A few statistical problems faced by the event reconstruction in ALICE experiment at CERN are discussed in this paper. We outline several ad-hoc extensions of traditional Kalman-filter track finding which seem to increase the quality of tracks reconstructed in high multiplicity events anticipated for Pb-Pb collisions at LHC. These extensions, however, need a stricter formulation and justification from the theoretical side. The particle identification in ALICE is done by combining the information

more »

... from different detecting systems using a Bayesian method. Having many clear advantages, this approach introduces into the analysis additional complications which are also discussed here.