Abstract: A new variable order (or tap-length) algorithm is presented and applied to variable order adaptive lattice recursive least square filter. The adjustment of correlative parameters when updating the filter order is also discussed. The new algorithm compares timemean square errors of long and short filters in terms of decibel and employs adaptive taplength step size. It can update filter length and tap-weights fast at the same time before the tapweights converge, and converge to optimal order under noises of different magnitudes. The theory analysis and numerical simulation results verify the efficiency of new algorithm.