alb wrote:
When you change your sights to compensate for RANDOM variations in the location of your shot groups, then you are, in effect, changing your sights RANDOMLY. When you change your sights RANDOMLY throughout a match, all that you really accomplish is to reduce the accuracy of your pistol. When you you reduce the accuracy of your pistol, your aggregate group size will increase. This increase in aggregate group size, based on a RANDOM reduction in accuracy, TRANSLATES TO LOWER SCORES.
Regards,
Al B.
This makes a lot of sense to me. Never the less, I'm sure I've been able to pick up a few points toward the end of a match, by moving my sights a few clicks to compensate for group POI changes due to fatigue. ( Usually takes longer to notice than I'd like, especially if the targets don't have to be manually changed).
Anonymous wrote:I'd rather take advices from top shooters than from people spreading the shots all over the target. :-)
I'll start by making the perhaps unwarranted assumption that you aren't a "top shooter" -- otherwise, you would most likely have said so. I'll make the further assumption that you don't have a very solid foundation in math or much of an understanding of randomness and probablity, and the effect that randomness has in everyday life, and in particular, pistol shooting. Otherwise, you would have responded to my post in kind, rather than with an ad hominem attack.
So, I'll try to explain the point that I was making in non-mathematical terms.
If you look at the post that started this thread, you'll see a target that represents an atypically tight group for this shooter, based on his own statement, as well as his stetement that his best match score so far is 490 and the fact that he has only been shooting for four months. Based on this, I would estimate that the RANDOM variation in the location of the center of his 20-shot groups is at least 4 to 5 MOA (8 to 10 clicks), or about 0.5 inches -- roughly the diameter of the 10-ring. Furthermore, the RANDOM variation in the location of the center of his 5-shot groups is probably considerably greater than that.
Regards,
Al B.
It does not really matter how well I shoot. But since you mentioned my performance level several times in your post so... My competition pb in AP60 is 575p. Much above average in my country, but not in top 10. My point was that "fact" advices from people barely shooting groups should be taken with a big pinch of salt.
Of course the original poster should have moved his sights. It's clearly off center. Shooting 50+ shots that are off center at a competition isn't that smart - losing points is what's happening.
At a typical competition I shoot about 5-10 sighting shots, and one by one closely watch where the hole is and how I felt the shot. Often I'll have to adjust the sights a few clicks up or down during the sighting series. But seldom during the match. (but it happens) I'm not really interested in why the shots are a bit low or high that day, I'll just move the sights along and shoot more tens. :-)
I think you are trying to make shooting more complicated than it is - like a true engineer would. Naah, keep things simple instead and shoot more tens! :-)
pgfaini wrote:Never the less, I'm sure I've been able to pick up a few points toward the end of a match, by moving my sights a few clicks to compensate for group POI changes due to fatigue. ( Usually takes longer to notice than I'd like, especially if the targets don't have to be manually changed).
Paul,
You've raised an interesting point. As long as you are making changes that are consistent with your ability to detect real changes in POI, then changing your sights during a match will work for you.
For example, the threshold with which the original poster was able to detect real changes to POI was probably a minimum of 10 clicks or more, whereas a shooter who averages 570 or more can probably detect real changes in the 3 to 5 click range.
If the original poster were to start chasing 3-click changes in group placement, it would be the functional equivalent of trying to measure the thickness of a sheet of paper with a yardstick. Sometimes, he would think that the strategy was working great, other times, he would think he was having a bad day and nothing was working right. Neither would actually be true, however. And over the long run, he'd be hurting his average score.
Humans seem to be hard-wired to attribute success to what they are deliberately doing, and failure to bad karma. This is how superstitions are formed. Unfortunately, humans don't seem to be particularly well suited to thinking probabilistically.
If your personal best is 575, then your average is probably close to 570. If you were to shoot the group that was originally posted, then of course you should move your sights.
But that shooter isn't you. We all shoot groups, but some of us shoot bigger groups than others. Somebody that shoots in the 530 to 540 range can keep all but a couple of his 60 shots in the 7-ring. Somebody who averages 570 can put all but a couple of his 60 shots in the 9-ring (he is still missing the 10-ring about half of the time). The difference between 535 and 570 is that at 570, you're placing all of your shots into less than 1/5 of the area of the 535 shooter.
