Depends. The OP may go off his meds again, showing his ass and tender feelers at the same time. Creating the need for a delete to avoid looking like a total asshat...
- Thread starter Sniper King 2020
- Start date
Reason for keeping is the explanation posts from others. The other threads getting deleted lost some good. Acting like a retard on the internet should hurt some. Could they have been cleaned up? Probably good for a technical thread.
Pragmatically, it doesn't matter if you are 'sampling' enough shots to get a good idea of your actual SD. At least in my own experience, you need to have about 20-rounds +/- in order to get a good idea of your SD. The further below that you drop in terms of round count, the further into the noise you'll find yourself.
Here's a real life example with 33 rounds:
2628
2627
2638
2611
2612
2623
2617
2635
2625
2631
2630
2620
2620
2637
2645
2630
2610
2599
2596
2617
2611
2626
2600
2625
2638
2602
2609
2610
2598
2629
2615
2621
2629
STDEV.P = 12.665 fps
STDEV.S = 12.862 fps
Here's another 30 rounds from a different rifle (different cartridge, different ammo, different day)
2889
2876
2839
2865
2856
2856
2860
2834
2855
2860
2874
2868
2868
2894
2883
2854
2903
2900
2869
2915
2923
2862
2870
2848
2881
2915
2886
2877
2881
2892
STDEV.P = 21.732 fps
STDEV.S = 22.104 fps
Here's a third example (different barrel/ammo from the 2 posted above):
2995
2996
3028
2993
3022
2978
3003
2984
3003
2990
2975
2995
3048
3025
3009
3026
3037
3022
3028
3005
2996
3007
3038
3023
3035
3010
3018
3021
3020
3049
STDEV.P = 19.455 fps
STDEV.S = 19.788 fps
Once you get about 20 rounds+ (approximately) not only does the variation within each "method" become more stable, but the difference between the two "methods" becomes negligible from an application standpoint. It's really a non-issue unless you are one of these guys that thinks a 3-shot group is a meaningful test.
Here's a real life example with 33 rounds:
2628
2627
2638
2611
2612
2623
2617
2635
2625
2631
2630
2620
2620
2637
2645
2630
2610
2599
2596
2617
2611
2626
2600
2625
2638
2602
2609
2610
2598
2629
2615
2621
2629
STDEV.P = 12.665 fps
STDEV.S = 12.862 fps
Here's another 30 rounds from a different rifle (different cartridge, different ammo, different day)
2889
2876
2839
2865
2856
2856
2860
2834
2855
2860
2874
2868
2868
2894
2883
2854
2903
2900
2869
2915
2923
2862
2870
2848
2881
2915
2886
2877
2881
2892
STDEV.P = 21.732 fps
STDEV.S = 22.104 fps
Here's a third example (different barrel/ammo from the 2 posted above):
2995
2996
3028
2993
3022
2978
3003
2984
3003
2990
2975
2995
3048
3025
3009
3026
3037
3022
3028
3005
2996
3007
3038
3023
3035
3010
3018
3021
3020
3049
STDEV.P = 19.455 fps
STDEV.S = 19.788 fps
Once you get about 20 rounds+ (approximately) not only does the variation within each "method" become more stable, but the difference between the two "methods" becomes negligible from an application standpoint. It's really a non-issue unless you are one of these guys that thinks a 3-shot group is a meaningful test.
That was another point of contention. In practical terms, it really doesn't matter if you use N or N-1. As long as you don't mix the two, your data analysis will give you the same answer. And once you fire enough shots, the two SD's converge pretty close.
But, he was adamant that everyone except Garmin was "wrong."
I think this is more like SPAM.Have y'all not ever heard the phrase "don't feed the troll"?
I see you've met my students.....
Yes I am on a break this semester because of surgery..
In reality I was burned out from the very thing you are talking about--privileged little bastards (Graduate students) can barely do a regression and definitely can't write a report.. I'll admit I am a weirdo and get a giant boner when we start tacking about hessians, backpropagation, etc... I watch Youtube videos on Math (current subject Riemann Geometry Metric Tensors for General Relativity) for recreation (And the entire bear pit goes : "That explains so much")
My own "data scientist" irritates the shit outta me calling me up everyday with his latest bug that he can't figure out (look at your data Moron) and I want to scream "READ THE FUCKING ERROR MESSAGE AND GOOGLE THAT SHIT", but I politely assist him. My postdoc is less irritating but is very much like you describe: cut and paste. Took her 6 months to do this experiment that I did in one afternoon. (She wanted a Transformer to my CNN because TRANSFORMERS!!!--oh look the loss is 0.02 better that was worth it and its not even statistically significant--guess the old boy knows his shit after all). Ok swap out the models and go. FFS what is the hold up.
