A polymer chain containing $N$ monomers confined in a finite cylindrical tube of diameter $D$ grafted at a distance $L$ from the open end of the tube may undergo a rather abrupt transition, where part of the chain escapes from the tube to form a crow
n-like coil outside of the tube. When this problem is studied by Monte Carlo simulation of self-avoiding walks on the simple cubic lattice applying a cylindrical confinement and using the standard pruned-enriched Rosenbluth method (PERM), one obtains spurious results, however: with increasing chain length the transition gets weaker and weaker, due to insufficient sampling of the escaped states, as a detailed analysis shows. In order to solve this problem, a new variant of a biased sequential sampling algorithm with re-sampling is proposed, force-biased PERM: the difficulty of sampling both phases in the region of the first order transition with the correct weights is treated by applying a force at the free end pulling it out of the tube. Different strengths of this force need to be used and reweighting techniques are applied. Using rather long chains (up to N=18000) and wide tubes (up to D=29 lattice spacings), the free energy of the chain, its end-to-end distance, the number of imprisoned monomers can be estimated, as well as the order parameter and its distribution. It is suggested that this new algorithm should be useful for other problems involving state changes of polymers, where the different states belong to rather disjunct valleys in the phase space of the system.
