Skip to main content

Trade Labelling Module

The label.py module provides a single function, label_trades, that accepts raw DEX trade data and applies a series of rule-based checks to label each trade as wash trade or not.

These rules are imported from token_rules.py, allowing us to maintain a clean separation between rule logic and labeling orchestration.

Purpose​

The goal of this module is to return a labeled DataFrame with a new column is_wash_trade, where:

  • 1 means the trade is flagged as wash trade
  • 0 means the trade is considered normal

Understanding Code Logic​

Imports and Function Initialisation​

import pandas as pd
from token_rules import (
    detect_self_trades,
    detect_repeated_pairs,
    detect_loops,
    detect_spoofing,
    get_suspicious_summary
)

def label_trades(trade_data):
    # All the code written below is placed here
  • Imports all rule functions from token_rules.py.
  • Uses pandas to work with structured tabular data.

Flattening Trade Data​

Converts the nested JSON object into a flat pandas Dataframe.

    df = pd.json_normalize(trade_data)

Applying Rules​

Each function returns a filtered DataFrame of suspicious trades that match one rule and finaly returns suspicious trades.

    self_trades = detect_self_trades(df)
    repeated_pairs = detect_repeated_pairs(df)
    loops = detect_loops(df)
    spoofing = detect_spoofing(df)

    suspicious_tokens, suspicious_tx, suspicious_wallets = get_suspicious_summary(
        self_df=self_trades,
        repeated_df=repeated_pairs,
        loops_df=loops,
        spoofed_df=spoofing,
        original_df=df
    )

Return Labeled Data​

The suspicious trades are labelled as wash trades and returned.

    df["is_wash_trade"] = (
        df["Trade.Buy.Account.Address"].isin(suspicious_wallets) |
        df["Trade.Sell.Account.Address"].isin(suspicious_wallets) |
        df["Transaction.Signature"].isin(suspicious_tx)
    )

    return df