| paper | d4-conf-aaai-2015-0112 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Proposition 1 compares immediate trade costs but does not assert that a computational problem can be solved or is hard. The simulations report market outcomes without a numbered computational guarantee, so no anchor meets bit (a).
fails bit a — no named computational result to mirror
The sole numbered result, Proposition 1, is a one-step market-property comparison; the simulation findings on price discovery, spreads, volume, surplus, and loss have no stated computational guarantees and are not mirrored.
The paper offers a plausible high-multiplicity setting: a large prediction market could have many background traders sharing a small number of posterior-value types. Mass would represent each type’s share of traders. But that does not provide a qualifying anchor under your criteria.
The only numbered result is Proposition 1, proved here: when the LMSR market maker and order books start in the same state in PAR and INT, every next arriving trade incurs an immediate cost at least as high in INT as in PAR. This is a one-step market-property comparison, not a result classifying the complexity of a computational problem. The paper’s other findings come from simulations and are not stated as numbered computational results.
One could average Proposition 1 over a distribution of trader types, but its pointwise inequality already gives that average, and the population distribution does no work in the result. That would be an analytical restatement, not a useful continuous computational mirror. So I can’t honestly give a prover-ready continuous problem per anchor: the paper contains no named computational result that qualifies under the requested anchor rule.
The proponent identifies the decisive obstacle: the paper has no named result about the complexity of a computational problem to mirror. Proposition 1 compares the immediate cost of the next trade in INT and PAR, given identical market-maker states and order books. Its inequality holds trade by trade, so averaging over a distribution of trader types adds nothing; the population distribution does no work in the result.
The strongest alternative would be to model a large market with many traders sharing a few posterior-value types, then ask about expected cumulative loss or price discovery. But that is a new stochastic market-dynamics problem. To preserve the paper’s mechanism, it needs the evolving order books, order quantities, and arrival priority as well as trader types. The distribution of types alone does not describe that state. A fluid or mean-field treatment might be worthwhile, but it would be a separate analytic model, not a continuous mirror of a result in this paper. And despite its title, “continuous” here describes the auction’s operation over time, not a continuous population.
So no anchor survives: the one numbered proposition is not a computational-complexity result, and the simulations report finite-horizon outcomes rather than an algorithmic problem. The honest weakness in this negative case is its universal scope: someone could formulate a valuable computational question for a continuous-population market. But that would originate in a new model, not in a named computational result this paper supplies.
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.