Adaptive Multi-strategy Market-Making Agent for Volatile Markets

Crypto-currency market uncertainty drives the need to find adaptive solutions maximize gain or at least avoid loss throughout periods of trading activity. Given high dimensionality and complexity state-action space in this domain, it can be treated as a “Narrow AGI” problem with scope goals environments bound financial markets. Adaptive Multi-Strategy Agent approach for market-making introduces new solution positive “alpha” long-term handling limit order book (LOB) positions by using multiple sub-agents implementing different strategies dynamic selection these agents based on changing conditions. AMSA provides no specific strategy its own while being responsible segmenting activity into smaller execution sub-periods, performing internal backtesting historical data each doing sub-agent performance evaluation re-selection them end sub-period, collecting returns losses incrementally. With approach, return becomes function hyperparameters such granularity (refresh rate), sub-period duration, number active sub-agents, their individual strategies. Sub-agent next is made return/loss alpha values obtained during well real trading. Experiments have been performed under conditions relying proved probability case properly selected hyperparameters.