In this work, we present our findings of the so‐called CIR#, which is a modified version of the Cox, Ingersoll, and Ross (CIR) model, turned into a forecasting tool for any term structure. The main feature of the CIR# model is its ability to cope with negative interest rates, cluster volatility, and jumps. By considering a dataset composed of money market interest rates during turmoil and calmer periods, we show how the CIR# performs in terms of directionality of rates and forecasting error. Comparison is carried out with a revamped version of the CIR model (denoted
CIRadj), the Hull and White model, and the exponentially weighted moving average (EWMA) which is often adopted whenever no structure in data is assumed. To confirm the analysis, testing and validation is performed on both historical and ad hoc data with different metrics and clustering criteria.