View attachment 7794131
I don't really see a trend that I would bet money on just yet. I agree with his conclusion that .009" seating depth is better than the others in this data set, but with equal graduations being taken per step, simply having one dip down isn't necessarily kicking my "trend" button. I'm not saying it's not a thing, just that the above testing, even when you go a step further and correlate POI/POA for the various 5-shot groups isn't enough to convince me one way or the other. Interesting, though.
Before I get my shit jumped... Here's 100 shots of the same factory match ammo (Hornady 6mm ARC 108 ELDM). It is then broken down into 33x 3 shot groups, 20x 5 shot groups, 10x 10 shot groups, 5x 20 shot groups, and 3x 33 shot groups, with the same sequence of firing. The dots represent the average group size, and the 'wings' represent the total span of recorded group size. THIS IS WITH NO VARIABLES CHANGED-- The SAME ammo. Accuracy fixture, straight 1.25" no contour barrel, 200yd climate controlled indoor range. This also an excellent indicator to the level of trash "group size" is as a metric, but I digress... With this level of noise present in a "no" variable string, it makes a guy question what you're reading when you do change variables.
ETA: Important to note here that the 20-shot and 33-shot data is in itself small sample size data (only 3x or 5x of them), and would likely also grow a little with more testing-- however, with such large samples per test, the amount it would grow would be significantly less than 3-10 shot data sets.
ETA2.5: Okay, screwed myself with an F4 button in Excel, here's the corrected one.
View attachment 7794214
Here is the same data analyzed with mean radius and SD on individual shot radii from the MPOI. This is 2*(mean rad + 2* SD) to generate an estimate of group size. I can explain this if you'd like but all of the data I've collected has shown 4-4.5x SD + 2*MR to be pretty close to inclusive of 50+ shot group size (diminishing returns on group size growth past 50 shots)... The resulting value is for "worst case" predictions on hit probability.
View attachment 7794145
Note how much more even and expected the trend is of the averages (using more of the data from each shot, not just the 'worst' 2). Also note the wild variation that comes from trying to predict results with 3 and 5 shot groups. Wish the trend would die.
Another thing people like to try, is to average a boat load of small sample tests and say "Surely, this is as good as a single large sample test"... And you can see that the distribution obviously favors smaller group size with smaller sample size. Without a POA reference to tie multiple small sample tests together, you're operating with less data, even if the round counts are the same. Similar trends exist with ES/SD on MV.
The more you learn...