In other words, you're probably shooting more than 5 times as well as a 535 shooter.
Someone who shoots in the 470 to 490 range is not only going to have a lot more variation in individual shot placement, but will have a lot more variation in group placement, particularly for 3 to 5 shot groups. He will also have a lot more variation in his individual scores.
What works for you isn't going to work for him -- at least not to the same degree of precision. Chasing imaginary 3-click changes in POI will only reduce his scores.
I benefited greatly from what Al B the Engineer has to say .. Especially when he put the scientific basis of his observations in ordinary plain English that even guest could grasp.
Why do I have a sneaky feeling that guest shoots nowhere near the 570 he claims.... The over 570 shooters are well known in their clubs and in national competitions, surely if he gave his name and club someone on here can verify his claims and shut the likes of me up ?
Shooting is an ART and a Science . It is so simple that it comes down to point and shoot , and so sophisticated as to pose a challenge to the most adept of the scientifically oriented.
Even the Art part of shooting is in reality that part of Science that is beyond the grasp of the conscious mind.
A rather disheartening part of what Al B said was his observation ( almost certainly true and correct to my mind ) is that to move from my level of 526 in the latest match to 570 , I have to shoot at least FIVE TIMES better !! However , part of the pleasure of shooting is in the challenge. So now I have to start "not accepting 8's" and not just 7's !
- Assume you are a shooter with good technique
- When your good technique is working, you can "call your shots" with a fair degree of accuracy
- Assume your equipment is in good working order
Now you show up at a match and have no reason to suspect your sights were knocked about since the last time. Your shots are somewhat consistently falling "off call."
Do you adjust your sights?
Why?
There is probably an underlying technique issue that is causing your shots to deviate from call, therefore, you have two choices:
a. Adjusting your sights to compensate won't fix the problem; but will however "paper over" the technique flaw and allow you to continue, with your flaw, with your sights "optimized" to minimize impact of your technique flaw;
b. Solve your technique issue
Steve Swartz wrote:There is probably an underlying technique issue that is causing your shots to deviate from call, therefore, you have two choices:
a. Adjusting your sights to compensate won't fix the problem; but will however "paper over" the technique flaw and allow you to continue, with your flaw, with your sights "optimized" to minimize impact of your technique flaw;
b. Solve your technique issue
I frequently have the experience that my shots don't seem to be going where I call them, and they seem to be grouping somewhere else, particularly on a fresh target.
Since I regularly shoot 50 shots or more at a single target (I'm cheap, and at 25 yards I don't get too many hits on the same hole), I also have the experience that if I leave my sights alone, I eventually end up shooting a group that has about the same size and center as every other 50-shot group that I shoot. In other words, that seemingly off-center group of a small number of shots is off-center due to random clumping.
Unless you occasionally shoot a large number of shots at the same target, you're a lot less likely to understand and appreciate this phenomenon. It happens to 580-shooters just as much as it does to 480-shooters. It's just that with 580-shooters, the displacements and group sizes are a lot smaller.
When I start throwing shots during a match, and I can't figure out why, I'd personally rather spend a couple of minutes dry-firing, rather than touch my sights. In other words, I prefer option 'b'.
I think I have a semantic disconnect with something and need clarification please.
I have difficulty resolving your use of the term "random clumping" with the generally established fact that random dispersion of a match-grade gun shooting match grade pellets is pretty much undetectable.
Please clarify?
I think what you call "random clumping" is the same thing I am referring to as "technique issue(s)?"
In other words, the clump forms because *I* am doing something different for those shots.
If what I claim is correct, then moving your sights to center any particular "clump" will only mask the technique issue.
And when you do shoot a correct shot, it will be "out" of the clump- which has the side effect of 1) negatively reinforcing good behavior; and 2) confusing the shooter, who just saw a great shot with his/her own eyes only to discover a hole somewhere else.
Instead of "Centering Your Clump" perhaps we should be "Centering Our Technique?"
Steve Swartz wrote:In other words, the clump forms because *I* am doing something different for those shots.
If what I claim is correct, then moving your sights to center any particular "clump" will only mask the technique issue.
