Go with MV or SD?

[nCognitos]

I did some load testing with my 6.5CM using N150 and Lapua 139 Scenars.

From the data below, should I go with:

(i) the velocity node at 37.8?

(ii) the ES/SD node at 37.4?

(iii) stop overthinking it, the SD's are so close it makes no difference, just load for the MV I want?

GRT gives nodes at 37gn and 38gn and tells me the max load (38.4gn) is 6% below max pressure.  38gn would be 9% below PMax and 37.4 would be 15% below PMax.

Charge Weight	Average	High	Low	ES	SD
37.2	2682	2690	2675	15	6
37.4	2690	2694	2684	11	5
37.6	2712	2719	2705	14	6
37.8	2716	2726	2711	15	6
38	2725	2732	2718	13	5
38.2	2739	2750	2732	17	7
38.4	2751	2760	2744	16	6

[No i deer]

Accuracy is the most important..

Go with the lowest ES/SD and adjust the seating depth to find the sweet spot..

Go 3 thousandths increments not to miss it.

You can still get good accuracy from a 30fps ES..

The target will tell you what you need too know 👍

[nCognitos]

Thanks, my other problem is there was no appreciable change in group size across the entire range but that was on an Electronic target at 300yds.  Will go again with 37.4 to 38 on paper and see what that tells me.  

[Popsbengo]

30 minutes ago, nCognitos said:

Thanks, my other problem is there was no appreciable change in group size across the entire range but that was on an Electronic target at 300yds.  Will go again with 37.4 to 38 on paper and see what that tells me.  

How many rounds did you test at each increment of load ?

 

[SMLE]

As a point of reference, my 6.5 Creedmoor, shooting the same bullet at pretty much your velocity and SD will be at moa or just under all day long at 1000 yards.

Quick edit..... at your 37.4 load, mine achieved with RS62 though.

Edited June 10, 2022 by SMLE
Missed data

[nCognitos]

2 hours ago, Popsbengo said:

How many rounds did you test at each increment of load ?

 

5 rounds

20 step ladder prepared of 100 rounds, started in the middle with a view to work up or work down depending on MV and pressure signs, all looked good so worked up and stopped at 38.4, I had loads up to 39 but I thought I could see very feint pressure signs around the primer so I stopped at 38.4

Once I have it narrowed down, I will load up 10 at each increment for MV and grouping but I don't really want more than 5 increments for this.

I really am a bit bemused by this, my mate who was with me said load what you like, the gun doesn't care!

I did some .303 at the same time and got the results below, which are much more what I expected, the answer is obvious, go with 41gn.

Charge Weight	Average	High	Low	ES	SD
40.4	2370	2383	2349	34	16
40.6	2382	2421	2354	67	25
40.8	2414	2434	2400	34	12
41	2424	2437	2416	21	8
41.2	2435	2462	2418	44	17
41.4	2445	2467	2428	39	17
41.6	2459	2470	2444	26	11
41.8	2472	2486	2457	29	14

[Popsbengo]

Problem is, if you repeated the tests you'll get different results I'll bet a pound to a pinch of snuff.  Increments of 0.2 gr but ± what ?

Your results already overlap on ES and did you clean the barrel between sets? I'm not suggesting you should just that you're at risk of reading too much into the results - like so many others do.  Do you control for primer variation?   Of course not, it's not practical.  Did you ensure precise neck tensions and seating depth variation?  Possibly but if not then is a 0.2gr increment going to be visible in the 'noise' of other variables.

I'm not having a dig at you, I'm just amazed at how much time and barrel life is wasted by so many.  This whole tiny increment ladder testing with a few rounds is statistically challenged to say the least.  There's an awful lot of spherical gonads around load development IMHO.

I move in half grain increments (typical loads for .308/6.5CM/.338) to find something that groups best without pressure signs, try another batch and tweak seating depth - done.  No graphs, no bullshine misuse of stats.

Once I have a good load I do the utmost to ensure those variables are controlled batch after batch.  My results are stable over normal UK temps, repeatable, precise and accurate.  

[SMLE]

4 hours ago, Popsbengo said:

This whole tiny increment ladder testing with a few rounds is statistically challenged to say the least.  There's an awful lot of spherical gonads around load development IMHO.

