Set of programs which ultimately produce a reduced-variance payoff matrix from samples of a continuous double auction (CDA) simulation.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data
lib
src/sim/cda
README
report.pdf

README

Set of programs which ultimately produce a reduced-variance payoff matrix from samples of a continuous double auction (CDA) simulation.

For additional background, see http://augiehill.com/blog/cda-strategies

The pipeline works as follows:
 - GatherResults: collects simulation samples from the Michigan CAEN Advanced Computing cluster. The collected samples are available in /data/samples.zip. It's ~750MB unzipped.
 - CalculateAverageNormalizedUnitValue: name says it all.
 - CalculateAveragePayoffPerStratPerProf: average payoff per strategy per profile.
 - CalculateCoefficients: the 10 control variate coefficients from the average normalized unit values and average payoffs per strategy per profile.
 - CalculateRedVarAvgPayoffPerStratPerProf: produces a payoff matrix which can then be analyzed by an external game analysis script
 - CalculateVarianceReductionRatio: by how much was variance reduced?

The ConvertPayoffMatrix script is used to convert the results of the original study to the format used by the game analysis script.

A detailed explanation of the variance reduction process is available in report.pdf.

The source code for the CDA simulator which produced the samples is available at https://github.com/augie/cda-simulator

The produced JSON payoff matrix is analyzed using the game analysis script available on GitHub at https://github.com/egtaonline/GameAnalysis