And when you do shoot a correct shot, it will be "out" of the clump- which has the side effect of 1) negatively reinforcing good behavior; and 2) confusing the shooter, who just saw a great shot with his/her own eyes only to discover a hole somewhere else.
When Al refers to clumping, I think he is indeed talking about something that happens because of you, not the gun. As you point out, the guns are so accurate that in a vise, they can put every pellet through the same hole.
And I think you are correct in understanding his argument to mean that you should not adjust your sights to center one particular clump.
But his reasoning is a little different that yours, I think. I think you expect that your clump on some particular target might be displaced because of a particular problem in your technique that you might change on the next target. I think that's why you hypothesized the undesirable side effects you cited.
But I think Al is arguing purely probabilistically that you take a random sample of a random distribution, you'll often see a clump or at least what appears to be a center to the sample. Is that really the center of the distribution from which the sample was taken? The answer is, you don't know. The best you can do is calculate the probability that you've found the center +/- certain bounds. If you have a very large sample, your confidence is increased.
Consider just the simple problem of flipping a coin 5 times. Suppose you got 4 heads and 1 tails. Would you conclude that heads is 4x as likely as tails? Or would you take for granted (because you know coin-flipping is supposed to be 50:50) that you just need a larger sample?
Relating this back to the shooter, it's possible the shooter might display two distinct modes, one where he makes a specific error, one where he doesn't, and I think maybe you're thinking about that case. But even without any bimodal or multi-modal behavior, even if all that's happening is that the shooter is producing a large distribution about a single center, you'll still get the phenomenon Al is talking about, which is that if you take a small, randomly-chosen sample, you might find that all the shots form a nice tight clump, purely because it's a random phenomenon. And if you get one of those clumps, and then adjust the sights to center it, you could be wrong -- not because your behavior on the next target will be different, but because it will likely be the same, meaning chosen randomly from the same random distribution.
Exactly- so the issue is, "How do we rule out (pseudo random) clumping, implying the correct action would be 'Do Nothing'?"
First, my initial answer assumed a "process basically in control" or a shooter who can generally place an infinite number of shots inside the 8 ring, the vast majority of shots inside the 9 ring, and a pretty extensive number of shots inside the ten ring. And the process is centered on the center of the target.
Back to clumping for those who are still reading.
For a discrete distribution (coin flipping) the likelihood of "streaking" (clumping) can be calculated, and we see that the probability of getting 2, 3, or 4 heads in a row is not that big a deal.
For the XY continuous distribution however (pseudo random) clumping is much less likely, and the issue is complicated as the probability of two shots falling "together" is much more likely near the true underlying centroid of the distribution of shots- and much less likely as the "clump" forms away from the centroid. Even the newest newbie has experienced a couple of shots falling onto the same spot on the target (frequently somewhat removed from the center, at that!).
And the actual "likelihood" of a "clump" forming depends on how widely dispersed a group of shots you consider a "clump" vs. how widely dispersed your overall group is.
[Note: lawyers, judges and juries struggle with this issue daily in this country; as someone is always trying to "prove" that some "unexpected" occurence of cancers was "caused" by proximity to power lines, source of drinking water, etc. Billions of dollars have been extorted out of various industries in this country through the volatile mix of sympathy, junk science, and seemingly non-random events. But I digress (what a shock!)]
We have all experienced the phenomenon of clustering- indeed, we are trying to cluster our shots on the centroid in the first place!
The problem is the cluster that forms *away* from the centroid. A cluster of three shots at 9:00 in the tight-nine ring area is probably pseudo random clustering- even if it occurs "in a row" (as opposed to "out of 100 shots." Well, if the clump forms out of shots "in a row" that also lowers the likelihood that it was pseudo random clumping).
A cluster of six tight shots at 9:00 in the 8 ring, however, is very probably not pseudo random clustering . . . especially if it is "in a row," and arguably if it is "out of 100 shots."
Of course, this presumes a certain level of process variance- a "widely scattered cluster" of six shots out of 100 shots that are all "widely scattered" probably isn't going to tell you anything.
Anyhow
The process of delivering the shot isn't really "random" at all. That isn't to say it is "totally controllable" either- and the various physical factors that control the delivery of the shot can sure seem to be random at times!