I move in half grain increments (typical loads for .308/6.5CM/.338) to find something that groups best without pressure signs, try another batch and tweak seating depth - done.  No graphs, no bullshine misuse of stats.

Once I have a good load I do the utmost to ensure those variables are controlled batch after batch.  My results are stable over normal UK temps, repeatable, precise and accurate.  

Spot on Pops, spot on!

[nCognitos]

Dirty Barrel, I do 5 fouling shots with factory ammo before I start.

I disagree with you, the MV variation across one grain of load are equivalent to about 0.6MOA, which is the difference between a V and a 5.

To be honest, I'm not asking to be trolled about my methodology, simply asking for help in its interpretation as the results are so different to what I expect, as demonstrated by the .303 results.  

[Popsbengo]

25 minutes ago, nCognitos said:

Dirty Barrel, I do 5 fouling shots with factory ammo before I start.

I disagree with you, the MV variation across one grain of load are equivalent to about 0.6MOA, which is the difference between a V and a 5.

To be honest, I'm not asking to be trolled about my methodology, simply asking for help in its interpretation as the results are so different to what I expect, as demonstrated by the .303 results.  

Sorry, not trolling.  No offence intended, giving an honest appraisal of method is the way to explain confusing results.

1gr may give an observable 0.6moa but it does not follow that 0.1gr is 0.06 moa (ie tenths) as there's a collection of other variables.

Your results overlap in ES and with only five values per set it's pointless using standard deviation calculation.

What were your groups like?  That's a key criteria,  looking for tight groups that indicate a barrel harmonic node.

[martin_b]

13 hours ago, Popsbengo said:

'm not having a dig at you, I'm just amazed at how much time and barrel life is wasted by so many.  This whole tiny increment ladder testing with a few rounds is statistically challenged to say the least.  There's an awful lot of spherical gonads around load development IMHO.

Couldnt agree more!   With my latest rifle I ran quick load and got a nice mid range load for running the barrel in, but a few rounds through it an then took it out to 600yards seems to group well, then 1000y still good groups, so Im just going to stick with whats Ive got, 

[Andrew]

19 hours ago, Popsbengo said:

How many rounds did you test at each increment of load ?

 

The most important question...... ~Andrew

[nCognitos]

I really am not interested in a discussion on methodology, but if anyone has an opinion on the original question, I would be grateful.  I must say it's much more difficult with this rifle and calibre, everything else I have gives a clear answer straight away.  

[jcampbellsmith]

If you think that better results are possible, try changing primers or powder or bullet. 

Regards 

JCS 

[Chris-NZ]

8 hours ago, nCognitos said:

I really am not interested in a discussion on methodology, but if anyone has an opinion on the original question, I would be grateful.  ..

Ok, my "analysis" is your data is so limited is that you are unable to draw any valid conclusions. It's pointless talking SD with such minute samples. You'd warm up with an ES assessment with 10 shots and more realistically 20 if you're serious. As for pressure, only a strain gauge will tell you the actual pressure. Modelling will give you an idea of ballpark. I'd rely more on signs like primer flattening and difficult bolt lift ultimately.

I'd suggest you focus more on groups than "statistics"

[nCognitos]

So, there seems to be a real absence of understanding of the way statistics work.

There is no minimum number of samples for a meaningful results, what counts is confidence level generated by the margin of error, you keep taking data until you achieve the confidence level that you desire, and you can do this with 4, 5 or 6 samples, as long as they are representative.  More samples will increase your confidence but it's not linear and quickly becomes the subject of diminishing returns.

So for my data, taken at 37.2gn and running it through a stats calculator I have an average of 2,682ft/s.

From that I get a confidence of 68% that the results for the population will be 2,682 ± 2.49, and a 99% confidence that the results will be 2,682 ± 6.41.

I'm afraid that that is just the way maths works.  

Anyway, I got an answer to my question elsewhere so I will leave it at that.

[Popsbengo]

Just now, nCognitos said:

So, there seems to be a real absence of understanding of the way statistics work.

There is no minimum number of samples for a meaningful results, what counts is confidence level generated by the margin of error, you keep taking data until you achieve the confidence level that you desire, and you can do this with 4, 5 or 6 samples, as long as they are representative.  More samples will increase your confidence but it's not linear and quickly becomes the subject of diminishing returns.

So for my data, taken at 37.2gn and running it through a stats calculator I have an average of 2,682ft/s.

