Toward Explainable AI Traders: Analysis of Algorithmic Traders Created via Deep Learning on Level-2 Limit Order Book Data

We show that a Deep Learning Neural Network (DLNN) learns to be a high-performing algorithmic trading system, operating purely from training-data inputs generated by passive observation of an existing successful trader T. That is, we point our DLNN system at trader T and successfully have it learn from T 's trading activity, such that it then trades at least as well as T . Our system, DeepTrader, takes inputs derived from Level-2 market data, i.e. the market's Limit Order Book (LOB). Unusually, DeepTrader makes no explicit prediction of future prices. Instead, we train it purely on input-output pairs where in each pair the input is a snapshot S of Level-2 LOB data taken when T issued a quote Q (i.e. a bid or an ask) to the market; and DeepTrader's desired output is to produce Q when it is shown S. That is, we train our DLNN by showing it the LOB data S that T saw at the time when T issued quote Q; and in doing so our system comes to behave like T, acting as a profitable algorithmic trader. We evaluate DeepTrader against other algorithmic trading systems, including two that have repeatedly been shown to outperform human traders. DeepTrader matches or outperforms such pre-existing algorithmic trading systems. We analyze successful DeepTrader networks to identify what features are relied on and which features can be ignored. Our methods can in principle create an explainable copy of an arbitrary trader T via imitative deep learning methods.