Detecting and Quantifying Wash Trading on Decentralized Cryptocurrency Exchanges

This repository contains the implementation of a wash trade detection process for data from the limit order book-based decentralized exchanges IDEX and EtherDelta.

Instructions

You will need at least 32GB RAM, or large enough swap space to run everything. It took about 2.5 hours to run everything on a Ryzen 5 3600.

Data

The following data files can be downloaded from Zenodo, and should be placed in the data directory:

• EtherDeltaTrades.csv: EtherDelta trades
• IDEXTrades.csv: IDEX trades
• EtherDollarPrice.csv: daily ETH-USD prices
• token_decimals.json: decimals used by tokens for precision

Preprocess Trades

In order to preprocess the trades, you must have Python 3 installed on your machine.

For IDEX, run:

IDEXtrades_preprocessing.py -i <idexfile> -d <decimalsfile> -o <outputfile>

and for EtherDelta, run:

EtherDeltatrades_preprocessing.py -i <etherdeltafile> -d <decimalsfile> -o <outputfile>

providing the paths to the data files mentioned above.

Run Pipeline

In order to run the pipeline, you must have R version 3.5.1 installed on your machine, as well as the following R packages:

• optparse
• data.table
• igraph
• rjson
• Rcpp
• hash

The pipeline can be called by running Rscript pipeline_wash_trading_paper.R [options], using the following options. The trades must be the preprocessed ones.

Options:
	-d DEX, --dex=DEX
		name of DEX, must be either 'IDEX' or 'EtherDelta' [default= IDEX]

	-t TRADES, --trades=TRADES
		trade dataset file name [default= data/IDEXTrades-preprocessed.csv]

	-p PRICES, --prices=PRICES
		Ether-Dollar-Price file name [default= data/EtherDollarPrice.csv]

	-o OUTPUT, --output=OUTPUT
		output folder name [default= output_IDEX]

	--sccthresholdrank=SCCTHRESHOLDRANK
		threshold for relevant SCC: rank [default= 100]

	--washdetectionether=WASHDETECTIONETHER
		should wash trades be detected for Ether amounts (TRUE) or Token amounts (FALSE) [default= TRUE]

	-m MARGIN, --margin=MARGIN
		margin of mean left trader position for wash trade detection [default= 0.1]

	--washwindowsizesecondspass1=WASHWINDOWSIZESECONDSPASS1
		wash trade detection window size for first pass in seconds [default= 604800]

	--washwindowsizesecondspass2=WASHWINDOWSIZESECONDSPASS2
		wash trade detection window size for second pass in seconds [default= NULL]

	--washwindowsizesecondspass3=WASHWINDOWSIZESECONDSPASS3
		wash trade detection window size for third pass in seconds [default= NULL]

	-h, --help
		Show this help message and exit

How we ran it for the paper

Preprocessing:

python IDEXtrades_preprocessing.py -i data/IDEXTrades.csv -d data/token_decimals.json -o data/IDEXTrades-preprocessed.csv
python EtherDeltatrades_preprocessing.py -i data/EtherDeltaTrades.csv -d data/token_decimals.json -o data/EtherDeltaTrades-preprocessed.csv

Wash Trade Detection

We use the following parameters:

• Perform the analysis based on token amounts, not Ether amounts
• Use an SCC threshold of minimum 100 occurences
• A margin of 1%
• Analysis windows: 1 hour, 1 day, 1 week

Executing wash trade detection for IDEX (this takes about 2 hours):

Rscript pipeline_wash_trading_paper.R -d "IDEX" \
-t data/IDEXTrades-preprocessed.csv \
-p data/EtherDollarPrice.csv \
-o output/idex-t100-1h-1d-1w-1pmargin \
--sccthresholdrank=100 \
--washdetectionether=FALSE \
-m 0.01 \
--washwindowsizesecondspass1=3600 \
--washwindowsizesecondspass2=86400 \
--washwindowsizesecondspass3=604800

Executing wash trade detection for IDEX (this takes about 15 minutes)

Rscript pipeline_wash_trading_paper.R -d "EtherDelta" \
-t data/EtherDeltaTrades-preprocessed.csv \
-p data/EtherDollarPrice.csv \
-o output/etherdelta-t100-1h-1d-1w-1pmargin \
--sccthresholdrank=100 \
--washdetectionether=FALSE \
-m 0.01 \
--washwindowsizesecondspass1=3600 \
--washwindowsizesecondspass2=86400 \
--washwindowsizesecondspass3=604800

Statistics and Plots

For plotting, you will need the following additional R packages:

• ggplot2
• scales
• RColorBrewer
• ggthemes
• extrafont

To create the plots, run paper_plots.R. Depending on your output file location, you may need to adjust some variables at the beginning of that file.