3 shot group vs ladder tests

[209jones]

I do 5 loads per powder weight, lower half to max, depending on cartridge size, .3, .5 on the lower part, .3 on the upper area of the chg wt, from ballpark, say 2gr below max for a med sized cartridge (say 308), 1gr on the lower, .5 on the upper with a magnum size. Being as the bullet was already picked for the app, and that is THE bullet, doing it this way gives me a few things, how the powder pressure increase behaves, whether or not it looks like it wants to shoot with that bullet/powder combo, and a rough idea of what speed range that will get me to for that bullet.
If I like it what I see, I'll test what seems to be up or down from a point that looks promising, then perhaps another with what looks like a good spot, but, change seating depth a bit to see what that does, or not. If I don't like what I see on the first test, I'll try another powder. I usually get the first one right, often try a csecond one to see if or how much difference. I have had 3 powders not work, in which case I'll then swap bullets, unless I am wedded to using that bullet for that application, then the gun goes.
As far as I'm concerned, Satterlee method is garbage, ES can be overrrated unless you shoot at 500yds plus a lot. The vid Little Crow Gunworks did last week, illustrates what I am talking about explains it very well, there is a series there, watch the first one, and the last one., if they turn your crank, watch the rest of them. I do agree with about 95% of what he does there, and he explains it more coherently and better than me. I just don't use GRT, is really the only difference, and I don't chrono much after the initial test, til I get to the end, then I'll do it again. The reason I use 5 rds on the initial set is 2 are for pressure test., other three are to see what the load behaves like.

 

[ELR LVR]

Good Afternoon,
I grew up developing loads using the 3 shot group method.
Lately I have been noticing a lot of guys using the ladder method.
I have a chronograph and thought I would give it a go.
So what you are looking for is a place where the velocity does not change with charge weight, correct? So how much do you increment your powder charge for each step on the ladder typically? So from that I could assume that POA does not matter all that much?
Thanks

This is just me but many guys do it different
I like to arrive and center up on a good load as QUICKLY as possible.
Sometimes I use the 1 shot instead of the 3 shot method.
You can keep firing at the same aiming point and watch the shots spread out or center up
Or run a waterline test on different aim points.
First thing I do is find seating depth
To accomplish this - I use a known good powder for the cartridge, a powder which has typically worked well in other same calibers.
For instance W-748 ALWAYS...............Every time....works well in .223
Next I will increment starting low, 0.5 grns at a time until I find Max, and see pressure signs.
I want to know what powder charge to stay behind to not experience high pressures when it's hot.
Between Low and Max powder charge somewhere in there will show good accuracy.
This may only take 20 rounds.
THEN
I hone in on the optimum charge by honing in the load in .2 -.3 grn increments using the 3 shot method
And look for
1 Desired Velocity
2 Low SD numbers
3 Consistency in accuracy from day to day or week to week
-----------------------
If I don't get all 3 , then I will Change Powder first
Then , Change Primer,
Then Change Bullet
All brass must be uniform in weight and neck thickness
-------------------------
When I've found Seating Depth
And Powder Charge
I will then move on to testing at least 3 different primers with that charge weight.
You may NOT be getting good Low SD numbers until you for instance change away from Winchester to maybe Remington Primers, or from CCI to Federal ETC.
-------------------------
You have no idea which powder will work best so have an assortment of 3-5 different powders in the right burn rate,
Like all sample tests, the more you have to choose from the better and more complete the results will be.
So even trying 10 different powders will give you better statistical testing results
Some may not like this fact but we cannot use our feelings or emotions to dictate statistics.
This is science (rocket science to be more precise)
I also like to know which powder WILL NOT work well so I know to avoid them
I also like to know how many different powders WILL WORK, since our fav powder may not always be available at any given moment, and it's good to have a back up second choice.
-----------------------
Don't set a timeline for yourself, it may take time(weeks) to arrive at your optimum load but much can be accomplished in just one week with 5 trips to the range
-----------------------
I do NOT take the Emotion driven and personal opinionated attitude that "THIS GUN" is going to shoot with "THESE" bullets and "THIS" powder just because I bought them special like a christmas present just for that gun.
It's an inanimate object and doesn't care that you spent "Good Money" on it.
That is how people set themselves up for failure and disappointment.
I have gotten every gun I have ever touched to shoot very accurately using my methods,
including factory guns.
However factory guns may tend to walk as things heat up so we must be realistic with our expectations

