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

154 comments in this topic

Recommended Posts

7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

Are you ready for some GODDAMN SCIENCE?

 

Tashkent (I don't have Khabarovsk, but it serves for this test)

8% fire chance. 0 point commander (as my Tash commander has DE and would unfairly influence the results)

HcOEC9G.jpg

 

Shimakaze

7% fire chance. I hardly use my guns on Shimakaze and specialized for torpedoes and stealth so a stock commander wasn't necessary for the test.

ldhKA3P.jpg

 

Target ships: Yamato x12

Number of shell hits per test: 500~ (The times I went over didn't add to the fire count anyway)

 

Methodology: Shoot at specific "fire positions" on a target Yamato. When a section catches fire, ignore it and continue shooting at parts that are not burning. If the entire ship is ablaze, shift to a different Yamato until the fires run out. I tracked fires started in blocks of 100, for 500 shell hits over 2 tests-- a total number of shell hits of 1000 for Tashkent and 1000 for Shimakaze. Missed shells were not counted at all.

 

As Lert so generously already proved, HE shells can and will start fires regardless of if they penetrate or shatter, so the exact placement of shells did not matter for the course of the test. Only that I avoided shooting sections that already had a fire burning on it.

 

Tashkent Test 1:

Spoiler

Test 1

RNLBNv5.jpg

 

hs2tdzs.jpg

 

25 fires set over ~500 shells hit.

Admittedly I was still figuring out my methodology and actually ran out the clock, and didn't think I'd actually passed 500 shells hit until I saw the results.

Tashkent Test 2:

Spoiler

Mti89aF.jpg

 

EKyWROk.jpg

 

23 fires set.

The torpedoes were fired after I hit my 500 shell hits goal, in order to end the match faster.

Okay. So what about Shimakaze? Surely only 1% lower fire chance couldn't be that big of a difference.

Same exact methodology...

 

Shimakaze Test 1:

Spoiler

vD5mTCD.jpg

 

TptZnPM.jpg

 

As you can tell by my taskbar clock, I actually ran Tashkent Test 1, then Shima Test 1, then Tashkent Test 2, then Shima Test 2.

As before, torpedoes were used to end the match after passing 500 hits.

 

16 fires. Nearly half of what Tashkent got in her first test.

 

Shimakaze Test 2:

Spoiler

48a2Wci.jpg

 

OQnO8Xi.jpg

 

13 fires.

Again, remember, I used the exact same targeting discipline in Tashkent, not spamming shells into burning sections like some of you are already preparing to accuse me of.

Tables!

Spoiler
Shots Tashkent fires Shimakaze fires
100 6 2
200 4 2
300 6 4
400 5 3
500 4 5
100 3 3
200 4 3
300 8 2
400 5 2
500 3 3

 

I documented the fires started every 100 shots made. This is the resulting data.

 

Wait, seriously?

 

Conclusion

Spoiler

Remember that Tashkent is Tier 9, and actually has a penalty for starting fires on tier 10 battleships.

 

And yet in both tests she still started nearly double the fires Shimakaze started using the exact same techniques between both ships. Check the replays below to verify that no, I did not shoot at flaming sections in Shimakaze "to prove a point", as I know some of you will accuse.

 

Tashkent's fire chance is at least double what is listed. More testing is required to see exactly what it is.

 

These results do not in any way reflect a simple 1% difference between these ships' ability to start fires.

 

It is absolutely repeatable. However, 1000 is generally considered a good number in most surveys so I'm comfortable with this result.

 

If you doubt me, run your own tests. Those of you with Shimakaze and Khabarovsk can probably run even more accurate tier 10 comparison tests. Just make sure you put as much effort into making them accurate as I did. Because I know some of you are shady enough to fudge results just to prove a point (IE: Shooting burning sections just to make sure you won't start more fires than you want to)

 

As it is I'm actually rather pissed that I actually wasted what precious little time I have to play games after work on proving that I'm not a paranoid idiot to some random strangers on the internet.

 

 

Replays

 

 

  • Cool 1

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

Tashkent doesn't have a "penalty" for starting fires on a T10 BB.  A T10 BB just reduces Tashkent (and Shima) fire chance by roughly 50%.

 

As far as your results, you're looking for an answer.  Case in point, you calling 16 fires "almost half" of 25 fires.  It's 2/3, which is not nearly 1/2.  

 

Also I would be careful about running "fire tests" with uneven fire percentages.  In this case, we have no conclusive proof of how 7% is rounded after it's halved.  Your numbers could be 4% vs 3% which would make the Russian numbers well within reason for a small sample test.  That is, you would expect the average to be 20 fires over 500 shots, you observed 23 and 25, but of those are well within reason.  It also puts your Shima expected average at 15, your observed was 16 and 13.  Almost perfectly what you would expect, even with the small sample size.

 

Equip fire flags to bring the Shima up to 8% and then run your test.  You don't lose flags in a training room.

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

Yeah because the first thing I want to do is waste another evening proving myself right when I already did it.

 

1% difference doesn't result in a 16/26 split over 1000 shells fired.