Then the best part: I have a bitching new server that just came up Mid December--I had a PAIN getting it installed. I set them both up on a brand spanking new Dual GPU 24GB Cards (I know not that impressive, but coming from a 16GB single server--again big stiffy). They still haven't logged in a month later. I stayed up all night the first night going WEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE like a 16year old with a hooker and a Ferarri.. abusing the shit outta that thing to "stress test it"
That's my story and I'm stickin to it.
I'd say we should get a beer and commiserate but I can't drink for a year as my stomach has gone missing.
Just waiting to find the right startup to cashout on myself. If you can't beat em, take the investors money and run....
OP can nuke it anytime until it’s locked, IIRC.
It’s nice knowing others are seeing the same issues. I thought I was going insane.I see you've met my students.....
Yes I am on a break this semester because of surgery..
In reality I was burned out from the very thing you are talking about--privileged little bastards (Graduate students) can barely do a regression and definitely can't write a report.. I'll admit I am a weirdo and get a giant boner when we start tacking about hessians, backpropagation, etc... I watch Youtube videos on Math (current subject Riemann Geometry Metric Tensors for General Relativity) for recreation (And the entire bear pit goes : "That explains so much")
My own "data scientist" irritates the shit outta me calling me up everyday with his latest bug that he can't figure out (look at your data Moron) and I want to scream "READ THE FUCKING ERROR MESSAGE AND GOOGLE THAT SHIT", but I politely assist him. My postdoc is less irritating but is very much like you describe: cut and paste. Took her 6 months to do this experiment that I did in one afternoon. (She wanted a Transformer to my CNN because TRANSFORMERS!!!--oh look the loss is 0.02 better that was worth it and its not even statistically significant--guess the old boy knows his shit after all). Ok swap out the models and go. FFS what is the hold up.
Then the best part: I have a bitching new server that just came up Mid December--I had a PAIN getting it installed. I set them both up on a brand spanking new Dual GPU 24GB Cards (I know not that impressive, but coming from a 16GB single server--again big stiffy). They still haven't logged in a month later. I stayed up all night the first night going WEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE like a 16year old with a hooker and a Ferarri.. abusing the shit outta that thing to "stress test it"
That's my story and I'm stickin to it.
I'd say we should get a beer and commiserate but I can't drink for a year as my stomach has gone missing.
Just waiting to find the right startup to cashout on myself. If you can't beat em, take the investors money and run....
I remember when I built a CNN model on my 16GB laptop and even with small batches it wasn’t enough memory (even with resizing). What’s worse is you need a ton of images to even make a good model so we need even more memory. So, I feel your pain on the memory constraints.
Seems like a garden variety forum suicide post to me.
I'm going to watch the Twilight Saga tonight after this. Wish me luck!
Weird. It was a joke.
Yes, you`re right, it is indeed a conscious choice to be here. One of the things one learns to do in seven decades on the planet is how to discern and filter out the useful nuggets of information and advice while simultaneously ignoring the bullshit and the assholes that provide it.I already know that you two are going to have a great discussion about how mean the forum is that you make a conscious choice to be on
STDEV.S is the correct method. End of story.
The difference between the two (open google--cause statisticians don't use those terms, there is standard deviation (STDEV.P)--statisticians use sigma and then there is ESTIMATED standard deviation,(STDDEV.S) which is known a 's'.
Its like your measured values is "x" and the mean is "mu"
English variables mean "measured" and greek variables are "intrinsic"--properties of the distribution
Standard Deviation is the width on a Normal or Gaussian Distribution. The formula is:
View attachment 8323682
sigma is the Std Dev. mu (the u thingy) is the mean).
Its a complicated formula, but just know that sigma controls the width, and mu controls the location:
View attachment 8323683
That's pretty standard stuff, but that's pure THEORY (egads a running gag in my own posts). When you MEASURE something you are sampling from that theoretical distribution (you get points at random, from a random distribution). Assuming your distribution is normal (not always safe), you can estimate the TRUE Standard Deviation by using the Estimated Standard Deviation, s. They key here is there is some unknown TRUE sigma which we try and estimate by taking samples. The Estimated and True will only begin to approach equal once the number of samples becomes very very large.
So the TLDR version is STDEV.S would be the statistician's choice to ESTIMATE the standard deviation of your velocity.
