Cloud gaming streaming services, such as GeForce NOW, offer a vast collection of game titles, but onboarding or optimizing these titles can be time-consuming. This research proposes a proactive solution to identify game titles that will maximize the time invested by predicting which games will substantially increase (or decrease) in popularity relative to others. We leverage the Prophet forecasting library to forecast future demand for up to 24 weeks on the top 500 games on Steam. We use Ray[Tune] to perform parallelized training of forecasters and HyperOpt to hyper-tune parameters, allowing us to select the best model for each game. We evaluate our tuning results by showing that the average individual game forecast outperforms a baseline naive forecasting model. Additionally, we normalize the trend of each game relative to all games combined, which enables the shortlisting of outlier games with extreme trends. Our approach provides insights that can inform decision-making for game streaming services and game developers alike.