Forecasting Daily Highs and Lows of Liquid Assets with Neural Networks

We use Historically Consistent Neural Networks (HCNN) to forecast intraday highs and lows of liquid and volatile stocks. To build our forecast model we only use easily available open-high-low-close (OHLC) data. This is a novel application of HCNN to intraday data. It is important to note that model performance evaluation does not need tick data, which is more difficult to obtain and to handle. However, there is only few academic literature on forecasting intraday high-lows with neural networks. The present study aims at closing this gap. We measure the economic performance of a strategy using forecast high-low data. The strategy is intraday. It exits all positions at the close. This reduces the risk of being caught in abrupt price moves without the ability to exit the position. We test the strategy on a sample of S&P500 stocks. It turns out that profit and reward to risk ratios are attractive and confirm the good results of previous studies on an emerging market.