Jump to content
You need to play a total of 50 battles to post in this section.
gopher31

The difference Sigma makes.

24 comments in this topic

Recommended Posts

[SHAD]
Players
2,470 posts
11,067 battles

The general consensus is that changing sigma makes a small difference only.

But.....

I have noticed  that between tier 8 and 9 cruiser accuracy takes a marked step up.

It's very clear to me that this is the case, could a sigma increase from the T8 2.0 to the T9 2.05 be the reason.

 

I've been comparing the Kutuzov (S 2.0) and Irian (S 2.05) apart from the sigma and range they are identical guns with identical dispersion. Is the Sigma the difference here?

 

I wonder if anyone else has noticed this?

 

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles

No I notice it across the board:

Chapayev to Donskoi

Hipper to Roon

all have the aiming mod.

 

the IJN line is far less pronounced but is still difference.

 

 

Share this post


Link to post
Share on other sites
[TTTX]
[TTTX]
Players
4,608 posts
8,081 battles
3 hours ago, gopher31 said:

The general consensus is that changing sigma makes a small difference only.

no it is not.

3 hours ago, gopher31 said:

I have noticed  that between tier 8 and 9 cruiser accuracy takes a marked step up.

no it does not.

  • Cool 1

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles
Just now, Tyrendian89 said:

no it is not.

no it does not.

So you are saying that Sigma makes a big difference?

Yet the sigma change between T8 and T9 does not improve accuracy?

Share this post


Link to post
Share on other sites
[ADRIA]
[ADRIA]
Players
5,061 posts
8,562 battles

Sigma can be anything between useless change and massive difference depending on what and how you compare

 

In a very tight dispersion ellipse you won’t miss much even with very low sigma

Similarly your sigma can be really high, if your dispersion area is massive you’ll still get lots and lots of misses

 

Identical sigma can have different impacts on ships accuracy if their dispersion isn’t identical. So you can compare ships like Kii vs Amagi (same dispersion), but saying that NC is more accurate than Amagi just bcuz it has better sigma - doesn’t work like that

 

Since sigma buffs / nerfs for ships don’t change their dispersion you can compare the pre and post version. In a case like this (I think it was at a range of 15km) difference of “hits out of the same number of shells fired” would be something along the lines of (keep in mind these numbers might be a bit wrong)

  • 0.1s -> less than 1%
  • 0.2s -> about 1...1.5%
  • 0.3s -> About 2.5...3%
  • 0.4s -> about 5%
  • 0.5s -> about 7%
  • 0.6s -> about 10%

So you can see it matters very little for any single game, but can matter in the long run. Same as fire chance

 

And that’s before you start looking at players themselves...

  • Cool 2

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles

Well I have seen LittleWhiteMouse’s comparison of Massachusetts (1.7 sigma) and Alabama (1.9 sigma) here:

Ioq9pyF.gif

 

The difference looks small but is 4 times the difference between 2.0 and 2.05.

unless there is something significant about sigma reaching 2.0.

 

I would like to understand what is going on but I note that no one has yet said that they observe what I observe...

Share this post


Link to post
Share on other sites
[COMFY]
Players
792 posts
2,025 battles

I say part of the dispersion thing is also affected by hidden dispersion fomulas.

one example is Yamato vs musashi sigma diffence, also done by LWM.

0NK79Y8.gif

Yamato's 2.1 sigma dispersion (orange) vs Musashi's 1.8 sigma (red).  180 shells fired at 15km, locked onto a Fuso.  Both ships are using Aiming Systems Modification 1 against an a target without camouflage.

You can see that past 2.0, signal makes noticably less difference imo. 

  • Cool 1

Share this post


Link to post
Share on other sites
[ADRIA]
[ADRIA]
Players
5,061 posts
8,562 battles
2 hours ago, gopher31 said:

Well I have seen LittleWhiteMouse’s comparison of Massachusetts (1.7 sigma) and Alabama (1.9 sigma) here:

 

The difference looks small but is 4 times the difference between 2.0 and 2.05.

unless there is something significant about sigma reaching 2.0.

You'd have to compare the "density of shells" at certain "radii" (altho it's an ellipse so that's not exactly the correct term) from the center to see that.

 

You can see how similar the central area is, as the "radius" there is small and thus it's easy to see / compare both. But at further distances from the center you may have identical number of shells spread visually differently, and so it "feels" like it's different when it's not.

 

Also, as I mentioned previously - 0.2s difference would be around 1...1.5% difference in accuracy there. That means that with 180 shells fired the difference would be... 2...3 shells less on target.

