We consider tick by tick hedging strategies for derivatives in a micromovement model of asset prices. The model assumes a latent intrinsic value process, which can be observed by agents only at trading times with market microstructure noises. The intrinsic value process, assumed to be a diffusion, is the usual price process in option pricing. Furthermore, this model incorporates the two stylized features of the ultra-high frequency data: random trading times and market microstructure noises. Because of these, the market becomes incomplete. We develop a new evolution representation of such a model in a form of marked point process. Then we derive the projected local risk minimization strategy with the observed prices alone. We also show that the computation of the local risk minimization strategy in our context leads to a smoothing and filtering problem in the nonlinear filtering literature.