Biasing soft output values based on a known or learned bit error rate function yields performance improvements in decoding algorithms adapted to work with soft values, such as soft output Viterbi algorithms (SOVA). For example, in a wireless receiver, the soft output values output from a signal demodulator may be biased to reflect the changes in bit error rate across a given burst or block of data. Such changes might arise, for example, due to increasing inaccuracies in the receiver's channel estimate, which is typically computed at the beginning or middle of a block of received data. The wireless receiver may store a table of scaling factors corresponding to the expected bit error rate distribution of the received signal. The table may be preloaded into the receiver, or may be determined during operation. In either case, the table may be updated during operation to reflect bit error incidence observed during operation.