 

It's absolutely repeatable. Do it yourself. Just make sure you don't falsify your results in Tashkent/Khabarovsk by shooting at sections already on fire.

Edited by AraAragami

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

Actually, it does.  A 15/20 split but with 500 hits the 16/26 are both possible.  The time you got 26 the expected was actually 20.5 or so, so that makes it even more likely to get a 26 fire result.

 

If you did anything you proved that the listed fire chances are about as advertised.  Possibly evidence that fire chances are rounded down to the nearest percent.

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

Then do your own test and prove me wrong.

 

Just don't falsify your results in your desire to ensure that I'm never right about something.

 

My night is over. I don't have any more time to waste proving something to strangers on the internet.

 

I'm actually pretty pissed that I decided to do this, you know. I could've done something fun instead with what limited time I have to play games.

Edited by AraAragami
  • Cool 1

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

You're the one who thinks there is "hidden Russian fire chance".  The burden of proof is on you.  Your test supports that there is no hidden fire chance.  

 

If you're happy to admit you're wrong and move on that's fine with me.  So long as you don't consider your test to be some sort of proof of hidden Russian bias.

  • Cool 2

Share this post


Link to post
Share on other sites
1,669
[TBW]
Members
6,319 posts
11,878 battles

That 1% chance per shell seems to make a big difference doesn't it. You would have to know the equation (algorithm) that is used to determine the chance of a fire before you can cast your judgement. The math might still be right. Your test does prove that 1% makes a difference though.

Edited by Sovereigndawg

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

I consider my test proof that Tashkent's fire chance is significantly higher than what is listed. I repeated it and the result was nearly identical.

 

If you want me to waste another hour of my life than you'll have to pay me. I usually only do tedious, repetitive, intellectually unfulfilling work for strangers who hate me at a rate of $20 per hour.

 

And in come the apologists.

Edited by AraAragami

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

Your expected is 20 fires and 20.5 fires.  Your observed was 23 and 26.  Those are well within normal result ranges.

 

I think you have a basic failure to understand probability.

  • Cool 1

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

Then prove me wrong. I've already presented my evidence. If you were right than Shimakaze should not have scored so much lower on completely separate tests. The gap should be narrower. Shimakaze even could've overtaken Tashkent's fires started.

 

But she didn't.

 

Instead, it's still Tashkent by a huge margin. On a simple 1% advantage. Against a target Tashkent should be having a harder time lighting fires on.

 

Hell if you really want than I'll repeat it tomorrow vs Myogis with fire chance flags to bump Shimakaze to 8% just to prove once and for all that I'm freakin' right and Tashkent's fire chance is significantly higher than advertised.

 

But I'm going to expect compensation for wasting more of my time just for you.

Edited by AraAragami

Share this post


Link to post
Share on other sites
1,669
[TBW]
Members
6,319 posts
11,878 battles

No hate involved here. Shell velocity, shell size, fuse time and other considerations are adjusted by an algorithm.

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

The Shima performed almost exactly as expected, +8% once -12% once.  The Tashkent was close at +15% and +18%.

 

Neither of those results is one that you can say didn't come from a random set of 500 shots fired and 3% and 4% occurrences.

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles

I maintain that a 1% difference in chance does not and should not result in a 30% variance. Their results should be much closer. I expected Tashkent to light more, but not to such a massive degree.

 

Across 1000 shell hits, it's 48 fires vs 29. This is a much bigger gap than 1%. It's nearly goddamn double.

 

You want more tests, you'll get more tests.

 

Though I'm starting to suspect you just enjoy getting me to waste my time and will find some excuse to claim new tests aren't valid either.

Edited by AraAragami

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles
7 minutes ago, AraAragami said:

I maintain that a 1% difference in chance does not and should not result in a 30% variance. Their results should be much closer. I expected Tashkent to light more, but not to such a massive degree.

 

Across 1000 shell hits, it's 48 fires vs 29. This is a much bigger gap than 1%. It's nearly goddamn double.

 

Your problem is that you didn't get a 30% variance.  You're looking for it.

 

If you ran your test 100 times each you would take the time the Shima got 8 fires and the time the Tashkent got 32 fires and jump up and down screaming about a 400% difference.  This isn't logical or scientific, but that's not your objective.  You are out to prove a point and damn the facts.

 

Over 1000 shots your expected numbers would be 30 fires for the Shima and 40 for the Tash.  Your observed is 29 and 48.  Both of those are perfectly reasonable values.

 

From a pure statistics point of view, with a 4% chance and 1000 events to have even a 70% confidence level to say that the chance wasn't 4% you would need to see more than an 11% result.  11% of 500 would be 55 fires.  If you had seen 55 fires from the Tashkent you could say there was a 70% chance that WG fudged the numbers.  Instead you saw 4.8%  That proves that the Tashkent performs as advertised.

Edited by Grizley

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles
4 minutes ago, Grizley said:

Over 1000 shots your expected numbers would be 30 fires for the Shima and 40 for the Tash.  Your observed is 29 and 48.  Both of those are perfectly reasonable values.

