Finally, the definitive answer about Warships ranked

I was pretty disappointed in not finding definitive link between play time and rank so I kept looking, this time putting the horse before the cart. If we look for multiple factors, we must first find unrelated factors. Winrate isn’t unrelated from battle count, since the lower the winrate, the more battles you need to gain the same rank. XP and damage aren’t unrelated either because of the tier switch. Those who played lots of battles in T10 will have higher XP and damage than those who played only in T8. But frags should be unrelated. Is it? To see, first I’ve split the 37K players in EU season 9 into 10 groups by frags and 10 groups by battle count, trying to create somewhat equal groups. If I’m right, there won’t be any pattern:

defi1

There are some overpopulated areas (most notably the low battle, high winrate corner where sealclubbers resided), but generally the grid is neutral. Battle count and frag count does not create large, correlating areas, they are pretty much unrelated. But to make sure, here is a linear fit between them, with absolutely horrible correlation:
corel

Now, with two independent variables, we can look for the result, how they contribute to rank, calculating the average rank of the groups of players defined by the above grid:

result

One rarely sees so perfect correlation. You need both enough battles and high frag rate to reach good rank. No matter how good you are in one field if you lack in the other. So the definitive answer is that casuals likely won’t rank out, but no-lifers with bad skills neither.

Finally, a personal note. My results were 42 battles, 0.93 frags. This places me to the last column, 4th row. The average rank of that group is 9.8, so my rank (10) is exactly where it supposed to be. Since my frag rate is already top 10%, all I have to do is play more battles next season.

Update: the standard deviations of the groups:
sd
It’s between 1 and 3 everywhere, meaning that approximately 2/3 of the players are 1-3 rank away from the value defined by their frag and battle count. Since the matching brackets are 5 rank wide, this SD is pretty good. The high battle count rows are on the wider SD zone, so the population is more diverse there, but still within a bracket range.

Author: Gevlon

My blog: https://greedygoblinblog.wordpress.com/

8 thoughts on “Finally, the definitive answer about Warships ranked”

  1. Average values are nice but a table with standard deviations of obtained ranks would allow to see the accuracy of predicted rank.
    “The average rank of that group is 9.8, so my rank is exactly where it supposed to be. ” I presume you got the 10th rank but explicit statement might be needed.

    Like

  2. @Gevlon: thanks for additions. As expected, standard deviations were low.
    Is expression “approximately 2/3 of the players are 1-3 rank away from the value defined by their frag and battle count.” based on https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule ? Maybe more appropriate is to use https://en.wikipedia.org/wiki/Chebyshev%27s_inequality , since rank can’t be less than one.
    High amount of battles seemingly predicts approximately 0.6 frags.

    Like

  3. @Gevlon: thank you for calculations.
    About 15% of groups have the amount of elements out of 1 sigma greater than 0.39 and about 10% of groups have the amount of elements out of 1 sigma less than 0.20, so it balances out.

    Like

  4. I don’t think this correlation says anything.

    If frag count is a good proxy for performance then it predicts win rate and rate of saving stars.
    The number of ranks gained is a direct product of the number of battles and win/star saving rates.
    Add a bit of variance for starting position and you get exactly this.
    It’s nor correlation, it’s equation. Equivalent to stating “better win rate produces more stars”. Duh.

    Like

  5. @Stawek: obvious for me and you, but not to those who claim that you should sacrifice for your team and push like crazy. The thing is that nowadays saying the obvious is important.

    Like

  6. @Gevlon

    Are there any RNG elements that would tend to push or skew the standard deviations on an individual, or team basis?

    Like

Leave a comment

Occasional Hero

Adventures in Part Time Gaming

Me Vs. Myself and I

A little bit of everything, a whole lot of nothing.

Gnomecore

World of Warcraft | Final Fantasy XIV Blog

I HAS PC

Life and Interwebs

In An Age

The adventure I was hoping for was in a place like this

Why I Game

Wandering worlds, wondering words...

Bio Break

MMOs, retro gaming, music, and more

GamingSF

Online gaming blog