I made a statement in an earlier post about learning not to trust 5 shot groups. In case anyone thought I was just whinging, I decided to do a little investigation, and came up with some results which may surprise some. I won’t bore you with the exact method except to say that it was done with a simulation using statistical functions in MS Excel. Anyone who is interested, please PM me and I’ll send you a copy of the spreadsheet.
Here is what I found. Suppose you have 4 loads, of which the accuracy is known exactly in terms of their SD (standard deviation) in MOA from the centre of the group. Let’s say they are 0.2, 0.25, 0.31 and 0.39 respectively. Note that after the first one, each subsequent one increases by a factor of about 1.25, and the worst one (0.39) is about double the best (0.2). This means in effect that if you were able to fire hundreds of shots with these four loads, average group sizes for the best load would be about half those of the worst, with the other two loads somewhere in between.
Now here’s the catch :
if you fire a 5 shot group with each of these loads, there is less than a 50% chance that the best load will produce the smallest group ![Shocked :shock:](./images/smilies/icon_eek.gif)
. And from another perspective, there is about a 25% chance that the best group will come from one of the two worst loads
![Rolling Eyes :roll:](./images/smilies/icon_rolleyes.gif)
!
One of my own testing methods has been to shoot 4 groups of 5 shots, each with something (e.g. powder charge) slightly different. Based on these simulation results, the assumption that the smallest group will indicate the most accurate load is at best very flakey.
Okay, then what should you do? I'm not sure actually - extending the simulation to 10 shots, there's still only about a 60% chance the best load will produce the smallest group. All I can say with any certainty is that the more shots you fire, the better idea you'll have of the accuracy.
Alternatively, you could just keep buying barrels until you get a hummer
![Cool 8)](./images/smilies/icon_cool.gif)
.
Alan