So let's add another branch to our decision tree:
We Observe X shots out of Y to be falling near to each other, away from the process centroid.
1) It is random clustering, ignore it.
2) The sights are off, adjust them.
3) The process is off, adjust it.
Your Mileage May Vary as to how you make the decision to 1. recognize that a cluster has formed (definition of "X out of Y" and "near to each other"), and then 2. treat it with corrective action 1) vs. 3).
For me, FWIW, a cluster might be recognized when:
"X out of Y" = 3 out of 6 or so
"Near to Each Other" = hole edges within 1/2 pellet diameter
and the solution is always
3) refocus on proper technique, from the ground up
Action 3 may be only a placebo, but it generally doesn't hurt anything either . . .
Al made a point which I think may be helpful, Steve, which is that the concept of randomness is not generally well-understood in the lay population even though it affects pretty nearly everything around us. One common result of that misunderstanding is that most people attribute too much significance to seemingly "unusual" events. One of the most widely-discussed examples of this is the so-called Birthday problem which asks that you estimate the probability that if you assemble some number of randomly-chosen people in a room, that at least 2 of them will have the same birthday (or alternately stated, how many it would take for the probability to be greater than some percent.) Invariably, people underestimate the probability or overestimate the number of people required.
I think if you'd really like to understand how probability works, you probably need to do some reading or go take a course, which is certainly beyond the scope of a discussion on TargetTalk.
So instead, let me just focus on a couple points. First, natural events, including shots fired, are random. It's just a question of identifying the distribution from which they're chosen.
Second, pseudo-random has a particular meaning different than you imagine. It does not refer to clustering in a sample. What it refers to are sequences that "appear" to be random but which aren't really. The most common example is a computerized pseudo-random number generator: They're not really random because if you choose the same starting point (called a "seed") they will always generate the same sequence.
Third, the probability of getting any particular unique sample is the product of the probabilities of the individual events. If the true distribution of a shooter's shots is uniformly distributed in all directions about the center of the black, then the probability of getting a sample of 3 shots that all landed on top of each other on the 7 ring at 3 o'clock is exactly the same as getting a sample with one there, one at 7 o'clock on the 7 ring and one at 11 o'clock on the 7 ring. (Yes, I know this completely counter-intuitive.) But this is an example of why I said that statisticians avoid claiming that a sample, even a large one, tells them the characteristics of the distribution from it was taken. Instead, they'll calculate the probability they know those characteristics.
Well, my previous 2 answers didn't go through so here's the "Reader's Digest" version:
1. There is a difference between believing that
a. underlying natural processes are random; and
b. it is useful to model what we can't yet understand at a deterministic level as "random" (ie the underlying meaning of the "error term" at the end of every natural "law")
2. I agree with your (commonly accepted) definition of "Pseudo Random." That's why I used it. That was my point.
3. I appreciate your suggestion to "study up on" Prob/Stat. Tha one really made my day! =8^)
Steve "Acolyte of Kuhn, Popper, and Determinism In General" Swartz
Let me see if I can put things somewhat back into perspective. This thread started over a year ago when a relatively new shooter, who probably averages in the 470's, based on his assertion that his personal best was 490, posted a target with a fairly tight, 20-shot group whose center was roughly 1/3 of an inch high and to the right of the center of the target. A lengthy discussion ensued regarding whether or not the shooter should adjust his sights.
As I pointed out to 'Guest' a few posts back (he claims a personal best of 575, which suggests that he averages close to 570), if he had shot that group, then of course he should move his sights. On the other hand, if a merely average shooter, such as myself, had shot that group, then he should leave his sights alone. The reason for this is that for an average shooter, a 1/3 inch radial displacement in the center of a 20-shot group is well within 1 standard deviation from the center, and is most likely due to random variation, rather than some enduring problem with the shooter's technique.
One of the difficulties most people have when discussing accuracy is that they don't really understand randomness and dispersion. Your air pistol has a certain amount of innacuracy associated with it. It's so small that you can barely detect it. You and your pistol, as a system, also have a certain amount of innaccuracy. Try as you might, you just can't put all 60 shots through the same hole. No amount of 'control' that you can exert is going to make you a 600 shooter. The variation in your shot placement from shot to shot is basically random (assuming that you shoot circular groups).
