Jump to content
You need to play a total of 5 battles to post in this section.
Pseudovector

Another visualization of the changes to AA in 8.5

36 comments in this topic

Recommended Posts

Members
182 posts
2,817 battles

I've seen a few graphs plotting plane kills vs. damage, but those looked like qualitative graphs or I couldn't tell where the data came from.
https://forum.worldofwarships.com/topic/183885-why-aa-sucks-for-everyone-and-how-to-fix-it/
https://www.reddit.com/r/WorldOfWarships/comments/bv3h3x/the_problem_with_continuous_aa_its_math/
https://forum.worldofwarships.com/topic/194693-notser-85-analysis-of-aa/?page=3&tab=comments#comment-4565492

Wanting to see a quantitative version, I wrote a simulation to produce such a graph.

Figure_n10000.png

  • Cool 3

Share this post


Link to post
Share on other sites
708
[HELLS]
Members
2,497 posts
24,838 battles

If I read this correctly, and based on play experience in CVs since the update, the current setup buffs AA and nerfs CV play compared to the other two methods. My impression that AA was reinforced at all tiers is not wrong then....

Share this post


Link to post
Share on other sites
1,801
[GWG]
Members
6,690 posts

So basically, there is no longer an advantage to hitting the return button after one run.   Planes are lost regardless. 

New tactics for cheating the system need to be developed.

Share this post


Link to post
Share on other sites
601
[PISD]
Members
950 posts
4,707 battles
24 minutes ago, AVR_Project said:

So basically, there is no longer an advantage to hitting the return button after one run.   Planes are lost regardless. 

New tactics for cheating the system need to be developed.

Flak still exist and still hurt all the planes.

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
On 6/30/2019 at 5:35 PM, GrandAdmiral_2016 said:

If I read this correctly, and based on play experience in CVs since the update, the current setup buffs AA and nerfs CV play compared to the other two methods. My impression that AA was reinforced at all tiers is not wrong then....

It's definitely a buff to AA; planes losses start much sooner than before.  Even if you don't take less damage, the CV will be depleting reserves faster, unless they were losing whole squads, then there's no difference.

The 2 curves labeled "target random plane" both represent AA in 8.4, but are different ways of averaging the data.
One averages the amount of damage done to the squadron when each plane was shot down.
The other averages the number of planes shot down after a given amount of damage had been done to the squadron.

On 6/30/2019 at 6:39 PM, AVR_Project said:

So basically, there is no longer an advantage to hitting the return button after one run.   Planes are lost regardless. 

This says nothing about the return button.  What it does show is that planes start being shot down sooner, so there may be more losses by the time the attack run finishes.  Any planes that survive that can still be recalled.  If you had at least one plane to recall in 8.4, in an identical situation in 8.5, you'd still have at least one plane to recall.  The number of planes might be lower if you had more than one, but this change didn't make the button pointless.

Share this post


Link to post
Share on other sites
1,801
[GWG]
Members
6,690 posts
6 hours ago, Pseudovector said:

This says nothing about the return button.  What it does show is that planes start being shot down sooner, so there may be more losses by the time the attack run finishes.  Any planes that survive that can still be recalled.  If you had at least one plane to recall in 8.4, in an identical situation in 8.5, you'd still have at least one plane to recall.  The number of planes might be lower if you had more than one, but this change didn't make the button pointless.

Situation is that often (under 8.4), you'd attack a high DPS target, score a hit, then after being clear of the AA, look down and see all 6 remaining planes are in the red.  I'd be stupid if I didn't return them right away and prevent any from being lost.

In an identical situation (under 8.5), I'd look down and see my reserves completely wiped out.  The only survivors would be the planes I attacked with.

So you say, 'Don't attack high DPS targets'...  Yeah, right !   Like I can control MM to where my Shokaku is not facing a wall of Des Moines and Worchesters.

You deal with what you get.  In that scenario, leave the planes parked and charge in for secondary damage.  Planes are expensive.

  • Cool 2

Share this post


Link to post
Share on other sites
Members
627 posts
8,271 battles
On 6/30/2019 at 8:39 PM, AVR_Project said:

So basically, there is no longer an advantage to hitting the return button after one run.   Planes are lost regardless. 

