How Bubble Experience Changes Reinforcement Learning Traders

Summary

This study investigates whether reinforcement-learning traders change their strategies after experiencing asset-price bubbles, and whether that learning affects later bubbles. The researchers train agents in the ABIDES multi-agent market simulation and compare traders trained with bubble experience against traders trained without it. Their reported behavioral distinction is that agents without bubble experience act like short-term momentum traders, while agents exposed to bubbles behave more like value traders.

In the simulation, inexperienced agents amplify bubbles, whereas experienced agents tend to suppress them and sometimes prevent them. The authors interpret this as a possible boom-and-bust mechanism: the memory of a bubble collapse can make new bubbles less likely for a time, with their likelihood rising again as that experience fades. These findings come from a simulated market and trained agents, so they do not establish how human traders or live markets would behave. The summary also gives no details about the simulation’s parameter settings, robustness checks, or the duration of the memory effect.

Key ideas

• The study compares reinforcement-learning traders trained with and without bubble experience in ABIDES.
• Traders without bubble experience are reported to follow short-term momentum behavior.
• Traders with bubble experience behave more like value traders.
• In the simulation, inexperienced agents amplify bubbles while experienced agents tend to suppress them.
• The authors propose that fading memory of a crash may allow bubble risk to return over time.

Tags

# Once Burned, Twice Shy? The Effect of Stock Market Bubbles on Traders that Learn by Experience


# Once Burned, Twice Shy? The Effect of Stock Market Bubbles on Traders that Learn by Experience









We study how experience with asset price bubbles changes the trading strategies of reinforcement learning (RL) traders and ask whether the change in trading strategies helps to prevent future bubbles. We train the RL traders in a multi-agent market simulation platform, ABIDES, and compare the strategies of traders trained with and without bubble experience. We find that RL traders without bubble experience behave like short-term momentum traders, whereas traders with bubble experience behave like value traders. Therefore, RL traders without bubble experience amplify bubbles, whereas RL traders with bubble experience tend to suppress and sometimes prevent them. This finding suggests that learning from experience is a mechanism for a boom and bust cycle where the experience of a collapsing bubble makes future bubbles less likely for a period of time until the memory fades and bubbles become more likely to form again.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.