In this paper, we consider a generalisation of the Hobson–Rogers model proposed by Foschi and Pascucci (Decis Eocon Finance 31(1):1–20, <CitationRef CitationID="CR9">2008</CitationRef>) for financial markets where the evolution of the prices of the assets depends not only on the current value but also on past values. Using differentiability of stochastic processes with respect to the initial condition, we analyse the robustness of such a model with respect to the so-called offset function, which generally depends on the entire past of the risky asset and is thus not fully observable. In doing this, we extend previous results of Blaka Hallulli and Vargiolu (<CitationRef CitationID="CR3">2007</CitationRef>) to contingent claims, which are globally Lipschitz with respect to the price of the underlying asset, and we improve the dependence of the necessary observation window on the maturity of the contingent claim, which now becomes of linear type, while in Blaka Hallulli and Vargiolu (<CitationRef CitationID="CR3">2007</CitationRef>), it was quadratic. Finally, in this framework, we give a characterisation of the stationarity assumption used in Blaka Hallulli and Vargiolu (<CitationRef CitationID="CR3">2007</CitationRef>), and prove that this model is stationary if and only if it is reduced to the original Hobson–Rogers model. We conclude by calibrating the model to the prices of two indexes using two different volatility shapes. Copyright Springer-Verlag 2013
