I'm still experimenting, but I'm getting there gradually. I've rounded the belt, included all the collision options, and increased the calculations per step (thanks svdl); I used a multi-stage approach, exporting the best frame from each simulation and using as the input to the next (thanks diolma), and I cut small gaps in the sides of the belt where it won't be touching the tabard (thanks trav). All these things have made a small improvement, I think it's fair to say without being over-scientific about it. It seems from what everyone is saying, and my own experience so far, that it's going to be a process of incremental improvement (as opposed to excremental improvement, which is something else entirely). It's now down to the point where the cloth is getting 'pinched' between the belt and the figure. I could play endlessly with the offset parameters, or I could just do the rest with a magnet or three. Since my aim was to model something useable rather than explore the depths of the cloth room, it's magnets for me. ;) I can then make the area under the belt constrained (thanks svdl) or decorated (thanks fls13) so that it doesn't interfere when the rest of the tabard is used dynamically.