The amount of variation in radial error in shot placement, group size and group center is different for each shooter/gun combination. 570-shooters have much less variance than 520-shooters -- in shot placement, in group size, and in radial error in group center.
It's important to understand this when deciding how to proceed during a match, and what advise to give to another shooter. Because the exact same target can mean something completely different when you shoot it vs. when I shoot it, particularly when we are talking about a relatively small number of shots. The fact is, I shoot much larger groups than you do (I believe you wrote in another thread that you used to average 570). 3 eights in a row at 9 o'clock just doesn't have any significance at all when I do it.
Still, a lot of people will change their sights anyway, even when they shouldn't, because that don't understand randomness and dispersion, and are therefore likely to lose confidence in themselves and their equipment a lot more quickly than someone who does.
As far as clumping is concerned, it's a fundamental characteristic of randomness. Just look at the night sky sometime. The stars aren't uniformly distributed in the sky. Neither are the grains silver oxide on a photographic plate, or the hairs on your head. They are all random phenomena, and they all exhibit 'clumping'. So too do the shots in a shot group, although the distribution is different.
For the mathematically inclined, the concept of a 'confidence interval' might be helpful. I.e., if I shoot an X-shot group, its center will be within some distance 'd' of the center of the target 95% of the time, for example.
The distance 'd' would be given by:
d = (Zvalue * SD) / sqrt(n)
Where the Zvalue corresponds to the degree of confidence that we want, assuming a 1-sided confidence interval. For a 1-sided 95% confidence interval, the Zvalue = 2.132 (assuming that shots are normally distributed in 2 dimensions, we use a T-distribution with 4 degrees of freedom).
SD is the standard deviation. If you put 95% of your shots inside the 9-ring (a 1 inch diameter circle), then 1 SD will be 1/2 the radius, or 0.25 inches (since 2 SDs is equal to the radius).
n is the number of shots in your group.
So:
d = 2.132 * 0.25 / sqrt(n)
For a 5-shot group:
d = 0.24 inches
In other words, for a 570-shooter, who places 57 out of 60 shots in the 9-ring, we can say with 95% confidence that if he shoots a 5-shot group centered more than 0.24 inches away from the center of the target, he should move his sights.
For a 500-shooter, who puts 95% of his shots inside the 6-ring (a 3 inch diameter circle):
d = 0.7 inches
In other words, we can say with 95% confidence that a 500-shooter who shoots a 5-shot group centered in the middle of the 8-ring should change his sights.
In the case of the original poster, he should leave his sights alone. In the case of 'Guest', who averages around 570, he should change his sights.
Okay, I promise this is my last post on this thread (again). Two parts: Part I for everyone (Ithink it's an audience of 3 at this point) and Part II for the inquisitive.
Part I: Adjusting Your Sights
What cracks me up is that we agree that you shouldn't adjust your sights!
- Al says "Don't adjust your sights because the clump is 'random' [sic]"
- Steve says "Don't adjust your sights because the clump isn't 'random' [sic]"
So we agree- you should only adjust your sights when you need to "move the centroid of a suitably large number of shots to coincide with the desired impact area."
So much for the important part- now for the triviata.
Part II: Random vs. Non-Random Clumping
I spent a bit of time in my misspent youth larnin about that there clumpin stuff and sech.
The issue- if you will stay with me here for a moment- is to separate the "random" "clumping" from the "non random" "clumping.
Back in the late 70s early 80s I had a couple of work-study gigs while at the University of Arizona. One of which involved sysadmin stuff for the structures lab; babysitting the Graphical Integrated Finite Transform Structures System (yeah, to satisfy the NASA sponsor, we needed a nifty acronym). That job involved modeling structures and then simulating breaking them. But that is irrelevnat to the current discussion.
The other work study job was planet hunting for the astronomy department. Back in the day we used Speckle Interferometry to look for reflective bodies orbiting suns. End result we found a bunch of Jovian and Near-Jovian candidates (some of which have been validated with more modern methods) using the technique. My job was to develop and manage the software used to find the planets. I didn't write the original program (written in "FORTH" by a grad student who had left the lab) but was responsible for upgrades, bug fixes, etc.