 

And 1.7 vs 1.9 compared to 2.0 vs 2.05 isn't x4 difference. It's actually more than that, as sigma doesn't scale linearly. The larger the gap between your 2 points, the more rapidly the difference will grow. Think: exponential function / growth.

 

And there's nothing particularly special about sigma being exactly 2.0 or passing it. It's simply generally regarded as accurate because of how relatively high it is in this game.

 

2 hours ago, gopher31 said:

I would like to understand what is going on but I note that no one has yet said that they observe what I observe...

All accuracy values work together - both dispersion (short (horizontal) and long (vertical) axis of the ellipse) and sigma. You can not look at one of them and just call that the defining factor. That's not how it works. Any of them can be anything between super important and completely irrelevant.

And even then there would be more points to add like impact angle and so on, which again can affect your chances of hitting even when your dispersion has already been set in stone

  • Cool 1

Share this post


Link to post
Share on other sites
Players
3,220 posts
2 hours ago, gopher31 said:

Well I have seen LittleWhiteMouse’s comparison of Massachusetts (1.7 sigma) and Alabama (1.9 sigma) here:

Ioq9pyF.gif

 

The difference looks small but is 4 times the difference between 2.0 and 2.05.

unless there is something significant about sigma reaching 2.0.

 

I would like to understand what is going on but I note that no one has yet said that they observe what I observe...

How does one lock onto a target?

Share this post


Link to post
Share on other sites
Moderator, Players
1,208 posts
8,496 battles

Sigma and dispersion are two different things. Dispersion is how far away your shell can land away from where you are aiming. Sigma basically influences the probability of how often your shell is going to miss by however much. Higher sigma increases your chances of seeing similar dispersion, so it affects shell grouping as a result. It's why a ship like Yamato is more accurate than Musashi (Keep in mind Yamato has 275m and 26.63km shell dispersion/range and Musashi 274m and 26.5km). But because of Yamato's higher sigma, the salvo spreads are more repetitive while Musashi is more erratic.

 

Ideally, you want both good dispersion and good sigma. And the best BB at that is the Yamato.

 

Now, is there a big step up between T8 and T9? In one sense not really, but T9 unlocks modules that can decrease dispersion and therefore you can stand a better chance of getting hits. But to get a good picture, you would have to compare maximum dispersion, sigma and range. And I am not in the mood to make any graphs at the moment. But from having been playing for more than 3 years and having grinded through every single tech tree to T10, I feel I can say confidently that there is not a big difference in accuracy between T8 and T9.

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles
1 hour ago, wilkatis_LV said:

You'd have to compare the "density of shells" at certain "radii" (altho it's an ellipse so that's not exactly the correct term) from the center to see that.

 

You can see how similar the central area is, as the "radius" there is small and thus it's easy to see / compare both. But at further distances from the center you may have identical number of shells spread visually differently, and so it "feels" like it's different when it's not.

 

Also, as I mentioned previously - 0.2s difference would be around 1...1.5% difference in accuracy there. That means that with 180 shells fired the difference would be... 2...3 shells less on target.

 

And 1.7 vs 1.9 compared to 2.0 vs 2.05 isn't x4 difference. It's actually more than that, as sigma doesn't scale linearly. The larger the gap between your 2 points, the more rapidly the difference will grow. Think: exponential function / growth.

 

And there's nothing particularly special about sigma being exactly 2.0 or passing it. It's simply generally regarded as accurate because of how relatively high it is in this game.

 

All accuracy values work together - both dispersion (short (horizontal) and long (vertical) axis of the ellipse) and sigma. You can not look at one of them and just call that the defining factor. That's not how it works. Any of them can be anything between super important and completely irrelevant.

And even then there would be more points to add like impact angle and so on, which again can affect your chances of hitting even when your dispersion has already been set in stone

 

Thank you for your input, it's very valuable even if it  doesn't give me a reason  for the patterns I see.

 

Has Wargaming ever set out how they work out shell groupings?

 

I have seen this post: 

A bit old now. Informative but not from the devs so I'm unsure of it's accuracy.

I'm not really sure I have a clear idea of how sigma values affect shell groupings  other than more = better.

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles
4 hours ago, howardxu_23 said:

I say part of the dispersion thing is also affected by hidden dispersion fomulas.

one example is Yamato vs musashi sigma diffence, also done by LWM.

0NK79Y8.gif

Yamato's 2.1 sigma dispersion (orange) vs Musashi's 1.8 sigma (red).  180 shells fired at 15km, locked onto a Fuso.  Both ships are using Aiming Systems Modification 1 against an a target without camouflage.

You can see that past 2.0, signal makes noticably less difference imo. 

That is very clear.

