Yes, this could be done, in several steps, something like this:
1- Collect the scores (whether current scores, and/or complete history of scores, or past x years of scores, etc).
2- Determine relative performance percentiles for each score and/or range of scores.
3- Identify the cut score for the 70th percentile.
4- Graph the scores which are above the cut score.
5- Draw a line to connect the actual graphed data points. The distribution may form a "normal curve" (symmetrical bell curve) or may be somewhat bumpy depending upon the data's sample size, actual scores, and frequency with which each score occurs.
6 - Draw a line to smooth out the curve, showing it to be as close to a "normal curve" (symmetrical bell curve) as possible.
Does this high-level process description answer your question???
