Jump to content
Sign in to follow this  
You need to play a total of 20 battles to post in this section.
A_Horde_of_Sharks

A detailed look at tier 7 with a histogram and frequency table

14 comments in this topic

Recommended Posts

Members
370 posts
5,045 battles

I decided to take a look at World of Warships from a scientific perspective, hopefully to bring some real quantitative data to the table.  There is plenty of qualitative data, but very little in numbers backing it up. 

 

Observation:

There is a large amount of talk regarding the strength of tier 7 due to it's matchmaking spread, but is it true?  Is tier 7 one of the strongest tiers in the game?  If so, where is the strength coming from?

 

Hypothesis:  Tier 7 is one of the strongest tiers in the game.  Ships that benefit the most, are ships that gain the greatest yield when top tier.

 

First some basic data:  I decided not to include the Flint due to it's precarious nature of being driven by highly skilled players.  I also did not include any clone Myoko (ARP, Southern Dragon.) 

 

Constructing a Frequency table: 

Finding the range  (highest value - lowest value)  Belfast  55.67 - Ranger 47.25 = 8.42 rounded off to 8.4

(range/ # of classes) will determine the class width.  Class number chosen 10.  8.4/10 = .84 

Ranger will be the first lower class limit.  .84 is the class width. Note: 1 additional class was added to account for the Belfast with a total of 11 classes.

 

 Classes          Win rate       #of occ.                        Additional info                     

class 1:    47.25%---48.08%    1               CV                                          Ranger

 

class 2:    48.09%----48.92%   4        DD,DD,BB,CL        Akatsuki  Mahan  Colorado Pensacola

 

class 3:    48.93%----49.76%   4         CL,DD,DD,CL         Yorck   Sims   Hatsuharu  Atlanta

 

class 4:    49.77%----50.60%    2            CL,BB                 Indianapolis  Nagato

 

class 5:    50.61%----51.44%    3         CL,CL,BB              Shchors  Myoko  Gneisenau

 

class 6:    51.45%---52.28%     1                  DD                  Kiev

 

class 7:    52.29%---53.12%      4      DD,DD,BB,CV        Leningrad  Leberecht Maass  Scharnhorst  Hiryu

 

class 8:    53.13%---53.96%      2              CL,DD             Fiji  Blyskawica

 

class 9:    53.97%---54.80%      2            DD,CV             Shiratsuyu  Saipan

 

class 10:   54.81%---55.64%     0               N/A                         N/A

 

class 11:   55.65%---56.48%     1                 CL                      belfast

 

As presented by the data, the mean is 51.86%  The ship spread is also interesting, it shows that the majority of cruisers are between classes 1-5, making them fairly weak at this tier.  Only the Fiji and Belfast are in the upper classes of 8 and 11.  Destroyers are heavily polarized, with the American and most of the Japanese dds occupying classes 2 and 3. This puts them close to the bottom.  Their russian counterparts, along with the german Leberecht mass and the japanese Shiratsuyu however, occupy classes 6-9, showing a much stronger performance.

 

Here is a Histogram to better visualize how the data is spread out.

 

     4 |        |- -- --|-- - --|                          |- -- --|

        |        |        |        |                          |        |

     3 |        |        |        |        |- -- --|        |        |

        |        |        |        |        |        |        |        |       

     2 |        |        |        |-- -- -|        |        |        |- -- --|- -- --|       

        |        |        |        |        |        |        |        |        |        |      

     1 |-- - --|        |        |        |        |-- - --|        |        |        |        |- -- --|

        |        |        |        |        |        |        |        |        |        |        |        |

     0 |____|____|____|____|____|____|____|____|____|____|____|

          C1     C2    C3    C4     C5    C6     C7    C8    C9    C10   C11

 

As you can see, the data on this histogram is skewed to the left and not uniform.  This means there are only a few ships at the top of the pile that perform far above average to compensate for the vast majority on the bottom.  

 

Conclusion: does the information support the hypothesis?  Somewhat, but it would not be enough to put this into a working theory.  More testing will be needed.  The two most heavily armored ships at this tier, the german BBs, are high performing.  It also shows that Cruisers, the class that arguably gains the least when top tier, is lagging behind at tier 7, with only the Fiji and Belfast on top.  This is important, because it shows through the scientific law of subtraction, that chances are, if the ship isn't a cruiser, it will perform better, with only a few exceptions.   

 

Thanks for reading!

 

 

(Source: na.warships.today 02/08/2017) 

 

Edit: Data was collected from all-time ship performance.  Any further articles I will post will feature data that is 3 months or newer, to assure better accuracy while maintaining good sample size.

 

 

 

Edited by A_Horde_of_Sharks
  • Cool 2

Share this post


Link to post
Share on other sites
Members
823 posts
561 battles

Hey, maybe put that table into a google doc and then copy and paste it into the thread to make it easier to read. It works, you just can't edit it on the forum.

Edited by Zehroflcopter

Share this post


Link to post
Share on other sites
Members
370 posts
5,045 battles

Nice work. Very detailed. Think you can do one for Ranked?

 

I may make another post featuring a new set of charts for ranked when the season starts. Until then, it's questionable how much will change when switching into the ranked meta.  If there is any consolation however, I believe that this data will certainly help choose a ship for ranked.     

