We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:


References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Science and Game Theory

Title: Strictly Proper Contract Functions Can Be Arbitrage-Free

Abstract: We consider mechanisms for truthfully eliciting probabilistic predictions from a group of experts. The standard approach -- using a proper scoring rule to separately reward each expert -- is not robust to collusion: experts may collude to misreport their beliefs in a way that guarantees them a larger total reward no matter the eventual outcome. Chun and Shachter (2011) termed any such collusion "arbitrage" and asked whether there is any truthful elicitation mechanism that makes arbitrage impossible. We resolve this question positively, exhibiting a class of strictly proper arbitrage-free contract functions. These contract functions have two parts: one ensures that the total reward of a coalition of experts depends only on the average of their reports; the other ensures that changing this average report hurts the experts under at least one outcome.
Comments: 11 pages
Subjects: Computer Science and Game Theory (cs.GT)
Cite as: arXiv:2109.01308 [cs.GT]
  (or arXiv:2109.01308v2 [cs.GT] for this version)

Submission history

From: Eric Neyman [view email]
[v1] Fri, 3 Sep 2021 04:47:53 GMT (13kb,D)
[v2] Mon, 6 Sep 2021 03:23:07 GMT (13kb,D)

Link back to: arXiv, form interface, contact.