Techniques are provided for decoding tail-biting convolutional codes by using information within the received data stream that traditionally has not been used or been available to the convolutional decoder, e.g., cyclic redundancy check (CRC) and bit information known by both the transmitter and receiver. Further, a single parallel trace-back is used that reduces implementation complexity. In addition, the least reliable decisions made during forward processing may be reversed in order to generate additional possible codeword candidates. These techniques can be used to reduce false detection rates (FDRs) and/or detection error rates (DERs).