ES vs SD

[3rd Monkey]

Alright, fresh to the forum and need to see what the deal is (to know if I'm cut out for this place).

ES or SD... It's not a "both" kind of question.

Which is KING!?

If it's been discussed before... I don't give a shit... Wasn't here.

 

[redbullitt]

If ES is acceptable for your intended distance and target, or time and budget constraint, then SD all the way until you exceed same time/budget wall lol.

 

[spife7980]

 

[3rd Monkey]

View attachment 7853180

Glad you let me know you're here...

 

[3rd Monkey]

If ES is acceptable for your intended distance and target, or time and budget constraint, then SD all the way until you exceed same time/budget wall lol.

Do you chase SD or mitigate ES? That's the question...

 

[straightshooter1]

Alright, fresh to the forum and need to see what the deal is (to know if I'm cut out for this place).

ES or SD... It's not a "both" kind of question.

Which is KING!?

If it's been discussed before... I don't give a shit... Wasn't here.

Read this:

Muzzle Velocity Stats – Statistics for Shooters Part 2

Consistent muzzle velocity is key for long-range shooting, otherwise, bullets that leave the muzzle...

precisionrifleblog.com

 

[nn8734]

Do you chase SD or mitigate ES? That's the question...

neither. You look at your target and compare it to your expectations and adjust if/as necessary.

 

[b6graham]

once you realize you're not likely to shoot enough through a chrono to have statically significant data to actual determine your rifle/reloading setup's capabilities its mostly unimportant which you use as a load one day could be ES 10 and SD 3 and the next day ES 25 and SD 13 by pure luck of which cartridges you shoot from the box of 100 you loaded. same goes for a satterlee load development, OCW, tuner, seating depth blah blah blah

2 years ago i got berger factory 144 6.5CM. first three shots identical speed. 1 hole group. prefect ammo right??? itll all be like that right??? in reality that load had an ES of 25 and shoots 1/2" when you add a few more groups/days of data

they both matter...somewhat, but how much time and effort and components you want to waste chasing 'great' data is up to you

pick an appropriate speed you want for the cartridge/bullet and send it

 

[JBoomhauer]

once you realize you're not likely to shoot enough through a chrono to have statically significant data to actual determine your rifle/reloading setup's capabilities its mostly unimportant which you use as a load one day could be ES 10 and SD 3 and the next day ES 25 and SD 13 by pure luck of which cartridges you shoot from the box of 100 you loaded. same goes for a satterlee load development, OCW, tuner, seating depth blah blah blah

2 years ago i got berger factory 144 6.5CM. first three shots identical speed. 1 hole group. prefect ammo right??? itll all be like that right??? in reality that load had an ES of 25 and shoots 1/2" when you add a few more groups/days of data

they both matter...somewhat, but how much time and effort and components you want to waste chasing 'great' data is up to you

pick an appropriate speed you want for the cartridge/bullet and send it

Lately I have been doing load development with the USPS method.

Pick a number and “ if it fits it ships”

 

[OREGUN]

This is painful.

Go for lowest possible ES. SD will be numerically small if your ES is small. With an ES of 10, All the shots will be close to the same velocity. Said another way, I’d rather have an ES of 10 even if all the shots were 10FPS apart, i.e. a range of 2900-2910 with all shots either 2900 or 2910, which is not enough to matter, than an ES of 100 with any SD….wherein at least some of the shots are going to be 100 FPS different than some of the others...which is enough to matter.

But also what everyone else said.

 

[Feniks Technologies]

I just made a post about this on social media a few days ago.

Here’s a copy/past:

For anyone who doesn’t know exactly what your Standard Deviation really means, here’s a simple break down with visual aid.

The example:

Avg Muzzle Velocity: 2950 fps

SD: 5 fps

So, if we took 100 shots the breakdown would look like this rounding down/up)

68 shots = 2945 - 2955 fps
27 shots = 2940 - 2960 fps
5 shots = 2935 - 2965 fps

There is still always a chance (0.3%) of having a MV outside of this range. As well as you need enough rounds tested to have a confident avg velocity as well as SD.

What this also means is if you have a 10sd, you will have about a 60es. If you’re shooting practical/prs stuff, it won’t matter unless the targets are very small and very far away, which is usually rare.

To have a sub 20 extreme spread, you’ll need a ~ 3sd. Which is world class ammo (despite what your 5 shot string chrono told you.

This is also why you have to be careful when just looking at your Extreme Spread. It may or may not be telling you the whole story. Standard Deviation will give you a much better idea of what your ammo is going to do in the long run.

Explaining this to shooters who are having issues with their loading usually helps to lessen the stress and expectations they impose on themselves.

 

[Feniks Technologies]

I’m going to point out again, because it’s that important:

Your ES will always be ~6 times your SD. 99.7% of the time to be exact. With 95% of the time your ES is 4 times your SD.

