a continuous witness improves both payoffs by margin \(\eta>0\), rounding it to whole districts should preserve improvement once each type has sufficiently large multiplicity relative to the rounding error. Conversely, rational distributions \(\mu_t=n_t/N\) are precisely the normalized high-multiplicity instances.

The scope is deliberately narrow. This mirrors only the paper's main measure-zero existence theorem, not every statement about budget-only transfers, and not the external characterization from [13]. It generates several useful follow-up questions: whether the exact decision and witness problem is polynomial in the bit model; how the exceptional sets in equations (4)–(9) pull back to the distribution simplex; and whether type-dependent campaign costs or restrictions on which district types may be transferred produce a genuinely richer LP or pricing problem.

The weakest point is that the paper's theorem is analytic, not computational, and its payoff depends only on \((\phi_1,\phi_2,X_1,X_2)\). A sceptical referee could therefore say that \(\mu\) is merely a verbose encoding of two aggregate numbers, while the paper's real continuity concerns divisible battlefield value rather than a continuous population. That objection is serious. The positive case survives only as a modest Class-A mirror: the paper supplies a natural high-multiplicity campaign regime and a precise continuous transfer problem, but it does not by itself supply the kind of computational complexity landscape that ChoCo is ultimately seeking.

The case AGAINST (opponent, writing after the proponent)

The negative case is stronger than the proponent admits: there is no qualifying computational anchor in the paper. Its only result is the unnumbered Theorem that the set of games without a mutually beneficial joint transfer has measure zero. That is an analytic genericity statement over \(G\subset\mathbb{R}^4\), not a computational result: the paper specifies no finite input representation, decision or optimization problem, bit model, complexity bound, approximation guarantee, or algorithmic theorem. The gradient observation is an informal procedure for generic instances, not a named computational result. Defining \(\Delta^*>0\) afterwards manufactures a new problem rather than mirroring one stated by the paper.

Even if that source objection is waived, the proposed population mirror collapses. In the paper, battlefields are not agents with preferences or identities. They are divisible resources, and every payoff and adversary response depends on them only through the two aggregate valuations \(\phi_1\) and \(\phi_2\), together with \(X_1\) and \(X_2\). In the proponent’s notation, define

\[ \pi(\mu)=\left(\sum_{t\in T_1}\mu_t v_t,\sum_{t\in T_2}\mu_t v_t\right). \]

For any two distributions with the same \(\pi(\mu)\), the payoff functions, adversary allocation, feasible range of \(\tau_v\), and value of \(\Delta^*\) are identical. Because type mass is divisible, every value transfer in \((-\phi_2,\phi_1)\) can be implemented by taking suitable fractions of the available types. The variables \(z_t\) merely provide a verbose encoding of the scalar \(\tau_v\).

Thus the proposed society distribution is not the computational object. It is compressed immediately to four numbers, \((\phi_1,\phi_2,X_1,X_2)\), exactly the paper’s original parameterization. This is a genuine continuum degeneracy: the population axis disappears, rather than becoming a high-multiplicity instance on which pricing, rounding, or population-sensitive complexity could arise. The issue is not that the resulting question is easy; it is that \(\mu\) contributes no information to the question.

Calling the battlefield units electoral districts does not repair this. The districts still have no preferences, strategic behavior, or individual welfare; they are transferable pieces of battlefield valuation. Replicating the three strategic agents instead is no better: the theorem depends on two named players and one named adversary, and replacing them by exchangeable types changes the coalition and payoff semantics. Replicating battlefields preserves the original model, but that multiplicity has already been quotiented out by the paper’s assumption of arbitrarily divisible battlefield value.

The only way to make \(\mu\) matter is to add restrictions absent from the paper: type-dependent transfer costs, spatial or eligibility constraints, indivisible transfer blocks, type-specific campaign technologies, or an adversary that allocates resources across battlefield types. Those may define worthwhile new Blotto problems, but they break the paper’s central reduction to \((\phi_1,\phi_2,X_1,X_2)\) and invalidate the four-case gradient proof. They are extensions or re-modelings, not mirrors of the paper’s theorem. The dichotomy is unavoidable: preserve the theorem’s semantics and the population disappears; preserve population information and the theorem is no longer the source result.

There is also no canonical “almost everywhere” society theorem here. The paper’s null set is measured in the four-dimensional parameter space \(G\). The map from a type simplex to \(G\) has positive-dimensional fibres, and the preimage of a null set under such a map need not be null. A chosen type space may even constrain all societies to a lower-dimensional subset of \(G\). Pulling the exceptional curves back to \(\Delta(T)\) would therefore require a new modelling and measure argument, not a routine translation.

A Blotto-inspired continuous population model could certainly be invented, especially with heterogeneous and constrained battlefield types. But this paper supplies no computational result to anchor it, and its own joint-transfer theorem survives only after eliminating the population distribution. On the programme’s standards, the sole proposed anchor therefore fails both the source gate and the non-degeneracy test.

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.