Share this post


Link to post
Share on other sites
[-RSN-]
Players
13 posts
6,623 battles

I will try to explain it as simply as I could. In statistics, sigma represents "standard deviation". Sigma of 1 would mean that there is equal chance that the shell will land at the center and at the edge. As the sigma goes up, there is higher chance to get shells closer to the middle. How much higher? We do not know the equation used by WG to make the call.

 

In my experience dispersion plays much bigger role. I find my musashi to be more accurate than german BBs even though both have 1.8 sigma. It is just that the dispersion of the IJN BBs is much better. Similarly, on Missouri, I can equip the -11% dispersion module (available to US BBs for T9/10). The guns became substantially more accurate.

 

I remember iChase mentioned somewhere that the difference between Alabama (1.9 sigma) and NC (2.0 sigma) was around 1%. I do not know where he took that number from but frankly, I can hardly see any difference when using both ships. In contrast, when I changed my Bismarck to standard BB build instead of secondary (changing the 3rd slot module for hte dispersion one -7% dispersion), the difference was tangible.

 

Lastly, it is all chance in the end. Also your personal view is a bias that will shift your perception in one direction or another.

  • Cool 1

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles
6 minutes ago, Tegli4 said:

I will try to explain it as simply as I could. In statistics, sigma represents "standard deviation". Sigma of 1 would mean that there is equal chance that the shell will land at the center and at the edge. As the sigma goes up, there is higher chance to get shells closer to the middle. How much higher? We do not know the equation used by WG to make the call.

 

In my experience dispersion plays much bigger role. I find my musashi to be more accurate than german BBs even though both have 1.8 sigma. It is just that the dispersion of the IJN BBs is much better. Similarly, on Missouri, I can equip the -11% dispersion module (available to US BBs for T9/10). The guns became substantially more accurate.

 

I remember iChase mentioned somewhere that the difference between Alabama (1.9 sigma) and NC (2.0 sigma) was around 1%. I do not know where he took that number from but frankly, I can hardly see any difference when using both ships. In contrast, when I changed my Bismarck to standard BB build instead of secondary (changing the 3rd slot module for hte dispersion one -7% dispersion), the difference was tangible.

 

Lastly, it is all chance in the end. Also your personal view is a bias that will shift your perception in one direction or another.

That’s the value I think I need.

 

I’m remaining open minded  about my perception. Although the difference  to me seems very obvious, not one person has said they’ve noticed the same thing.

I cannot ignore this.

Share this post


Link to post
Share on other sites
[ADRIA]
[ADRIA]
Players
5,061 posts
8,562 battles
25 minutes ago, gopher31 said:

even if it  doesn't give me a reason  for the patterns I see.

Humans see patterns where none exist. That's how it has been and will be.

 

27 minutes ago, gopher31 said:

I have seen this post: 

A bit old now. Informative but not from the devs so I'm unsure of it's accuracy.

Practically everything we know about dispersion comes from players testing & studying it. Something not coming from devs doesn't mean it's wrong

 

However that thread is a 50:50, as some things are correct, some aren't. Like that 2nd picture "Current vs Realistic dispersion model" is wrong. The realistic is the one we have - vertical dispersion (along the line of fire) is much much larger than horizontal (left-to-right)

And Looking at 1st pic - while horizontal dispersion indeed expands with range, the same can not be guaranteed for vertical dispersion. Noone has done the tests to prove / disprove it, and there are arguments for both. So it's unknown

 

31 minutes ago, gopher31 said:

I'm not really sure I have a clear idea of how sigma values affect shell groupings  other than more = better. 

Sigma just shows how steep the normal distribution curve will be, a.k.a. the higher sigma, the more shells will land closer to the exact center point of your aim. Lower sigmas have higher chances for shells to go out to maximum possible distance, higher will give you tighter central groupings

You could put it like this - low sigma will be a shotgun spraying bullets everywhere, while high sigma will be much more consistent with its ability to hit.

 

And better - in general, yes. Any single ship would be better with higher sigma, as you're keeping the dispersion constant in this case. Meanwhile a cruiser may have a bit lower sigma, but it will still be significantly more accurate than a BB with higher sigma value due to cruisers much smaller dispersion ellipse.

 


 

Also what @Tegli4 said is pretty much spot on

  • Cool 1

Share this post


Link to post
Share on other sites
[-RSN-]
Players
13 posts
6,623 battles

Reverse engineering the formula by the results "is possible" if we actually get the results. For comparision, in classic World of Warcraft, the players were able to determine a lot of things about how the melee combat system had been working. But the big "but" is that they were actually getting exact numbers for hits/misses/glancing blows/etc. Here you get dots on the screen with no value of where exactly the shell had landed. Frankly, what you see by LittleWhiteMouse/Lert and the dispersion gifs is the most visual result you can hope for.