So, even those times when you shoot a string and it’s 20es with an 8sd……your ES isn’t actually 20.

 

[marchboom]

ES or SD? Same thing as MOA or MILs. Both give you the info you need, just via a larger or smaller number.

 

[Doom]

I’m going to point out again, because it’s that important:

Your ES will always be ~6 times your SD. 99.7% of the time to be exact. With 95% of the time your ES is 4 times your SD.

So, even those times when you shoot a string and it’s 20es with an 8sd……your ES isn’t actually 20.

Put another way, the TWO values that make up ES have the following effect in terms of effect on SD

3 shots 67% of the SD data
5 shots 40% of the SD data
10 shots 20% of the SD data
20 shots 10% of the SD data
50 shots 4% of the SD data
100 shots 2% of the SD data

Statistically speaking ES has no meaning other than the max/min values of a test data set. As test sample size grows ES will grow because the likely hood of testing a single extreme value increases. At the same time the actual tested sample SD will approach the true SD of a large population and the ES values represent extreme values unlikely to occur with any frequency.

 

[OREGUN]

Are bullet velocities distributed normally around the mean? Just curious. Not teeing up any brilliant ideas.

 

[Tchitcherine]

Amazon product ASIN 1400067936

 

[OREGUN]

I’d still rather shoot ammo with a small ES…the only thing I’m really worried about is how badly it will hurt me if the next bullet down range is one of the nearly impossible outliers that are beyond 3 SDs of the mean.

 

[Feniks Technologies]

I’d still rather shoot ammo with a small ES…the only thing I’m really worried about is how badly it will hurt me if the next bullet down range is one of the nearly impossible outliers that are beyond 3 SDs of the mean.

That’s the thing, the ES will always be what it’s going to be in the long run.

If you have a high ES load with a 5sd and a low ES load with a 5sd….

What has happened is you just haven’t ran across the rounds yet on the low ES rounds.

Remember, just about 70% of the rounds will be a 10es and 95% will be 20es. So in one sample size you got something in that 5%

 

[Feniks Technologies]

ES or SD? Same thing as MOA or MILs. Both give you the info you need, just via a larger or smaller number.

Yea…..no. Lol

 

[marchboom]

That’s the thing, the ES will always be what it’s going to be in the long run.

If you have a high ES load with a 5sd and a low ES load with a 5sd….

What has happened is you just haven’t ran across the rounds yet on the low ES rounds.

Remember, just about 70% of the rounds will be a 10es and 95% will be 20es. So in one sample size you got something in that 5%

Kindly explain what you stated. Doesn't make sense to me.

 

[Feniks Technologies]

Kindly explain what you stated. Doesn't make sense to me.

Read my post above about what SD really is.

Your true ES will be 6x your SD 99.7% of the time.

 

[Gohring65]

Go out and shoot your ammo until you’re good enough that you can’t explain your misses, then go down that rabbit hole.

 

[smoothy8500]

Kindly explain what you stated. Doesn't make sense to me.

To simplify it, the range of the velocities is broken into 6 segments (SD):

 

[Doom]

Are bullet velocities distributed normally around the mean? Just curious. Not teeing up any brilliant ideas.

Yes. Almost everything you test will follow a normal distribution about a mean. That is why it is university accepted. What is almost universally missed is that the mean of a population is not necessarily the tested mean. Making sense of the test data requires analysis called ANOVA. The Blog referenced in post #6 has a three part series on how test data should be analyzed.

You may also find this of interest:

Stats for velocities and groups.

So why is a 5 shot the standard for groups but not for velocity? I’ve been seeing a lot of discussions lately about SD numbers being useless without a large sample size. I agree that the SD may be unreliable with a smaller sample size but I don’t think it is completely worthless. I use the 5...

www.snipershide.com

 

[semperfireloading]

Let me ask the question if a different way.

Which color do you like better? BLUE? OORRRR BLUE!?

I like blue, its better

 

[OREGUN]

I’m team blue. For sure.

 

[3rd Monkey]

ES or SD? Same thing as MOA or MILs. Both give you the info you need, just via a larger or smaller number.

I disagree, along with hundreds of F class shooters (tell them they are shooting 1/10th mil instead of 1/4 moa) but agree with your gearing towards the mil/moa debate being strikingly similar to ES and SD (just in a different aspect). Bravo!

 

[hlee]

The TLDR version is that shooters aren’t generating enough data from which to generate meaningful statistics. The question is nearly as meaningless as a 5 round SD or ES value.

 

[3rd Monkey]

The TLDR version is that shooters aren’t generating enough data from which to generate meaningful statistics. The question is nearly as meaningless as a 5 round SD or ES value.

Not really. You can gather meaningful and tunable stats from one, not NECESSARILY from the other... In my opinion. Do you think they are equal? Curious... not picking.

