The Firing Line Forums - View Single Post

Rimfire5 · November 30, 2012, 08:42 PM

Even a blind pig finds an acorn eventually and even a marginal load occasionally produces a good group, probably because the shooter and all the other variables happen to align. A single serendipitous event doesn't prove the pig can see or that a load is a good load.

As you know, lots of things can mess up a group:
a) neck sizing variations might not result in a repeatable neck pressure and cause a velocity difference
b) powder drops might not be exact so velocity can be different round to round.
Velocity will vary anyway as part of the normal process and the point of impact will change. But if the variation is cause by the loading process, then the process is not testing the same load in every cartridge, even though that was the intent.
c) seating depth might not be the same from cartridge to cartridge and might cause some change in the point of impact.
d) the shooter might not get behind the rifle the same way for every shot and actually not be aiming at the exact same aim point.
e) the barrel might get hot during the process of shooting the group and cause a change in the point of impact for one or more rounds.
f) the shooter can jerk the trigger or manhandle the stock, changing the point of impact for one or more rounds.
g) wind gusts can cause drift and open up a group
AND there are lots of other possible causes of variation in the process of shooting groups.

In any statistical process control effort, the first thing that is done is to identify any measurements that have known causes for variation and eliminate them from the analysis. They represent causes of variation that are not elemental to the process and don't represent what can be controlled.
If you know you messed up one of the shots in a group, or something went awry, it is best call it and don't count the group.
Including a group that has a known error only messes up your ability to make any conclusion about the load.

I measure at least 4 groups when I am investigating loads and take an average.
The ones that look very promising, I duplicate and shoot at least 10 groups.
There is always some range of group sizes. The loads that groups sizes that are similar are very promising. Those are the ones that I will examine in more detail. The others I record but don't necessarily do much with them.

When I shoot more groups, I look at the median and standard deviation of the groups for each load.
The loads with the best averages and medians and with low standard deviations will produce the best groups over the long term.
For the really good loads, I gather data on at least 20 groups so there is statistical significance. I then calculate the range that 90% of the groups will fall into so I know what the load will proably produce based upon those two measurements. My 10 best loads fpr a rifle may have upwards of 30 groups and if those loads retain a small standard deviation over a large sample, I can be confident that those loads will produce similar results time and again.

Some people seem to believe that a very small, single group indicates a great load.
All one group indicates is that everything went just right for that instance, or more likely, enough compensating things went wrong in just the right combination that the bullets happened to hit in the same spot.
If the next set of groups doesn't come close to matching that great group, then either the shooter can't repeat the techhnique, there is some anomaly like the wind that is causing deviations, or the load is not repeatable.

November 30, 2012, 08:42 PM	#8
Rimfire5 Senior Member Join Date: January 2, 2009 Location: Northern Virginia Posts: 923	Even a blind pig finds an acorn eventually and even a marginal load occasionally produces a good group, probably because the shooter and all the other variables happen to align. A single serendipitous event doesn't prove the pig can see or that a load is a good load. As you know, lots of things can mess up a group: a) neck sizing variations might not result in a repeatable neck pressure and cause a velocity difference b) powder drops might not be exact so velocity can be different round to round. Velocity will vary anyway as part of the normal process and the point of impact will change. But if the variation is cause by the loading process, then the process is not testing the same load in every cartridge, even though that was the intent. c) seating depth might not be the same from cartridge to cartridge and might cause some change in the point of impact. d) the shooter might not get behind the rifle the same way for every shot and actually not be aiming at the exact same aim point. e) the barrel might get hot during the process of shooting the group and cause a change in the point of impact for one or more rounds. f) the shooter can jerk the trigger or manhandle the stock, changing the point of impact for one or more rounds. g) wind gusts can cause drift and open up a group AND there are lots of other possible causes of variation in the process of shooting groups. In any statistical process control effort, the first thing that is done is to identify any measurements that have known causes for variation and eliminate them from the analysis. They represent causes of variation that are not elemental to the process and don't represent what can be controlled. If you know you messed up one of the shots in a group, or something went awry, it is best call it and don't count the group. Including a group that has a known error only messes up your ability to make any conclusion about the load. I measure at least 4 groups when I am investigating loads and take an average. The ones that look very promising, I duplicate and shoot at least 10 groups. There is always some range of group sizes. The loads that groups sizes that are similar are very promising. Those are the ones that I will examine in more detail. The others I record but don't necessarily do much with them. When I shoot more groups, I look at the median and standard deviation of the groups for each load. The loads with the best averages and medians and with low standard deviations will produce the best groups over the long term. For the really good loads, I gather data on at least 20 groups so there is statistical significance. I then calculate the range that 90% of the groups will fall into so I know what the load will proably produce based upon those two measurements. My 10 best loads fpr a rifle may have upwards of 30 groups and if those loads retain a small standard deviation over a large sample, I can be confident that those loads will produce similar results time and again. Some people seem to believe that a very small, single group indicates a great load. All one group indicates is that everything went just right for that instance, or more likely, enough compensating things went wrong in just the right combination that the bullets happened to hit in the same spot. If the next set of groups doesn't come close to matching that great group, then either the shooter can't repeat the techhnique, there is some anomaly like the wind that is causing deviations, or the load is not repeatable.