 

[Alex Wheeler]

First I want to say that your need to have a rifle that actually shoots. If you have noise in the system from a poorly built rifle, components like bullets or a barrel that holds it back, poor shooting technique, poor tuning ability, or any other thing that holds the rifle back you may come to odd conclusions. Like tuners or annealing doesnt effect group size or is not repeatable, you may think seating depth or powder charge dont effect groups. You may think there are no nodes.

Assuming you have a rifle that shoots, ladders have a ton of value. They are not just for powder charge. They are for anything you want to test in a load. They are awesome for picking components like powder or primers. They are a very fast way to pick a neck bushing as well. I dont use them to pick an exact pwoder charge or seating depth. I use them to find the area. Then I follow up with 3 shot groups in the area but still in a ladder format.

Ladders have a lot of value, the vast majority of the LR BR records I am familiar with utilized that format of tuning. It may not be good enough for Bill Nye the science guy, but its good enough to break records. Shooting is about proving it. Ask the science guys to start showing up to matches and proving these ideas of what works and what doesnt.

 

[LVLAaron]

I love it when things are thoroughly debunked, but are repeatable.

 

[urbanrifleman]

Flat spots exist.

 

[urbanrifleman]

I love it when things are thoroughly debunked, but are repeatable.


View attachment 1673737

Don't fight the consensous...

It was in a book. It has to be right.

I know for a fact one of the great powders I use, (the one which we shall not speak it's name) absolutely shows flat spots. Sometimes 2/3 of a grain wide.

 

[LVLAaron]

Because "ladders dont work" - I decided to shoot groups with my 300 WSM - A totally new cartridge to me. 66.4 gave me the smallest and very respectable group at 500 yards.

I just scribbled the avg velocity of each group in my book for that barrel. For the sake of ruffling feathers I just graphed it.



All roads lead to the flat spot.

 

[Varminterror]

I love it when things are thoroughly debunked, but are repeatable.


View attachment 1673737

What is your 30 shot ES for this load? Apply appropriate error bars to this data and run a linear regression against the average.

You’re drinking the same koolaid most of us drank for years.

ETA: Taking liberty and about 75 seconds of photo editing time on my phone…

Knowing a 3 shot series offers very low confidence for accuracy of the average, but working with the data on the page, the trend pictured above offers ~12-15ES throughout the trend, which, for small samples, suggests the SD for the population is likely 6.5-7fps ballpark (not the 3 shot sample SD, the predictable population SD), which corresponds to a population ES around 30-35 for larger strings. So dropping a rough average line and +/-15fps boundary lines, straight lines, essentially all shots fall within the linear prediction when we include inherent error possible for any shot. If the hypothesis that flat spots exist doesn’t even survive the most basic analysis, but the hypothesis that flat spots DON’T exist does persist, eh, I don’t really like lying to myself about what data tells me.

 

[Varminterror]

Don't fight the consensous...

It was in a book. It has to be right.

I know for a fact one of the great powders I use, (the one which we shall not speak it's name) absolutely shows flat spots. Sometimes 2/3 of a grain wide.

Publish the data and refute the multiple publications which have proven the flat spots occur simply due to statistical anomalies.

The first trend pictured above is a great example of WANTING to find patterns in data where none exists.

The same in the second - the guy mentions the starting load 66.4grn offered the best group, then points to an inflection deviation in a data set which traced a nearly perfect straight line, and pretends a node must exist around 68.5grn.

It’s really simple - anyone can publish data, but not everyone making claims is willing to conduct defensible datasets to actually support the claim.