 

[hlee]

ES vs SD

Alright, fresh to the forum and need to see what the deal is (to know if I'm cut out for this place). ES or SD... It's not a "both" kind of question. Which is KING!? :p If it's been discussed before... I don't give a shit... Wasn't here.

www.snipershide.com

ES vs SD

Alright, fresh to the forum and need to see what the deal is (to know if I'm cut out for this place). ES or SD... It's not a "both" kind of question. Which is KING!? :p If it's been discussed before... I don't give a shit... Wasn't here.

www.snipershide.com

I typed and deleted a lot of stuff before I realized that the 2 posts above neatly summarize my thoughts on the matter.

 

[Pharmseller]

about a mean. That is why it is university accepted. the mean of a population is not necessarily the tested mean.

 

[Feniks Technologies]

Not really. You can gather meaningful and tunable stats from one, not NECESSARILY from the other... In my opinion. Do you think they are equal? Curious... not picking.

The only thing you’ll usually be able to gain from small sample size is what isn’t working. Not always, but that’s about the only thing you can do with it.

 

[Feniks Technologies]

Here’s a very good example. Snagged this off a FB post. 10 shots over AB doppler.

Avg: 2857
SD: 8
ES: 25

Let’s assume the 8 sd is the actual long term SD.

If they were to shoot 100 shots, the ES would be ~48fps.

You could take another load and do the same thing and get an 8sd and a 48es.

And people would want to pick the 25es.

When in reality, they are both the exact same.


Remember, for this shooter:

68%: 16es (68 shots)
95%: 32es (27 shots)
99.7%: 48es (5 shots)

His ES of 25 was 1.5 standard deviations (12.5 each way). It was 14 up and 11 down. So, pretty close.

Which considering 68% is 16es…..you can now see how easy it was for his ES to say 25 when the SD is 8.

This is why SD tells the whole picture. It’s counter intuitive for many because people say “every shot falls inside of the ES”. Which is technically true. But your ES is usually much less statistically confident than your SD.

 

[OREGUN]

Here’s a very good example. Snagged this off a FB post. 10 shots over AB doppler.

Avg: 2857
SD: 8
ES: 25

Let’s assume the 8 sd is the actual long term SD.

If they were to shoot 100 shots, the ES would be ~48fps.

You could take another load and do the same thing and get an 8sd and a 48es.

And people would want to pick the 25es.

When in reality, they are both the exact same.


Remember, for this shooter:

68%: 16es (68 shots)
95%: 32es (27 shots)
99.7%: 48es (5 shots)

His ES of 25 was 1.5 standard deviations (12.5 each way). It was 14 up and 11 down. So, pretty close.

Which considering 68% is 16es…..you can now see how easy it was for his ES to say 25 when the SD is 8.

This is why SD tells the whole picture. It’s counter intuitive for many because people say “every shot falls inside of the ES”. Which is technically true. But your ES is usually much less statistically confident than your SD.

Why would we assume that the SD is the accurate number? Couldn't you do the same thing by saying 25 is right and thus, the SD must be 4?

 

[Feniks Technologies]

Why would we assume that the SD is the accurate number? Couldn't you do the same thing by saying 25 is right and thus, the SD must be 4?

Because the ES is only 20% of the sample size and SD in general is almost always a more accurate prediction.

This is why you’ll never see a statistician reference ES.

 

[Feniks Technologies]

Why would we assume that the SD is the accurate number? Couldn't you do the same thing by saying 25 is right and thus, the SD must be 4?

This sums it up well:

Muzzle Velocity Stats – Statistics for Shooters Part 2

Consistent muzzle velocity is key for long-range shooting, otherwise, bullets that leave the muzzle...

precisionrifleblog.com

 

[Feniks Technologies]

Common example of what happens to SD and ES in the long term. SD is more stable once you have a few shots. ES can always be affected by a single shot. Making it far less important in the long run.

 

[marchboom]

Now that makes sense and puts the ES/SD issue into perspective.

 

[smoothy8500]

ES can always be affected by a single shot. Making it far less important in the long run.

I know this question will be off topic, but pertaining to how we choose what velocity to input for ballistic calcs. The "average" (mean) is affected by a single shot more so than the median or "middle" velocity number.

 

[Feniks Technologies]

I know this question will be off topic, but pertaining to how we choose what velocity to input for ballistic calcs. The "average" (mean) is affected by a single shot more so than the median or "middle" velocity number.

That might require a bit of work to figure out the true median though.

As in, your sd after 20-30 shots is going to be pretty stable from there on out. So you can use it to see your 95% and 99.7% ES.

But you would need quite a bit of data (probably 100 shots) to be able to accurately pinpoint the median. And even then it may be subject to basic variance.

Average is likely going to be a closer prediction.