  • Cool 1

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles

I understand the concept of Sigma but I’m unsure how it scales for example at which number would all shells land on dead centre? Does such a point exist?

 

Share this post


Link to post
Share on other sites
[SHAD]
Players
2,470 posts
11,067 battles
Just now, Tegli4 said:

Reverse engineering the formula by the results "is possible" if we actually get the results. For comparision, in classic World of Warcraft, the players were able to determine a lot of things about how the melee combat system had been working. But the big "but" is that they were actually getting exact numbers for hits/misses/glancing blows/etc. Here you get dots on the screen with no value of where exactly the shell had landed. Frankly, what you see by LittleWhiteMouse/Lert and the dispersion gifs is the most visual result you can hope for.

That might be the case. 

 

Thank you all for helping with this.

Share this post


Link to post
Share on other sites
[-RSN-]
Players
13 posts
6,623 battles
2 minutes ago, gopher31 said:

I understand the concept of Sigma but I’m unsure how it scales for example at which number would all shells land on dead centre? Does such a point exist?

 

If your dispersion was 0, the sigma would not matter. If you have any dispersion different than 0, it will be mathematically impossible.

 

Joke aside, I will give you an example that might be closer to something useful.

 

Helena has 133 meters of horizontal dispersion (taken from the ingame tooltip). The length of the ship is 185.52 (wikipedia). If you aim in the middle of the broadside, all shells will be within the ship. The vertical dispersion is hidden though, so not all shells will actually land. Have in mind though that the ship is above water, so you are also aiming at a "projection" of hte ship on the 2D plane), so more shells will land than if you look at it from just 2D point of view.

Share this post


Link to post
Share on other sites
[NWP]
Beta Tester
1,767 posts
16,594 battles
3 hours ago, NoobySkooby said:

How does one lock onto a target? 

Move the mouse over the target and/or press X?

 

You can also shoot unlocked, but the dispersion will be even worse (like shooting into smoke etc).

Share this post


Link to post
Share on other sites
Players
80 posts
5,221 battles
1 hour ago, gopher31 said:

I understand the concept of Sigma but I’m unsure how it scales for example at which number would all shells land on dead centre? Does such a point exist?

 

Spoiler

8aa9ff808602c27f1d9d63d7b2c115388a34f190

In the spoiler is the propability density function for the normal distribution. This function shows you a function of propability. The μ defines the expected value and is in this case 0 for the middle of the dispersion circle. It is where the reversed parabel has its maxima(highest propability to hit is around center of circle).

 

In WoWs, when talking about sigma, we are not actually talking about just "sigma" but "value sigma". So we don't say sigma is 2.1, but dispersion has 2.1 sigma. This is used to determine a cutoff point.

Spoiler

gIEwijd.jpg

Normal distribution.World of Wasrhips seems to use use σ = 1 and μ = 0.

 

The percentages are the propabilities that a shell will land in the specific interval of the x axis(dependent on horizontal dispersion). If you integrate the whole function you will always get 1(Propability of the shell landing somewhere is always 100% of course).

 

Spoiler

r1R1wZ8.jpg

Normal distribution with cutoff of 1.8 sigma with 273m dispersion(Bismarck at max range).

Now here comes the kicker:

Because of the cutoff, the shells that would fly outside of the cutoff range will instead be equally distributed in the dispersion circle. That means higher sigma cutoff has a lower chance that shells will be randomly distributed with equal distribution and instead use normal distribution which is better to hit desired spots.

 

TLDR:

No, shells will never 100% hit the center point. Apart from the fact that WG seems to use a sigma of 1. Even if you give a ship infinitely big sigma cutoff. Your shells will still fall coresponding to the normal distribution. The shells will just never be equally distributed in the circle.

 

MfG Boom

  • Cool 1

Share this post


Link to post
Share on other sites
Players
3,220 posts
1 hour ago, Ubertron_X said:

Move the mouse over the target and/or press X?

 

You can also shoot unlocked, but the dispersion will be even worse (like shooting into smoke etc).

many thanks for that kind sir, your the first to tell me that so have a drink on me, lol

Share this post


Link to post
Share on other sites
[501L]
Players
381 posts
4,327 battles
8 hours ago, NoobySkooby said:

many thanks for that kind sir, your the first to tell me that so have a drink on me, lol

I will point out that the game will automatically do this most of the time but it can be annoying when it suddenly decides to lock onto a ship that was just passing infront of your target, that's when you need to make use of the X key.

  • Cool 1

Share this post


Link to post
Share on other sites

×