Example: I ran created a distribution with mean 2800 and simga 10. I grabbed 5 samples at random, 10 times and took the STDEV.P--However, we KNOW its 10 (I made it!). Here are the results:
"True Method"
[13.275930373733024,
8.342655830573312,
7.23926247655942,
5.156843761443447,
8.550960467769086,
13.76093532000941,
3.2590625589140196,
7.783767404854264,
3.911747500033366,
7.031416934586289]
Estimated:
[14.842941390110614,
9.327372775023447,
8.09374150227517,
5.765526599966628,
9.560264439422538,
15.385193404759432,
3.6437427123280806,
8.702516519150631,
4.373466660444734,
7.861363121939068]
With the estimated std dev, you get a value that is higher (instead of dividing by num of samples, you divide by n-1 for math reasons--in reality using the "true" method induces a bias towards 0)--so you could get lucky or you could get unlucky.
OR
Based on the following random data I "sampled":
array([2794.70526224, 2816.48889134, 2797.31856425, 2785.55297754,
2791.62315356])
I would publish an s of 11.67 with 95% CI of (4.07, 14.80) meaning I am 95% sure the TRUE sigma lies on the interval 4.07 to 14.80
You use confidence intervals : since I have to publish shit for the dreaded peer review, I would say my estimated std dev is 10.4 (4.07, 14.80)
So I read that as a sample size of 5 means dick either way. But "s" is the correct method.
Also statisticians don't use excel, they use R, which is a filthy language.
I teach data science/AI so I use python:
Go to google colab and reproduce my results. I prob made a mistake, but I fuck it. That's how you think of this as a "statistician"
spaces are important in the code btw Due to RNG, your number may vary slightly from mine. I'm, not getting into psuedo-random numbers with howler monkies.
Google Colaboratory
colab.research.google.com
[FONT=courier new]import numpy as np p=[] s=[] n=10 size=5 for i in range( n ): x =np.random.normal(loc=2800, scale=10.0, size=size) p.append(np.std(x)) tmp = (x-x.mean())**2 tmp = tmp.sum() tmp = 1/(size-1) * tmp tmp = np.sqrt(tmp) s.append(tmp) import matplotlib.pyplot as plt plt.scatter(np.linspace(0,9,10),p,label="True") plt.scatter(np.linspace(0,9,10),s,label="Estimated") plt.legend() plt.show() sigma=[] for i in range(100): tmp = np.random.choice(x,5,replace=True) print(tmp) sigma.append(np.sqrt((((tmp-tmp.mean())**2).sum())/4)) print(np.percentile(sigma,95)) print(np.percentile(sigma,5)) print(np.sqrt((((x-x.mean())**2)/4).sum()))[/FONT]
Anyone else get a boner from this, or is it just me?
P
Anyone else find it hilarious that benchrest people hand out a trophy called "sniper king?" lol
I thought it was a joke at first... but nope... it's real.
I thought it was a joke at first... but nope... it's real.
Well, I vote for user name change to “Bryan the tard” of misfit toy island…..
Good to see yall triggered and looks like yall figured out the SD thing be sure to join the FB page dedicated to y’all
www.facebook.com
be sure to join the FB page dedicated to y’all Log into Facebook
Log into Facebook to start sharing and connecting with your friends, family, and people you know.
www.facebook.com I forgot to mention that I am proud to live rent free in your mindsGood to see yall triggered and looks like yall figured out the SD thing
Log into Facebook
Log into Facebook to start sharing and connecting with your friends, family, and people you know.
good stuff kids. Keep it up!
I forgot to mention that I am proud to live rent free in your minds
You are the latest in line of retards that span over twenty years. This might be your first time but it is nothing new for the Hide.I forgot to mention that I am proud to live rent free in your minds
Don’t you high road me, I’m a low road guy.I forgot to mention that I am proud to live rent free in your minds
And 6 posts today. He’s talking to himself again, and still not liking the answers I bet.Private group, 1 member.
How can one even be a sniper with no proof whatsoever of bow kills, concrete work, Oakley ownership or Velcro Crocs?
Nothing snipery here unless math is performed while wearing a ghillie suit overseeing a convoy withdrawing from a village in Eritrea on a secret mission for a Wyoming Senator who squeals like a piggy.
Oh and.
Sirhr
Nothing snipery here unless math is performed while wearing a ghillie suit overseeing a convoy withdrawing from a village in Eritrea on a secret mission for a Wyoming Senator who squeals like a piggy.
Oh and.
Sirhr
- Nov 3, 2010
- 8,645
- 17,545
Polishing them. In SF ROTC. Before renting a lambo on Haiti and cornering the scope market. All classic sniper tactics. C’mon man… get with the program!
STDEV.S is the correct method. End of story.