I had the same knee jerk reaction when Litz published, followed shortly after by Hornady’s podcast - so I went back to my own data, some of which had 50-100 data points per charge and a few dozen Satterlee curves, I was dead set on emailing Miles and Bryan a nasty email which proved out my velocity nodes repeated. Until I compiled the data and ran simple analysis of error and realized all of those nodes I had believed existed for years were simply anomalies within the standard error of the population, and a straight line behavior emerged. I was convinced I could publish data which refuted theirs, just using old data I had on hand from what I used to call “node confirmation tests” I would shoot before every match. But doing the actual evaluation of the data - which is inherently why we capture data - proved the observer bias to conclude nodes were repeating was simply false. The null set - the idea velocity nodes do not occur in a Satterlee test (which is really an Audette test Creighton published around the time Scott was born) might be less interesting, but it’s true nonetheless.

 

[LVLAaron]

I think my SD's have always been pretty good?

 

[LVLAaron]

I'm not trying to change anyone's mind here, but I am the only one sharing data that is my own. Yes, the first graphs are biased because I was looking for flat spots; hell, sometimes I'd just fire into dirt and gather velocity data and go from there.

To remove that bias, I shot the 300WSM at a target and didn't even care to graph the data, but when I did... my chosen load fell on a flat spot. Is that "statistically relevant" to a statistician who doesn't shoot guns? No probably not. Is it a method that works for me? Yes. I don't care how or why it works. But it sure is weird that it works over, and over, and over again.

 

[Varminterror]

I think my SD's have always been pretty good?

An SD of 14.9 across 20 rounds or 13.7 across 20 is roughly 3x looser than the SD I used to analyze your data trend above. The yellow lines would be 3x wider, with ALL shots within those boundaries being valid predictions (proportionately distributed, of course) for actual velocity opportunities. In other words, I applied a linear regression with a tighter standard, about 3x tighter than your actual loads, and the linear regression fit your data. Applying even looser SD standard would only improve the fit of the linear regression and further confirm the flat spots are anomalous.

(Recognizing this is 500yrd data and Bc variability is influencing the read velocities, so the 500yrd SD may not align with the muzzle SD).

As another confirmation of inherent error possible the data readings is the fact two of those strings were fired only an hour and 6min apart, but read 6fps difference in average velocity. So both averages - the 2309 and 2315 actually are within the same inherent error of the load you were shooting, and on a total set,





So based on that data alone, there remains a ~7fps overlap where your true average for that population of ammo exists, but there’s no way to tell what the actual average for that ammo really is… but we sure want to tell ourselves 2309 or 2315 are the true averages.

it sure is weird that it works over, and over, and over again.

The reality is that is “working over and over again” because you want it to “work,” and you believe data says something it isn’t saying. It’s like the kid on Big Daddy, the movie, playing cards with Rob Schneider - “different cards every time, but he says he wins.” It’s observation bias, and you’re always the observer.

 

[Varminterror]

With greater than 99% fit, here is some of my load development data with a linear regression applied.

I COULD look at the few places where flat spots seem to repeat, letting my primitive human instinct seek out patterns and then tell my advanced logical faculties that those patterns mean something, OR, I can avoid lying to myself, apply even the most rudimentary error projection to the analysis, recognizing that “standard deviation” measures how far any given shot is likely to be from truth, and accept that a straight line fits nearly perfectly.

The null set (that no inflection exists) just isn’t very interesting.

 

[quest450]

WHOA Nelly

 

[ballisticdaddy]

An SD of 14.9 across 20 rounds or 13.7 across 20 is roughly 3x looser than the SD I used to analyze your data trend above. The yellow lines would be 3x wider, with ALL shots within those boundaries being valid predictions (proportionately distributed, of course) for actual velocity opportunities. In other words, I applied a linear regression with a tighter standard, about 3x tighter than your actual loads, and the linear regression fit your data. Applying even looser SD standard would only improve the fit of the linear regression and further confirm the flat spots are anomalous.

The velocities you are referencing are generated from a shotmarker, any of us that have shot using this system are aware the velocities and SD's are meaningless.

