In wireless communications, the block fading (BF) channel is an important channel model. A key quality indicator in coded transmission is the word error rate (WER), which is the fraction of packets that cannot be decoded correctly at the receiver. We study Low-Density Parity-Check (LDPC) coded modulation with precoding, with the aim to minimize the WER on BF channels without channel state information at the transmitter. In the literature, it was not yet known how to optimize the system parameters for this channel model, mainly due to the fading gain distribution. One of the existing approaches to combining coding and modulation is bit-interleaved coded modulation with iterative decoding (BICM-ID). This work uses precoding to optimize BICM-ID with LDPC codes for BF channels, and is a continuation of previous work that used precoding to minimize the outage probability limit. We present the selection of the precoding matrix, the mapping function and the error-correcting code yielding a WER that closely approaches this minimum outage probability. Using a geometric approach, the off-line system optimization effort for the BF channel is limited to at most B+1 times the effort for Gaussian channels, where B is the number of blocks in the BF channel.