The best part about 100 shot strings is that whole n vs. n-1 makes a near-negligible difference...
All joking aside, everything I've done has been stdev.s, under the assumption (backed by a few thousands of rounds of documented P&V and dedicated barrel life testing) that through a certain portion of barrel life the results are close enough to "the same" to count that portion as the "population" of which my specimens are a "sample" subset to be used as a predictive tool for the population. As far as which the chronograph shits out, SD is almost pointless to look at below 15-16 rounds and to be taken with a grain of salt at that point anyway... When you start looking at sample sizes large enough to make SD worth looking at, it's not a very big delineation regardless.
For example, on my PRS load for 20 shots STDEV.S is 8.1fps and STDEV.P is 7.9fps. Big deal. On a 100 shot test of factory 108 6mm ARC ammo STDEV.S is 12.69fps and STDEV.P is 12.63fps.
Filter of realistic use-case for most people: Use whichever gives you a smaller number so you feel better about what you've done.