From that I get a confidence of 68% that the results for the population will be 2,682 ± 2.49, and a 99% confidence that the results will be 2,682 ± 6.41.

I'm afraid that that is just the way maths works.  

Anyway, I got an answer to my question elsewhere so I will leave it at that.

Really, ok that's great, chuffed you got sorted, maybe you can become our go-to expert on stats ?

[TC]

5 hours ago, nCognitos said:

So, there seems to be a real absence of understanding of the way statistics work.

There is no minimum number of samples for a meaningful results, what counts is confidence level generated by the margin of error, you keep taking data until you achieve the confidence level that you desire, and you can do this with 4, 5 or 6 samples, as long as they are representative.  More samples will increase your confidence but it's not linear and quickly becomes the subject of diminishing returns.

So for my data, taken at 37.2gn and running it through a stats calculator I have an average of 2,682ft/s.

From that I get a confidence of 68% that the results for the population will be 2,682 ± 2.49, and a 99% confidence that the results will be 2,682 ± 6.41.

I'm afraid that that is just the way maths works.  

Anyway, I got an answer to my question elsewhere so I will leave it at that.

You seem to have Googled Standard Deviation, now it’s  time you move on to Google Standard Error  to understand the impact of sample size on the relevance of your SD …….

Glad you are happy though 👍😉

[Popsbengo]

6 minutes ago, TC said:

You seem to have Googled Standard Deviation, now it’s  time you move on to Google Standard Error  to understand the impact of sample size on the relevance of your SD …….

Glad you are happy though 👍😉

I think there should be a test ?

😂

[Chris-NZ]

Yeh, I'm not qualified to comment as I only tutored stats at uni. Forgotten most now but still remember enough to know 5 data points is a token effort at best. In the end, ES is the telling thing with target work, ie, your worst outliers which cost the points.

[VarmLR]

There IS statistical relevance in sample size when it comes to load development.  It's been done to death many times but the long and short of it is that anyone going out shooting 2 or 3 round "groups" won't get anything very meaningful from that exercise which can accurately and consistently be relied upon further out.  5 to 7 rounds, statistically, is about the best ratio of economy to relevancy and anything much over 7 is probably not that worthwhile in economy terms.  Some comp target shooters will shoot 7 to 10.

I always use 5 shot groups and use OCW methods rather than stepped ladder testing, and usually find one or two nodes for each powder/projectile combo, then tweak the seating to improve precision/lower group size.  Beyond that and achieving a low enough SD/ES for the range being developed for, there's little more to be done, other than be consistent batch to batch with case prep.  You can fiddle about with primers to improve things further but if what you have is achieving the right ES/MV for you then stick with what you use.

If sticking with Vhit powders, my recommendation would be N150 over N140.  RS62 is what I use for this combo though and have never varied my loads over anything outside of 43.8 to 44gr for the 139 Scenar (small primer/flash hole brass) for very reliable, accurate results.  If you want to interpret graphs, one useful thing from them might be to see where your load is flattening out, ie where a 0.2gr load either side has very little difference (sub 10fps) on MV.  From this, I'd be playing around with the middle of that plateau and then tweaking seating until I had a consistent and accurate load. I look for very small vertical deviation on target for a decent harmonic node.  Just my tuppence worth.  Others will have their preferred methods but experience and lots of load dev has settled in my mind anyway the most reliable way to go about it for me anyway.

[nCognitos]

On 6/20/2022 at 7:27 PM, TC said:

You seem to have Googled Standard Deviation, now it’s  time you move on to Google Standard Error  to understand the impact of sample size on the relevance of your SD …….

Glad you are happy though 👍😉

Well, I could but the SE is 2.803 which if you work it out tells you very little more than the SD, so going for a 95% confidence level, thats 2,716±5.6 fps.

 

On 6/21/2022 at 3:43 PM, VarmLR said:

If sticking with Vhit powders, my recommendation would be N150 over N140.  RS62 is what I use for this combo though and have never varied my loads over anything outside of 43.8 to 44gr for the 139 Scenar (small primer/flash hole brass) for very reliable, accurate results.  If you want to interpret graphs, one useful thing from them might be to see where your load is flattening out, ie where a 0.2gr load either side has very little difference (sub 10fps) on MV.  From this, I'd be playing around with the middle of that plateau and then tweaking seating until I had a consistent and accurate load. I look for very small vertical deviation on target for a decent harmonic node.  Just my tuppence worth.  Others will have their preferred methods but experience and lots of load dev has settled in my mind anyway the most reliable way to go about it for me anyway.