New tactics for cheating the system need to be developed.

I've been testing the 8.5 changes myself (I always try the new system so as to better defeat it). I've been dropping two loads of torps into islands early in the game. I'll make the first runs with only 4 planes instead of 8. Same with DBs. I get one drop off, and there are none left for them to shoot down after that. It seems to be working.

 

Rocket planes are so frail, in general, that I use the entire squad on those attacks. But I don't go to the rocket planes until much later in the game.

 

Seems to be working pretty well, really.  I had one game yesterday with 4 kills using Shokaku. And I've had some relevance when bottom-tiered as well. 

 

I can't say that my win-rate is particularly good with CVs since 8.5, but I've played a lot of Shokaku and she is a particularly cursed ship. 100k and 3 or 4 kills and she still loses. Lex has done well, and Big E and Kaga have done very well. Haven't tried Ryujo and Ranger yet.

Share this post


Link to post
Share on other sites
332
[GRETA]
[GRETA]
Members
580 posts
16,507 battles

Hakuryu 8km torpedo bomber variant can circumvent the AA constant damage by dropping from 5-6km away and pulling out of the AA bubble quickly. Works very well against BBs that grouped up for mutual AA defense, provided that you have the proper lead and some advanced knowledge about the torpedo spread patterns. Having TA removes stealth drop but massively increases the effectiveness of the quad drop pattern.

As a general rule, sandwich your target between an island and your planes. This forces the target to turn in a single direction and makes it more susceptible to follow up strikes. Once your torps force him out of position and separate him from his allies, you can use AP dive bombers to finish them off, AA ship or not.

As for the lower tier CVs, go for the lower tier targets first. Conserve your planes as much as you can early on, even if it means not doing damage. Late game is where you shine.

The AA change only makes playing CV more challenging and feel that much more rewarding. :Smile_playing:

  • Boring 1

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles

I've seen the suggestion to distribute AA over an attacking group rather than one plane or a full squad.
There are 2 interpretations of this, both of which I've modelled.
One is that AA will focus on and kill one group, then move onto the next. eg. for a squad of 6, and an attacking group of 2, fire at planes 5 and 6 until both are shot down, then 3 and 4, ...
The other is that AA will be distributed over however many planes an attacking group has, regardless of how many planes are in the last group. eg. for the same numbers as above, fire at 5 and 6 until one of those is shot down, then shoot the remaining one and plane 4, ...
They are labelled "strict" and "fluid" in the graphs, respectively.

Figure_group_s_6.png

Figure_group_f_6.png

Share this post


Link to post
Share on other sites
35,138
[WG-CC]
WoWS Community Contributors
11,574 posts
9,401 battles

How many planes are you assuming to be in the squadron / attack flight for these graphs?  I'm assuming 6...

Edited by LittleWhiteMouse

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
15 minutes ago, LittleWhiteMouse said:

How many planes are you assuming to be in the squadron / attack flight for these graphs?  I'm assuming 6...

Yeah, 6 in the squadron, 2 in the attacking group, for now.  I was planning on running this for squad sizes actually in the game, but I hadn't gotten around to looking up the numbers yet.

Share this post


Link to post
Share on other sites
35,138
[WG-CC]
WoWS Community Contributors
11,574 posts
9,401 battles
1 hour ago, Pseudovector said:

Yeah, 6 in the squadron, 2 in the attacking group, for now.  I was planning on running this for squad sizes actually in the game, but I hadn't gotten around to looking up the numbers yet.

It will be an even more dramatic disparity in some of the larger 9-plane squadron sizes that are hella common at tiers 8+.

Share this post


Link to post
Share on other sites
1,353
[O-PN]
Members
2,573 posts
5,644 battles
1 hour ago, Pseudovector said:

I've seen the suggestion to distribute AA over an attacking group rather than one plane or a full squad.
There are 2 interpretations of this, both of which I've modelled.
One is that AA will focus on and kill one group, then move onto the next. eg. for a squad of 6, and an attacking group of 2, fire at planes 5 and 6 until both are shot down, then 3 and 4, ...
The other is that AA will be distributed over however many planes an attacking group has, regardless of how many planes are in the last group. eg. for the same numbers as above, fire at 5 and 6 until one of those is shot down, then shoot the remaining one and plane 4, ...
They are labelled "strict" and "fluid" in the graphs, respectively.

