The impulse-response (IR) model describes the relationship between athlete training history and performance. The model features five parameters, with two derived parameters providing context for interpretation in terms of exercise training. Despite some past successes, IR models are often poorly fitted. Here we describe a novel Bayesian approach to fit the IR model. We discuss the elicitation of informative priors, and justify the assumption that performance is multivariate normal distributed. MCMC via Gibbs sampling was used to sample the posterior. The method was applied to an international-class middle-distance runner, for which training was quantified as TRIMPi and performance as IAAF points achieved in a sanctioned race. The method produced well-constrained estimates of the five parameters, but the posterior intervals of the derived parameters were too wide to make reliable training optimization decisions. We conclude our approach could further improve the fit of the IR model。