 

[LVLAaron]

The velocities you are referencing are generated from a shotmarker, any of us that have shot using this system are aware the velocities and SD's are meaningless.

I think they're "OK" - They're consistent but not accurate, in my experience. Our club has one that's 100fps different than the rest. But I'm with you, I wouldn't use it's data for scientific rocket surgery.


@Varminterror - Don't mistake my position here. I don't care. I'm just sharing my observations and data from guns I shoot. I'm not saying my way is the right way. This just happens to be the way that works out for me. My good 300wsm load happened to land on a flat spot. You can't tell me I'm "wrong" or "biased" for noticing that and sharing my data. I dont see anyone else sharing theirs. So please, if you want to continue, show us your scores and how you got there.

EDIT: Any time the statisticians start getting their underpants in a wad I like to remember that in order to satisfy the data they want: A) You would more than burn out a barrel and B) Barrels are living breathing tools; they change every time you fire them. A single load isn't going to last the entire life of the barrel, or 500 shots if we're being realistic.

 

[Varminterror]

The velocities you are referencing are generated from a shotmarker, any of us that have shot using this system are aware the velocities and SD's are meaningless.

Not my data - it was the reference provided to support an argument towards one idea, so inherently became the only evidence to be used to unpack the proposed idea.

You’ll note, I specifically referenced above that the shot marker data is not representative of MV data due to its method of capture.

 

[tom]

Ladders work great, but I'm talking holes, not velocities. I keep the data, but really I could get to the same goal if I forgot the crony at home.

Tom

 

[ballisticdaddy]

Not my data - it was the reference provided to support an argument towards one idea, so inherently became the only evidence to be used to unpack the proposed idea.

You’ll note, I specifically referenced above that the shot marker data is not representative of MV data due to its method of capture.

I realize that was not your data, but feel you are missing the whole point...the scores were all clean and there were no velocities indicated on the scoring sheet. For there to be 3 clean targets in a row then the rifle, load development and shooter were more than adequate as the score sheet confirms.

 

[Varminterror]

Any time the statisticians start getting their underpants in a wad I like to remember that in order to satisfy the data they want: A) You would more than burn out a barrel and B) Barrels are living breathing tools; they change every time you fire them. A single load isn't going to last the entire life of the barrel, or 500 shots if we're being realistic.

This is kind of a nonsense response that a lot of folks put together to defend ignoring their own results when they are presented back. But it’s really not true - if we’re being realistic, the good statistician will tell you NOT to burn out a barrel by seeking minute differentiation, but rather that the influence of these variables is so inconsequential that no shooter can truly differentiate the variables.

A good statistician, I am not, but I AM a good experimental designer. What those of us in this trade know is how to differentiate inductive vs. deductive experimentation. Folks have believed for at least 60 years that the Audette/Satterlee curve produces flat spots which are reliable and repeatable nodes - believing we can deduce “the best load” by identifying it among the rest. But in reality, this method, among many others, has worked for generations because we are simply doing INDUCTIVE experimentation - we set out to find a good load, and we find one.

So the “good statistician,” or good experimental designer, will run an ANOVA on these potential input variables and determine there is no inflection, no distinct shift in influence, so they will NOT tell you to shoot some exhaustive round count - they will tell you to not waste rounds or energy conducting unproductive and inconclusive analysis. They’ll tell you to save your ammo, and simply live happy knowing there is no sudden inflection. Good loads are good.

@Alex Wheeler has a video out there - at least one - from a few years ago where he describes (might have been a collaboration with Cortina? I know there were two talking heads, and they reviewed the 7 PRCW as “new” at the time) that velocity nodes aren’t real, that so many folks end up having good results at the same speeds because we’re shooting the same cartridges with the same bullets, same powders, same barrel lengths - so the “velocity nodes” we think happen with some bullets are really coincidental to the fact we’re using all of the same components.

Getting axle wrapped on data which doesn’t matter just isn’t productive. Picking a charge weight by a flat spot is largely just throwing a dart at a board with a blindfold.