We peered at distant suns and digitized the images falling on out photoreceptor plate. Basically a point source of light (the distant sun) would travel through the atmosphere and become a series of points on the plate over a time period. You collect a series of point in two-dimensional space and can draw inferences about the light source by analyzing the cloud of points.
Sound familiar?
Our job was to separate the "random" [sic] effects of atmospheric distortion from the "non-random" point source and reflected source(s) origins.
If you are still interested, here's how we did it.
The photoreceptor data consisted of a bunch of points. The points showed "clumping." Our job was to analyze the clumps in order to remove atmospheric distortion.
1. We assumed atmospheric distortion would be uniformly/randomly distributed acoss the plate
2. We assumed the point source(s) would not follow a uniform/random distribution across the plate
Two approaches:
1. Model the random effect of atmosphere as a normal (gaussian) random variate away from the mean in both the X and Y axis. On a histogram, we would see a big spike over the underlying "True" (point source) value for the X axis; and the same thing for the Y axis. A deviation from this pattern; eg a secondary "hump" (bimodalism) in the data would suggest another "True" light source (reflecting planetary companion) as a companion to the primary light source (star). Of course, the selection of an X axis and Y axis are somewhat arbitrary(!) so we fooled around with rotating the image through 90 degrees etc. hocus pocus to pick up more subtle effects; ended up going to method 2 . . .
2. Model the data in global coordinates, assume the random dispersion would be uniformly/randomly distributed away from the centroid in all vectors. On a histogram, the "True" value represented the max tail of an exponential distribution anchored on the centroid. Rotate the data through 360 degrees and look for deviation from the exponential distribution (again, looking for bimodalism & a secondary "spike"), which would represent a second "True" ppoint source of reflected planetary light.
Worked like a champ. Every morning, I would download the data and process it (why we used "FORTH" is a story in and of itself). Any deviation from the normal/exponential distribution would be flagged for further analysis. Every day the lab would analyze the data and refine the target/candidate list of stars. Some nights we went back and took more data. While I was there we looked at a buttload of stars and found quite a few candidates.
The moral of the story: read up on Speckle Interferometry (in ABI/INFORM, EbscoHost, or some such) and you will find some satisfactory algorithms for teasing out the non-random clumps from the random clumps.
Sounds like you are somewhat of a hobbyist in this regard and you may find that what we used analytically back in the 80s was somewhat unsophisticated compared to what you could do in a simple stats program like SAS or SPSS today.
Just digitize your shots either in global coordinates or XY coordinates (one or the other, then convert) and have at it!
If you want to chat more, let's do it offline 'cause I wouldn't want to waste any more precious bandwidth!
- The point sources ("True" value) would be represented by, well, a singularity
- The atmospheric distortion was represented by a uniform distribution
- The *combined* distribution would then be a normal/exponential
What we actually found was that while a normal (cartesian) or exponential (global) distribution was a good "model" for the purpose of identifying a second true source, the actual distributions in practice were somewhat different.
What cracks me up is that we agree that you shouldn't adjust your sights!
- Al says "Don't adjust your sights because the clump is 'random' [sic]"
- Steve says "Don't adjust your sights because the clump isn't 'random' [sic]"
So we agree- you should only adjust your sights when you need to "move the centroid of a suitably large number of shots to coincide with the desired impact area."
...
Sounds like you are somewhat of a hobbyist in this regard ...
Steve
Actually, Steve, I'm a systems engineer, and I'm currently doing performance engineering fo a living, so this kind of thing is more than just a 'hobby' to me.
At least we agree that you shouldn't move your sights unless the center of your shot group is sufficiently far off center -- how far off center is the key issue.
Your mention of 'Forth' takes me back a few years -- I knew Chuck Moore, the creator of 'Forth', back when his company did a contract for us back in the late 70's.
By the way, if you were involved in astronomy 25 years ago, you should know that stars, or 'point sources' as you referred to them, are Poisson-distributed in 2 dimensions in the night sky. As such, they exhibit a phenomenon known as 'Poisson clumping'. Shots on a target follow a normal or Gaussian distribution in 2 dimensions. They also demonstrate random 'clumping'.
My point was that in order to distinguish between random 'clumping' and non-random effects, you need to have a large enough sample set to draw a statistical inference for a shot group located a specified distance from the center of the target. This necessarily involves an understanding of the degree of accuracy of the shooter.