Figure_group_s_6.png

Figure_group_f_6.png

You have just put more analysis into AA than WG throughout the entire CV rework.

  • Cool 1

Share this post


Link to post
Share on other sites
1,353
[O-PN]
Members
2,573 posts
5,644 battles
Spoiler

 

You seem to have the system down -- would you model for...
- 9 aircraft
- damage is random to all 9 until x, then damage focus to last aircraft
- x1 = 50% of total aircraft 
-x2 = 60% of total aircraft health

Thus AA is random on all aircraft until total combined HP of aircraft = 50% then one at a time deletion begins, and another at 60% of total combined.  The point being to see if if one at a time deletion starting only after all aircraft health is diminished thus allowing both TB heal and AR cap'n skills to enable.

Graph should have 8.4 AA and 8.5 AA overlaid as well.

Edited by NoSoMo

Share this post


Link to post
Share on other sites
Members
78 posts
On 7/1/2019 at 1:12 AM, Pseudovector said:

I've seen a few graphs plotting plane kills vs. damage, but those looked like qualitative graphs or I couldn't tell where the data came from.
https://forum.worldofwarships.com/topic/183885-why-aa-sucks-for-everyone-and-how-to-fix-it/
https://www.reddit.com/r/WorldOfWarships/comments/bv3h3x/the_problem_with_continuous_aa_its_math/
https://forum.worldofwarships.com/topic/194693-notser-85-analysis-of-aa/?page=3&tab=comments#comment-4565492

Wanting to see a quantitative version, I wrote a simulation to produce such a graph.

Figure_n10000.png

 

Thanks for taking the time to make and and post this.

 

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
17 hours ago, LittleWhiteMouse said:

It will be an even more dramatic disparity in some of the larger 9-plane squadron sizes that are hella common at tiers 8+.

Now that I've looked at squad/flight sizes, I see that 9 plane squads are quite common.
I didn't realize when I started that 6:2 squad:flight sizes were also quite common; those were just convenient numbers for prototyping/testing code.
8:2 seems to be the next most common, so here are graphs for 9:3 and 8:2.

I'm thinking that the 12 plane squads on Midway and Hakuryu are probably worth doing too.
For torp bombers on Hakuryu, do people generally prefer 12:2 or 12:4, or are both common?

Sidenote: I note that the wiki was of no help, not having any of these numbers.  I looked briefly at the page source for the ship pages, and couldn't figure out how the ship stats tables are generated; I'm guessing it's in the template to do a lookup somewhere.

group_s_9_3.png

group_f_9_3.png

group_s_8_2.png

group_f_8_2.png

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
18 hours ago, NoSoMo said:

You seem to have the system down -- would you model for...
- 9 aircraft
- damage is random to all 9 until x, then damage focus to last aircraft
- x1 = 50% of total aircraft 
-x2 = 60% of total aircraft health

Thus AA is random on all aircraft until total combined HP of aircraft = 50% then one at a time deletion begins, and another at 60% of total combined.  The point being to see if if one at a time deletion starting only after all aircraft health is diminished thus allowing both TB heal and AR cap'n skills to enable.

Graph should have 8.4 AA and 8.5 AA overlaid as well. 

Interesting idea for preserving that functionality, and fairly easy to code.

The curve looks like it follows the random one until the threshold, then is fairly linear to the last plane.

I think the thresholds suggested might be too low, and goes back to having no planes killed until too late, like pre-8.5, but a higher one could be interesting middle ground.  How much HP does the heal give?

NoSoMo_50_9_3.png

NoSoMo_60_9_3.png

  • Cool 1

Share this post


Link to post
Share on other sites
1,353
[O-PN]
Members
2,573 posts
5,644 battles
17 hours ago, Pseudovector said:

NoSoMo_60_9_3.png

While the 60% might seem too much by simply looking at the graph, if you graph lines to what people would complain about like "only shooting down 2 planes".  In the 60%, when you draw a line from what used to be 2 plane kills to intersect my suggested fix, it's more than double the destruction rate of full random.  6 aircraft hangar pulls takes 7--9 minutes.

