Are Renko charts suitable for algorithmic trading? What are their advantages and disadvantages? How can they be represented for convolutional neural networks (CNNs)? We contrast a classical trading approach (Renko charts) with a machine learning approach (CNN) and then integrate these two approaches into a single method. We show that Renko charts offer an opportunity to train a CNN classifier with noise filtered out and demonstrate how they can be applied in algorithmic trading. We use Kibot as a source of historical market data and evaluate our approach with 50 stocks and exchange-traded funds (ETFs) that have the highest trading volume.