An Application of HodgeRank to Predicting the Outcome of Competitive Events: A Case Study of Horse Racing

Abstract

Financial markets rely heavily on the informational efficiency of their participants, setting security prices by aggregating the decisions made by these participants. Errors in the decision making process can pervade financial markets, creating opportunities for savvy traders to exploit and achieve excess risk-adjusted returns.

Systematic errors arise when the complexity of a decision problem exceeds the capabilities of an individual who, in turn, employs heuristics to reduce the complexity of the task. These so called cognitive biases have been observed in human behaviour affecting attitudes to risk, reliance on certain information sources, and the veracity of judgements.

Inconsistent data, containing conflicting information, increases the complexity of decision problems and the likelihood that individuals will make sub-optimal decisions. This project studies the impact on decision making and market efficiency of inconsistent ranking data, data used for ranking purposes containing intransitive patterns of preferences.

HodgeRank, a topologically-inspired ranking algorithm, is used to understand and explore inconsistent ranking data (Jiang et al. 2011). This technique separates pairwise comparison matrices into consistent and inconsistent components, and derives a ranking solution from the consistent ranking data. This project extends the algorithm to account for (i) the reliability of the information contained in the data and (ii) information contained in the inconsistent ranking data.

A study of parimutuel horserace wagering markets is undertaken to establish whether inconsistent ranking data is fully accounted for in real world settings. The results of a statistical and economic evaluation
demonstrate that the presence of inconsistent ranking data reduces the quality of decisions made by bettors and that market inefficiencies exist as a result. HodgeRank, in conjunction with conditional logit models and Kelly wagering strategies, is capable of exploiting this inefficiency and achieving abnormal returns.

More information

Identifiers

ORCID for Ruben Sanchez Garcia: orcid.org/0000-0001-6479-3028

Catalogue record

Export record