Share this post


Link to post
Share on other sites
Beta Testers
105 posts
5,508 battles

Did you use a range date to display this data (i.e. in the last 3 months, or since 0.6.0).  I think games played during the last couple patches are most relevant instead of total games played since the inception of that line.  Additionally, it would be interesting to see # of matches played  in this data as that also has relevance on the performance of a ship.   Stats tend to be skewed high and trend towards normalization as frequency of games increases.

 

Thanks for the charts, was very interesting to look at.

Share this post


Link to post
Share on other sites
3,513
Members
16,315 posts
12,285 battles

     4 |        |- -- --|-- - --|                          |- -- --|

 

        |        |        |        |                          |        |

 

     3 |        |        |        |        |- -- --|        |        |

 

        |        |        |        |        |        |        |        |       

 

     2 |        |        |        |-- -- -|        |        |        |- -- --|- -- --|       

 

        |        |        |        |        |        |        |        |        |        |      

 

     1 |-- - --|        |        |        |        |-- - --|        |        |        |        |- -- --|

 

        |        |        |        |        |        |        |        |        |        |        |        |

 

     0 |____|____|____|____|____|____|____|____|____|____|____|

 

          C1     C2    C3    C4     C5    C6     C7    C8    C9    C10   C11

 

thats a map of a dungeon in https://en.m.wikipedia.org/wiki/Eye_of_the_Beholder_(video_game)
Edited by slak__
  • Cool 2

Share this post


Link to post
Share on other sites
Members
370 posts
5,045 battles

Did you use a range date to display this data (i.e. in the last 3 months, or since 0.6.0).  I think games played during the last couple patches are most relevant instead of total games played since the inception of that line.  

 

Good point, I will add more detail in the source at the bottom of the article.  I used the data of their total games played since they were released.  Perhaps I will limit the data to a specific patch for any future article to assure more reliable findings.

Share this post


Link to post
Share on other sites
10,399
[BLNCE]
Members
13,459 posts
44,054 battles

Use the two week data. I expect the data will be even more heavily skewed, because of WG's decision to turn T5 and T6 into sacrifices for T7 and 8.

Share this post


Link to post
Share on other sites
Members
765 posts
2,849 battles

Look, im not trying to argue your point (at all).  I just want to point out that Warships WR% stats are inflated.  I used to assume that the high rate of good and premium ship play accounted for this... but after re-evaluating... they are simply over-inflated by about 1%.  Perhaps this accounts for 1% dropouts.. i have no idea.

Share this post


Link to post
Share on other sites
Members
3,229 posts
7,820 battles

The win rates for all of the stat sites are out of whack.  If you take a little time and add up all the battles from all the ships, from all-time, you get a greater than 50% average win rate.  That is statistically impossible.  There are certain players who are not being included in the surveys and given the skewing of winrates to the high side, I would suggest that many of the below average to bad players are not being factored in.  This isn't a deliberate act on the part of the 3rd party stat sites.  It's just a limitation based on the information that can be pulled from WG's servers.

Share this post


Link to post
Share on other sites
Members
370 posts
5,045 battles

The win rates for all of the stat sites are out of whack.  If you take a little time and add up all the battles from all the ships, from all-time, you get a greater than 50% average win rate.  That is statistically impossible.  There are certain players who are not being included in the surveys and given the skewing of winrates to the high side, I would suggest that many of the below average to bad players are not being factored in.  This isn't a deliberate act on the part of the 3rd party stat sites.  It's just a limitation based on the information that can be pulled from WG's servers.

 

Interesting, I will edit out the 50% comparison and perhaps replace it with the average of tier 7. :hiding:  Either way, it will be pulled from supporting the hypothesis.
  • Cool 1

Share this post


Link to post
Share on other sites
2,319
[CVA16]
Members
8,890 posts
27,128 battles

Would be interesting to see these same stats for all the mid tiers (5-8). T5 especially since it gets bottom tiered most of the time.

Share this post


Link to post
Share on other sites
4,615
[-K-]
[-K-]
Members
6,121 posts
27,992 battles

Interesting, I will edit out the 50% comparison and perhaps replace it with the average of tier 7. :hiding:  Either way, it will be pulled from supporting the hypothesis.

 

More accurately, there is a WR inflation bias across all tiers due to the higher probability of a low-performing player hiding their stats than a high-performing player. If you wanted to address the question of whether 7 was a "strong tier", you could compare its WR against the WR of other tiers, although any such comparison is confounded about whether the observed differences were due to Tier 7 ships being strong, or Tier 7 getting a higher percentage of games as top tier.

Share this post


Link to post
Share on other sites
Members
765 posts
2,849 battles

Another WR% factor that skews perception:  

 

Using a hypothetical 4 ship set

Discontinued Premium  WR%=75%

Available Premium WR%=55%

USA tree ship=51%

Russian tree ship=47%

 

This might appear impossible because the average WR of the ships is higher than 50%, however lets now include the number of battles.

 

Discontinued Premium 100 battles

Available Premium 1500 battles

USA tree ship 2000 battles

Russian tree ship 3500 battles 

 

Now, the average WR is about 50%

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×