Translocation of biopolymers through pores is implicated in many biological phenomena. Confinement within a pore often breaks ergodicity on experimental and/or biological time scales by creating large entropic barriers to conformational rearrangements of the chain. Here, we study one example of such hindered rearrangement, in which the chain reverses its direction inside a long pore. Our goal is twofold. First, we study the dependence of the time scale of polymer reversal on the pore size and on the polymer length. Second, we examine the ability of simple one-dimensional theories to quantitatively describe a transition in a system with a complex energy landscape by comparing them with the exact rate constant obtained using brute-force simulations and the forward flux sampling method. We find that one-dimensional transition state theory (TST) using the polymer extension along the pore axis as the reaction coordinate adequately accounts for the exponentially strong dependence of the reversal rate constant on the pore radius r and the polymer length N, while the transmission factor, i.e., the ratio of the exact rate and the TST approximation, has a much weaker power law r and N dependence. We have further attempted to estimate the transmission factor from Kramer's theory, which assumes the reaction coordinate dynamics to be governed by a Langevin equation. However, such an approximation was found to be inadequate. Finally, we examine the scaling behavior of the reversal rate constant with N and r and show that finite size effects are important even for chains with N up to several hundreds.