Thanks for that, I'm using N150, I'm going to test down the other way with smaller charges tomorrow, and see if I can get that flat spot or obvious best point.  Then I can focus on seating depth.  I've done load testing before but I've never had a situation where the SD's and ES were so similar across a grain variation in powder, with very similar group sizes, it struck me as odd but perhaps my friend is right, the barrel just doesn't care.  I'm going to shoot max 1,200yds so I have some leeway to bring the charge down and still have the speed.  My daughter was shooting at 1,000yds on Sunday and getting excellent results with slightly lighter loads that I put together for her to plink with, so I'd like to explore that space a little more.

[VarmLR]

Some barrels shoot some bullets better than others, granted, but all barrels are subject to laws of physics and most certainly harmonic nodes  determine optimum loads.  For each powder and bullet type I've generally found at least two nodes, sometimes three between min and max loads.  I generally chose the lowest MV which satisfied the required terminal velocity.

Group sizes tell you little at 100 or even 200 yards.  ES or SD are a far more reliable indicator for longer range where variations in MV only widen the terminal velocity differences, hence vertical and horizontal spread.  I'd always take a 0.7moa load with sub double figure SD/ES any day of the week than a sub 0.5 5 shot group at 100 yds with higher SD/ES.

 

You can't always rely on "flat spots" or graphs unless case prep is consistent and absolutely impeccably done for precision, and where loads are literally to within 0.1 or 0.2 grains.  That's been over popularised in You Tube videos where the shooters may have gone to extraordinary lengths for consistency in case prep and tried various primers for the best SD and consistency.  

At 1200 yds, you're on a hiding to nothing unless you can get SD and ES down to within double figures...But rule of thumb, at that range, calibre and bullet dependant, using the higher MV loads will always yield less variation given similar SD results and groups as lower MV nodes.

[Popsbengo]

8 hours ago, VarmLR said:

Some barrels shoot some bullets better than others, granted, but all barrels are subject to laws of physics and most certainly harmonic nodes  determine optimum loads.  For each powder and bullet type I've generally found at least two nodes, sometimes three between min and max loads.  I generally chose the lowest MV which satisfied the required terminal velocity.

Group sizes tell you little at 100 or even 200 yards.  ES or SD are a far more reliable indicator for longer range where variations in MV only widen the terminal velocity differences, hence vertical and horizontal spread.  I'd always take a 0.7moa load with sub double figure SD/ES any day of the week than a sub 0.5 5 shot group at 100 yds with higher SD/ES.

 

You can't always rely on "flat spots" or graphs unless case prep is consistent and absolutely impeccably done for precision, and where loads are literally to within 0.1 or 0.2 grains.  That's been over popularised in You Tube videos where the shooters may have gone to extraordinary lengths for consistency in case prep and tried various primers for the best SD and consistency.  

At 1200 yds, you're on a hiding to nothing unless you can get SD and ES down to within double figures...But rule of thumb, at that range, calibre and bullet dependant, using the higher MV loads will always yield less variation given similar SD results and groups as lower MV nodes.

You say "group size tells you little at 100 or even 200.."  Then how do you know you've found an harmonic  node ?  I use group size as my main criteria and so far I get excellent results at 1000yds for 308/6.5cm and I can shoot sub 1moa at a mile with .338LM provided the gods of wind agree.  For sure Mv consistency is very important but I'm not sure it's true to say groups tell you little.  A group is identifying the trajectory of the bullets, they can't miraculously come together at a diverging range except for the slim possibility of elevation due to velocity (Lee Enfield affect) - if the group spread left/right that's only going to get worse as they diverge further.

 

[VarmLR]

I meant that grouping alone at 100m won't be much good to you if you have an SD of 30, at 1000 yds or whatever.  By that distance, the spread will have increased beyond proportionally.  I think you knew this already though.  Grouping is an indicator of a harmonic node, granted but MV consistency from the load data is as important at range. OCW is a good way to see the POI variations "around the clock" so you can concentrate on those with low SD/ES about a similar vertical plane, with low vertical spread.