To give credit where credit is due, this article comes as a request by Bill_Halsey on the NA forums as part of the How to Control your Win Rate 1 guide discussion written by LittleWhiteMouse. Further, as a disclaimer, this article will be very stat heavy.

The point of this article is to identify a relationship, if any, between Win Rate and Survival Rate, Ship Type (Cruiser, BB, DD, & CV) and Tier. From both personal experience and previous research and analysis on this subject, it would have been my guess that there is indeed a relationship. It should be small overall, but significant, particularly between certain tiers. So with that, I have analyzed the data using correlations and a linear regression with multiple predictors  as a means to determine these relationships and their significance. So here goes.

The Data

All the data was obtained from in 2017 which was using the Wargaming API to collect data on every player in every ship on both the North American (NA) and European (EU) servers. A statistical program called SPSS was used to download that data from and randomly selected 5% of the download data. This equated to over 3,250,000 cases of over 308,000 Players by 223 Ships. Further, due to the nature of the analysis, the data was trimmed to account for players with less than 25 battles to correct for extremes. This brings the total data points down to 1,286,144; still a very hefty number. 


The first part of the analysis that I will discuss is the basic correlation between player win rate and survival rate across all tiers and ships but only if they have player 25 or more battles in a particular ship. In other words, if a player has played 50 ships, but only 30 of those ships that player has 25 or more battles in, we are only analyzing the stats for those 30 ships. The results of this correlation show a r=.3382(Fig. 1) with a significance level better than 0.01.3

Further correlations were conducted to identify how the relationship between win rate and survival rate changes when we account for Ship Tier and Type4 both individually and combined. The results, shown in figures 2, 3, and 4 respectively indicate that there is a slight decrease in the relationship between win rate and survival rate when we account for tier (fig. 2) as well Tier and Type (fig. 4) but no or minimal change when we account for Ship Type (fig. 3). All the correlations remain significant however despite these considerations. Due to the change in these correlations and as a way to further investigate how ship type, tier and survival rate interact with win rate, I conducted a linear regression. This will allow me to not only have a better idea of the of how the individual factors interact with win rate, but how the combination of them interact with win rate as well.

The results of the linear regression analysis (fig. 5) further prove that there is an overall relation between our factors when trying to predict win rate. The subsequent coefficients table (fig. 6) outlines how each of the factors interacts – both positively and negatively – with win rate. As previously noted in the earlier correlation with Win rate and survival rate when accounting for ship type, when no change was noted compared to the base correlation, the linear regression coefficients table identifies that Ship type as a predicting factor for win rate is a non-significant factor (beta=.001, t=.843, sig.=.399) while the factor of tier has an apparent negative relationship. This was a a little surprising as it would otherwise be expected that players would become better as they grind up the tiers, thus increasing their win rate.


While there are changes in win rate as survival rate changes, the effect that ship type has on overall win rate is almost non-existent, while as a player goes up in tier, their average win rate decreases. As previously mentioned, this was a little surprising, however, there does appear to be an overall increase in survival rate as players go up the tech tree (tiers; fig. 7). My conclusion here can be broken into a few different factors that ultimately equate to more passive play. The first is ship maintenance costs. As you climb into higher tiers, the cost to maintain your ship increases to a point where if you get sunk and have not had a stellar game, even on a win, you can lose money. The second factor is the power differences between the tiers. The difference between a tier 3 and tier 5 ship is not the same as that between a tier 8 and tier 10. Players are more likely to make bold plays or even play foolishly if there is less at steak (credits) than at the higher tiers. Couple that with the increase in ship maintenance costs and players are less likely to risk damage (repair costs), particularly if they are not on a premium account, to help secure a win.

What this comes down to is a more in-depth look at my previous article, Winrate as a Player Metric attempting to look at the further effects that survival rate, ship type and ship tier have on the a player’s win rate and how futile (though significant) those effects can be used to predict a player’s win rate. Though win rate still remains a valid metric to determine a player skill, it is certainly incomplete. The same can of course be said for survival rate as we cannot make a determination, without looking at multiple other data points, as to why the only survive such a percentage of the time.

Finally, though it was highlight earlier to show how survival rate does increase with tier, figure 7 shows the means and standard deviations for all ship types at each tier as requested by Bill_Halsey. Also, for all those interested in the full set of analyses that were run, you can download the PDF below.


It has come to my attention that this does need to be mentioned – sadly. The conclusions drawn here are based upon older data and further data that does not give all the details about certain instances such as when a player sank during a match or if their team won while they remained AFK. Further, due to the age of the data, the effect that certain ships or mechanics may have within a battle cannot be counted as that data does not exist in the data set that I currently have. Examples, in the case of battleship survival rate, are the Pan Asian destroyers with their low detection torpedoes, AP bombs on certain carriers, and of course the increase in the HE spam meta. Even as the overall skill of the player base increases, I would expect that survival rates would decrease for battleships at the lower and mid tiers due to the deep water torpedoes and decrease at the higher tiers due to AP bombs and increased HE fire.


Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Survival Rate Case Summaries with means, standard deviations, and case counts

Liked it? Take a second to support Warship News on Patreon!
  1. The request can be seen on page 18 here
  2. The value of .338 means that there is a positive relationship (one goes up, the other is expected to go up) between win rate and survival rate, but not necessarily causal. However, with the r=.338 value, we can account for about 11% of the relationship between the two leaving the remaining 89% to be due to other factors.
  3. This means that the chance of this correlation happening due to chance is less than 1%.
  4. Ship type is coded as 1-4; Battleship, Cruiser, Destroyer and Aircraft carrier respectively

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.