The paper introduces a novel approach, Fin-GAN, which uses Generative Adversarial Networks (GANs) to forecast and classify financial time series. The authors propose a new economics-driven loss function for the generator, which places GANs in a supervised learning setting and enables them to produce full conditional probability distributions of price returns given historical data. This approach moves beyond traditional point estimates and allows for uncertainty estimates. The Fin-GAN method is evaluated on daily stock ETF-excess returns and raw ETF returns, outperforming classical models like LSTMs and ARIMA in terms of Sharpe Ratios. The proposed loss function also helps alleviate mode collapse and improves the model's ability to capture cross-asset interactions. The paper includes a detailed methodology, experimental setup, and performance metrics, demonstrating the effectiveness of the Fin-GAN approach.The paper introduces a novel approach, Fin-GAN, which uses Generative Adversarial Networks (GANs) to forecast and classify financial time series. The authors propose a new economics-driven loss function for the generator, which places GANs in a supervised learning setting and enables them to produce full conditional probability distributions of price returns given historical data. This approach moves beyond traditional point estimates and allows for uncertainty estimates. The Fin-GAN method is evaluated on daily stock ETF-excess returns and raw ETF returns, outperforming classical models like LSTMs and ARIMA in terms of Sharpe Ratios. The proposed loss function also helps alleviate mode collapse and improves the model's ability to capture cross-asset interactions. The paper includes a detailed methodology, experimental setup, and performance metrics, demonstrating the effectiveness of the Fin-GAN approach.