Basically, it’s damn near impossible to theoretically predict a median. So you can’t use it without quite a large sample size. Hence average/mean being the better statistic as we are forced to be working theoretically in nature with shooting.

 

[Doom]

Median is useful in evaluating a set of test data, in particular when the data is skewed. Consider the following 5 shot group.

2580 2585 2593 2600 2700

The median is 2593, and the mean is 2611.6. In this case the data isn’t normally distributed. The mean is higher than 4 out of five shots. It is possible that the true mean is closer to 2593 than 2611.6. This data is not normally distributed.

So what does this mean? Likely the 2700 was not loaded correctly (it could be an outlier) or by odd chance we loaded consistently for 5 rounds and have one of the two shots that make up the real extreme spread. In either case, we don’t have enough data to make a decision.

 

[mtrmn]

Y'all makin' my head hurt. I'm running with Boomhauer.

Lately I have been doing load development with the USPS method.

Pick a number and “ if it fits it ships”

 

[3rd Monkey]

Common example of what happens to SD and ES in the long term. SD is more stable once you have a few shots. ES can always be affected by a single shot. Making it far less important in the long run.

ES directly affects the SD, and vice versa if you look at it from that aspect. 1 high/low fps shot will change the dynamics of the SD, no?

That's not where I compare SD to ES. In fact, I would compare ES to COAL before I would SD, in terms of tunability and in my opinion.

Ever had a small ES/SD and still grouped for shit?

 

[Feniks Technologies]

ES directly affects the SD, and vice versa if you look at it from that aspect. 1 high/low fps shot will change the dynamics of the SD, no?

That's not where I compare SD to ES. In fact, I would compare ES to COAL before I would SD, in terms of tunability and in my opinion.

Ever had a small ES/SD and still grouped for shit?

No. Once you have enough data to have a solid SD, a high or low velocity shot will not change it much at all. As shown in the graph above.

Hence having a 5sd will have a 20es 95% of the time. And 5% of the time it will be 30es. That 5% doesn’t change the sd.

And the comparison to OAL isn’t really applicable. Completely different. Also, you can make any load shoot small or large via seating depth. Completely different part of the puzzle.

 

[Feniks Technologies]

Again, SD is theoretical as in it predicts what will happen. And with enough data, it’s extremely accurate at predicting what will happen.

ES is just an absolute measurement of the high and low. And it’s only valid if you happen to have stumbled upon the actual high and low shots in your string.

70% of the time, that won’t be the case (it’s probably a bit different than 68%, but that’s the easiest way to think about it).

 

[straightshooter1]

Again, SD is theoretical as in it predicts what will happen. And with enough data, it’s extremely accurate at predicting what will happen.

ES is just an absolute measurement of the high and low. And it’s only valid if you happen to have stumbled upon the actual high and low shots in your string.

70% of the time, that won’t be the case (it’s probably a bit different than 68%, but that’s the easiest way to think about it).

Hmmmm??? I wouldn't characterize SD as something that "predicts' anything. SD tells you what you've done . . . much like a baseball batting average. Whatever the SD is for the next set of numbers is going to be is simply unknown. It could be lower, it could be higher, it could be the same . . . all depending on the variables. If one can be consistent with the variables, then one can see the same consistent SD. . . .but it's simply not predicting anything. It's a measure of the past, not the future.

 

[Feniks Technologies]

Hmmmm??? I wouldn't characterize SD as something that "predicts' anything. SD tells you what you've done . . . much like a baseball batting average. Whatever the SD is for the next set of numbers is going to be is simply unknown. It could be lower, it could be higher, it could be the same . . . all depending on the variables. If one can be consistent with the variables, then one can see the same consistent SD. . . .but it's simply not predicting anything. It's a measure of the past, not the future.

It’s a measurement of the past that we will be using to predict the future.

Batting average is a terrible example.

If using the same powder/bullets/barrel etc etc……

SD gives an extremely accurate prediction of future velocities.

So much so, that like I stated earlier, you can take a load with a 5sd/10es and know that load is going to very, very likely perform the exact same as one with a 5sd/30es.

If we couldn’t use SD to predict the future, we wouldn’t use the chrono numbers to work with load development.

 

[Feniks Technologies]

Which is why SD is used to predict future volatility and many other things.

I likely used the wrong wording as I didn’t mean the measurement is an actual measurement of the future (which I shouldn’t have to say because it’s impossible to measure the future).

But when comparing SD and ES, SD can be used to predict what will happen much more accurately than ES. As the only time ES can be used to predict is when you have sufficient data to be confident in the ES.

Which is not needed for the most part, because if you have a confident SD (which takes far fewer shots), then you can simply extrapolate your ES.

 

[Pharmseller]

How many data points are necessary to have confidence in SD?

 

[Feniks Technologies]

How many data points are necessary to have confidence in SD?

10 is usually around the minimum. 30 is where you won’t get much change past that.