 

We'll see what tomorrow's tests turn out.

 

Though don't expect me to be in a good mood about wasting more of my precious little time proving you wrong just so you can try to spin it so you're not.

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles
Just now, AraAragami said:

 

We'll see what tomorrow's tests turn out.

 

Though don't expect me to be in a good mood about wasting more of my precious little time proving you wrong just so you can try to spin it so you're not.

 

Don't bother, if you're bent out of shape with 1000 shots then getting a sample on a 4% chance event that will prove to any sort of standard that it's not 4% will drive you crazy.

 

Just a for example.  If you did 10k shots instead of 1k, and you saw 6% fires out of the Tashkent then it's still expected.  You would need to see 7% to say with 70% confidence that it isn't 4%.  You don't have the patience for that, and I don't blame you.  Just accept that the fire chance is as advertised and move on.

  • Cool 1

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles
1 minute ago, Grizley said:

 

Don't bother, if you're bent out of shape with 1000 shots then getting a sample on a 4% chance event that will prove to any sort of standard that it's not 4% will drive you crazy.

 

Just a for example.  If you did 10k shots instead of 1k, and you saw 6% fires out of the Tashkent then it's still expected.  You would need to see 7% to say with 70% confidence that it isn't 4%.  You don't have the patience for that, and I don't blame you.  Just accept that the fire chance is as advertised and move on.

No. I'm going to pursue this because I need to prove I'm not a goddamn idiot like you and everyone else on this forum capable of spelling their own name constantly claim I am. Even in this thread.

Edited by AraAragami

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles

OK, just do a little research into what you need to do to use statistics to prove that 4% is really 5%.  It's not trivial.

Share this post


Link to post
Share on other sites
3,612
[INTEL]
Members
8,416 posts
25,261 battles

I posted this to the other thread...

====================================

The 1% difference (Shim vs Tash) has enormous implications if you consider it across a number of shells rather than as base fire chance.


For example, lets imagine your chance of NOT setting a fire assuming all salvos result in 6 hits. That's a dependent calculation, since each shell in a six shell salvo must fail to ignite for whatever reason in order to avoid setting a fire. 

 

If you fire a single salvo with Tash (8% base fire chance), your chance of NOT getting a fire is .60 (.92^6). For Shima (7%)... it is .64.

Iterating that difference across 500 shells, 83.3 salvos, that 4% chance of NOT getting a fire on any salvo will indeed result in the different numbers you see. 

The Shima simply has a roughly 3-4% higher chance of not setting fires for any given salvo than the Tash. That's assuming equal hits and dispersion and no effect from the target. 

According to your numbers after 500 shells, the Tash set 25 and 22 fires , the Shima, 16 in the first test and 13 in the second. 

Yammy's fire coefficient is .5. That means each shell has half its normal chance. For Tash that fire chance is 4%, for Shima, 3.5%, giving a Tash a .78 chance of NOT getting a fire. For Shima, the chance of firing a salvo and not getting a fire is .80. 

That means the Shima will avoid a fire on 66 of those 83 salvos (.80 x 83.3), meaning that you will set around 17 fires. Which is very near your actual results of 16 and 13.

For Tash, ~65 salvos will not get a fire, meaning that 18 will. Your outcomes of 22 and 25 appear to my Mark 1 eyeball to be within normal variation. This is because the larger the sample (83 salvos here), the greater the chance of getting outcomes on the high or low side.

To meaningfully learn whether RU DDs really do have secretly boosted fire chances, you would have to repeated this test numerous times and then see what the data look like. If the Tash persisted in getting numbers above 20 for test after test, you would have a case.

Edited by Taichunger

Share this post


Link to post
Share on other sites
Members
1,044 posts
8,288 battles
2 minutes ago, AraAragami said:

No. I'm going to pursue this because I need to prove I'm not a goddamn idiot like you and everyone else on this forum capable of spelling their own name constantly claim I am. Even in this thread.

ZOqdT71.gif

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles
Just now, Grizley said:

OK, just do a little research into what you need to do to use statistics to prove that 4% is really 5%.  It's not trivial.

 

I'll let the data do the talking and watch the apologists scramble to explain away what's staring them in the face.

Share this post


Link to post
Share on other sites
1,497
Beta Testers
6,868 posts
4,189 battles
1 minute ago, AraAragami said:

 

I'll let the data do the talking and watch the apologists scramble to explain away what's staring them in the face.

 

I can say that in all my RU DD games I've never had one that would hit even 70% confidence level of the observed fires not matching the given fire %.

Share this post


Link to post
Share on other sites
7,208
[NMKJT]
Beta Testers
21,176 posts
3,867 battles
Just now, Grizley said:

I can say that in all my RU DD games I've never had one that would hit even 70% confidence level of the observed fires not matching the given fire %.

 

Because you never actually compared it to its contemporaries in controlled conditions where the intent is to start fires, not do as much damage as quickly as possible.

 

In regular gameplay how much time do you spend raking a superstructure even after fires are lit, hoovering up that HE alpha damage? How do you think that affects fires set per salvo?

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×