Response.xls files are a feature of MediaLab, but they would be a step in the right direction if you were using DirectRT embedded within MediaLab. You can use the response.xls file to calculate the standard deviation, as you want to. However, I feel the sticking point is in then using DirectRT to take that number and apply it to later trials. The MinMax values are defined in the style file, and as far as I can tell, there is no way to import new values into the style file during the experiment.
I think the closest you could do is have the practice round, then summon the experimenter to configure the second round with the data just collected.