A paper worth knowing about if you're into quant trading and generative AI:
TRADES: Generating Realistic Market Simulations with Diffusion Models
Authors: Berti, Prenkaj, Velardi (Sapienza University of Rome + TU Munich)
What's inside:
> A transformer-based diffusion model for generating orders in the Limit Order Book (LOB) - realistic simulation of the market order book
> Two key metrics - realism (how close it is to real market behavior) and responsiveness (how the market reacts to an agent's actions)
> Results: ×3.27 and ×3.48 improvement over SoTA on the predictive score across two stocks
> DeepMarket - the first open-source Python framework for deep-learning-based LOB simulations
> Comes with a synthetic LOB dataset generated by the model itself
A step toward backtesting trading strategies and training RL agents on data that behaves like a real market instead of random noise.
They keep having sleepless nights over another man’s success. You’re just a B!tch ass nigga if another man’s success and wealth gives you headache. 🤕 🤣