Standard deviation (usually abbreviated as the lower case “s” or Greek sigma, or sometimes sd or SD) is one of a number of different but somewhat related measures of the variability, or dispersion, in a set of measurements, or in the population from which the sample set is drawn. Others include the variance (which is the square of s), the range, the coefficient of variation, and the mean or average deviation. Each of these can be useful, but s has the advantage that, in a normally distributed population (and most “velocity” – technically, because it’s a scalar not a vector, “speed” – measurements associated with reloaded ammunition that I’ve tested are, in fact, normally distributed or acceptably close to being so), the mean plus or minus s will include about two-thirds of the population, +/- 2s will include 95%, and +/- 3s will include 99%.
I think most reloaders would agree that, all else being equal, consistent velocity is a desirable characteristic of reloads – it’s hard to imagine a situation where an increase in variability would be a desirable goal. So, to that extent (but see below), your friend is most definitely incorrect. What there’s much less agreement about is how small s should be for a particular load to be considered “good.” And that leads us into one of the deficiencies of s as a measure of variability, namely that it will vary (i.e. is correlated) with the mean velocity. Consider, for example, a handgun load that has an average velocity of perhaps several hundred fps vs. a rifle load humming along as something closer to 4,000 fps. A value of s of, say, 50 means something very different for each of those. The coefficient of variation (s divided by the mean then converted to a percent by multiplying by 100) takes that into account, in effect “adjusting” s by the value of the mean. I rarely see much discussion of CV in reloading circles, but I think of it as more valuable than s, although again there will be differing opinions on what’s a “good” CV.
All of that said, the most important consideration is how well the loads perform in your firearm with you at the controls. The results – and there are much better measures than group size, but that’s a whole ‘nother discussion - are incorporating the net effect of all sources of variability, not just the variability in velocity. If your groups are acceptably small then it really doesn’t matter much what s is. So, in that particular case, I’d have to say your friend has a point. However, if your groups are larger than you like, knowing the associated value for s can help to isolate the reason.
Sorry for the long-winded reply. I used to work with this stuff at one time, but don’t have much opportunity to discuss it now in retirement.