The difference between the two (open google--cause statisticians don't use those terms, there is standard deviation (STDEV.P)--statisticians use sigma and then there is ESTIMATED standard deviation,(STDDEV.S) which is known a 's'.
Its like your measured values is "x" and the mean is "mu"
English variables mean "measured" and greek variables are "intrinsic"--properties of the distribution
Standard Deviation is the width on a Normal or Gaussian Distribution. The formula is:
View attachment 8323682
sigma is the Std Dev. mu (the u thingy) is the mean).
Its a complicated formula, but just know that sigma controls the width, and mu controls the location:
View attachment 8323683
That's pretty standard stuff, but that's pure THEORY (egads a running gag in my own posts). When you MEASURE something you are sampling from that theoretical distribution (you get points at random, from a random distribution). Assuming your distribution is normal (not always safe), you can estimate the TRUE Standard Deviation by using the Estimated Standard Deviation, s. They key here is there is some unknown TRUE sigma which we try and estimate by taking samples. The Estimated and True will only begin to approach equal once the number of samples becomes very very large.
So the TLDR version is STDEV.S would be the statistician's choice to ESTIMATE the standard deviation of your velocity.
Example: I ran created a distribution with mean 2800 and simga 10. I grabbed 5 samples at random, 10 times and took the STDEV.P--However, we KNOW its 10 (I made it!). Here are the results:
"True Method"
[13.275930373733024,
8.342655830573312,
7.23926247655942,
5.156843761443447,
8.550960467769086,
13.76093532000941,
3.2590625589140196,
7.783767404854264,
3.911747500033366,
7.031416934586289]
Estimated:
[14.842941390110614,
9.327372775023447,
8.09374150227517,
5.765526599966628,
9.560264439422538,
15.385193404759432,
3.6437427123280806,
8.702516519150631,
4.373466660444734,
7.861363121939068]
With the estimated std dev, you get a value that is higher (instead of dividing by num of samples, you divide by n-1 for math reasons--in reality using the "true" method induces a bias towards 0)--so you could get lucky or you could get unlucky.
OR
Based on the following random data I "sampled":
array([2794.70526224, 2816.48889134, 2797.31856425, 2785.55297754,
2791.62315356])
I would publish an s of 11.67 with 95% CI of (4.07, 14.80) meaning I am 95% sure the TRUE sigma lies on the interval 4.07 to 14.80
You use confidence intervals : since I have to publish shit for the dreaded peer review, I would say my estimated std dev is 10.4 (4.07, 14.80)
So I read that as a sample size of 5 means dick either way. But "s" is the correct method.
Also statisticians don't use excel, they use R, which is a filthy language.
I teach data science/AI so I use python:
Go to google colab and reproduce my results. I prob made a mistake, but I fuck it. That's how you think of this as a "statistician"
spaces are important in the code btw Due to RNG, your number may vary slightly from mine. I'm, not getting into psuedo-random numbers with howler monkies.Google Colaboratory
[FONT=courier new]import numpy as np p=[] s=[] n=10 size=5 for i in range( n ): x =np.random.normal(loc=2800, scale=10.0, size=size) p.append(np.std(x)) tmp = (x-x.mean())**2 tmp = tmp.sum() tmp = 1/(size-1) * tmp tmp = np.sqrt(tmp) s.append(tmp) import matplotlib.pyplot as plt plt.scatter(np.linspace(0,9,10),p,label="True") plt.scatter(np.linspace(0,9,10),s,label="Estimated") plt.legend() plt.show() sigma=[] for i in range(100): tmp = np.random.choice(x,5,replace=True) print(tmp) sigma.append(np.sqrt((((tmp-tmp.mean())**2).sum())/4)) print(np.percentile(sigma,95)) print(np.percentile(sigma,5)) print(np.sqrt((((x-x.mean())**2)/4).sum()))[/FONT]
Am I the only one that fapped to this?
So, he is starting thread and special Facebook page. And he is living our heads and we are triggered. Yep yep that stands up to scrutiny. 
- Nov 3, 2010
- 8,645
- 17,545
Good to see yall triggered and looks like yall figured out the SD thing
Log into Facebook
Log into Facebook to start sharing and connecting with your friends, family, and people you know.
Attachments
Is that your brain or are you just happy to see me?
We need to start handing out the Veer award again
Turkeytugger must not have a FB profile. I’m sure those two could jerk each other off for hours in a group like that.
What’s the probability that Flips will show up again…including the standard deviation?
The extreme spread will be wider.
Me too lol
Bear Pit does not care how you feel.
Similar threads