You seemed to be saying the same thing, but you appeared to be making the distinction based on your own degree if accuracy, and the rules-of-thumb that you've developed over the years, intuitively, based on the way that you shoot.
I made the calculations that I did in order to demonstrate the effect of shooter accuracy in this regard. I'm not entirely sure of the validity of using a t-distribution for this purpose. But then, I don't much like using the t-distribution anyway. After all, it is what it is, i.e., a way to try to draw a statistical inference from a small sample set.
However, the numerical results that I got seem to be consistent with your rules-of-thumb and my own experience. So, I'll agree to agree if you will.
I want my shots to "clump" in the ten ring.
One time a few years ago I got my pistol back from my pistolsmith on the day of the match. It was zeroed nicely prior to sending it out for the trigger work, but due to circumstances, I had no chance to test it before the match. At 50 yards I had 10 shots that all counted. First shot was a clump of one way low and left. So do I confirm with a second shot? No. I cranked in a bunch of up and right, and fired a second shot. A little high and still some left. Another sight adjustment and I found the edge of the ten ring on the third shot. A little more right and I clumped the remaining shots in the tenring (and a little 9 ring). No science involved.
This thread is making me not want to log into anymore.
Come on, If you want to be a brainiac or prove you are smart you have done it, however, what does it matter if you move your sights or not.
It should be a personal issue, not a scientific, random calculation or algorithm.
You have good technique (execution) and can call your shots going somewhere other than called then move your sight accordingly.
You are not so good caller, and your technique is less than perfect, leave your sights alone.
They are called ADJUSTABLE sights because they are there for you to adjust them. Otherwise they would be called FIXED sights.
Everybody will at one time or another need a sight adjustment and if you need to do it during a match then do it!
Don't wait until you have mastered ZEN or have taken 20,000 samples to make a decision.
come on people wrote:
They are called ADJUSTABLE sights because they are there for you to adjust them. Otherwise they would be called FIXED sights.
I burst out laughing when I read this. Unfortunately, I was in the midst of giving my students a final exam. I could use a good "happy laugh" during finals week.
Steve, in the example you suggested, you're separating signal from noise based on the knowledge of their governing distributions, which I don't think is applicable in analyzing the distribution of shots (well, actually we have to look at it as time series but whatever), as we have no separable sources of signal and noise. Instead, we have the same source of error but with varying parameters:
We have a shooter, and at any given time we can model the random variable of shot placement with some probability distribution (something like a 2D normal distribution, but not quite because triggering errors will fatten up the tail significantly). The main problem is that the distribution parameters, the centroid and variance, are variable (making a reference to someone mentioning non-stable system in the beginning of the thread).
We may start with a centroid at zero, and some variance, then get jittery and start throwing the shots all over the place (centroid is still zero, but high variance), then we get tired and our NPA shifts, resulting in change of centroid, etc.
Now lets say we've shot a group away from the center, and we're trying to decide whether to adjust the sites. If the artifact is due to increased variance, we need to get our act together. If we think that our variance is small enough then the change in sights is warranted. (I understand that you can try looking for the source of deviation in centroid, etc, but if the error in technique is not immediately apparent, then why not change the sights and keep doing what you were doing? If you're grouping well you must be doing something right!)
Now, how do we know if the artifact is due to increased variance or shift in centroid, or even a combination of both? We may draw some weak inferences by analyzing thousands of shots, massaging the data post mortem in SPSS and what not, but a much better and more valid answer
is not on the target, but inside the shooter. Is your technique OK? Sights stay still, good folowthrough and you called good shots? Last 5 well-executed shots produced a good-sized group at least half a group from the center? (your rule of thumb may vary) - Adjust the sites without a second thought. If you can't positively say your technique was OK you weren't paying attention; in that case do two more shots while observing your technique and the answer should be clear.
Now before you go "but but but.. how do you KNOW that the group did not end up there by random error?" The whole damn point is that you would have/should have SEEN the errors in your technique with your own eyes and other senses!
The bottom line:
a) If your technique is good but your grouping is off, adjust the sites.
b) If your technique is not good, pull your act together; see a)