Share this post


Link to post
Share on other sites
18
[MUSK]
Members
45 posts
392 battles
On 7/2/2019 at 11:47 PM, Pseudovector said:

I've seen the suggestion to distribute AA over an attacking group rather than one plane or a full squad.
There are 2 interpretations of this, both of which I've modelled.
One is that AA will focus on and kill one group, then move onto the next. eg. for a squad of 6, and an attacking group of 2, fire at planes 5 and 6 until both are shot down, then 3 and 4, ...
The other is that AA will be distributed over however many planes an attacking group has, regardless of how many planes are in the last group. eg. for the same numbers as above, fire at 5 and 6 until one of those is shot down, then shoot the remaining one and plane 4, ...
They are labelled "strict" and "fluid" in the graphs, respectively.

Figure_group_s_6.png

Figure_group_f_6.png

How about random planes but weaker planes are more likely to be hit?

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
10 hours ago, NoSoMo said:

While the 60% might seem too much by simply looking at the graph, if you graph lines to what people would complain about like "only shooting down 2 planes".  In the 60%, when you draw a line from what used to be 2 plane kills to intersect my suggested fix, it's more than double the destruction rate of full random.  6 aircraft hangar pulls takes 7--9 minutes.

True, ships that were shooting down roughly 2 planes would shoot down more, but there are ships with weaker AA that were not able to shoot down even 1 plane pre-8.5, and those ships would go back to not being able to shoot down anything.  I don't think it would be good for those ships to be nerfed that much.

I plotted the 70% and 80% cases just for comparison.

NoSoMo_70_9_3.png

NoSoMo_80_9_3.png

Share this post


Link to post
Share on other sites
1,353
[O-PN]
Members
2,573 posts
5,644 battles
2 hours ago, Pseudovector said:

True, ships that were shooting down roughly 2 planes would shoot down more, but there are ships with weaker AA that were not able to shoot down even 1 plane pre-8.5, and those ships would go back to not being able to shoot down anything.  I don't think it would be good for those ships to be nerfed that much.

I plotted the 70% and 80% cases just for comparison.

Such is the issue with AA.  The difference between min and max AA levels is entirely to large and substantially magnified by overlapping.

The very FIRST step for the look at AA is "what is the least that should happen in an AA interaction, and what is the most"  All ships and their respective damage should have been put between those 2 points.  Worcester fully spec'd for AA in 8.4 was entirely too much being able to completely kill 9 aircraft w/ out an attack landing while at the same time some DDs could have been circled for 5 minutes not harming an aircraft.  WG is overtly ignorant on the issue.   They've done everything but come out and say "we don't know what we're doing".

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles
On 7/4/2019 at 2:13 PM, lucas1899 said:

How about random planes but weaker planes are more likely to be hit?

I've been thinking about this, but to simulate, this idea isn't precise enough.  You need to be able to describe mathematically, how much more likely a plane is to be hit given its HP.

One way this could be done is to make the probability of hitting a plane inversely proportional to its HP.  A plane with twice the HP of another plane is half as likely to be targeted.
I ran this simulation for squad sizes 6, 8, 9, and 12.

invHP_6.png

invHP_8.png

invHP_9.png

invHP_12.png

Share this post


Link to post
Share on other sites
Members
182 posts
2,817 battles

Weighting probabilities by the inverse of HP looks too close to the random curve to me.
The next idea I had was to use the inverse of the square of HP.
Again, simulated for squad sizes of 6, 8, 9, and 12.

invHP2_6.png

invHP2_8.png

invHP2_9.png

invHP2_12.png

Share this post


Link to post
Share on other sites
Members
777 posts
62 battles
On 7/1/2019 at 2:35 AM, GrandAdmiral_2016 said:

If I read this correctly, and based on play experience in CVs since the update, the current setup buffs AA and nerfs CV play compared to the other two methods. My impression that AA was reinforced at all tiers is not wrong then....

Indeed. Quite an extreme buff to be honest 

  • Cool 1

Share this post


Link to post
Share on other sites
221
[WTAF]
Members
503 posts
16,619 battles

Naturally. CVs have been getting nerfed on average every other "update" for over a year. Excluding, of course, the big nerf.

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×