Well, it could very well be used as popularity chart as well:)
BRs with most popular vehicles have noticeable dips in winrates.
Because lots of new players spend money, who puts the most money into WT?
2 Likes
Click bait M1AIM F-5C F-4S AH-64GR you name it no wonder why they lose so much
2 Likes
As @Miragen said raw statistics like this aren’t even relevant. Winrate shouldn’t even be used to decide which vehicles must be nerfed or buffed (at least if it isn’t played by enough players).
Here are both the Winrate and Number of games made by each countries, we can clearly see that minor nations aren’t played by many players and their statistics are consequently biased.
The stats come from WT Data Project
So firstlt the number of people are faking the stats but on top of that, new players (and inexperimented tend to play bigger nations) resulting in bad stats for those nations whereas most of the small trees enjoyer are experimented players.
Honestly i dont really know the solution…winrate at least uses some sort of feedback, but would not work at all in arcade, so i doubt it is what Gaijin is using. Kill ratios or some more complex metric might be better…but TBH i dont really know, would have to test different systems…hopefully someone is getting paid for it.
I am a bit sceptical of “manual” systems…players don’t agree between themselves…there are many occurrences where some are “completely” sure of two fundamentally opposing views, particularly when talking about vehicles of different nations. And goes without saying that devs doing manual only changes will probably not be accepted well…
Some hybrid system might work, but devs said somewhere it would be “work intensive” and not lead to more acceptance anyway…
Gaijin clearly use winratr to nerf or buff tank. Just look at how the Leclerc is mistreated only because it’s winrate is high
There is also a K/D spreadsheet from K2 Kit Krabiwe, if you’re interested.
Spoiler
2 Likes
What I’d really like to see is if there is a way to check the statistics of individual vehicles. Like their average score, K/D etc. Though I’m not sure how to go about that
This exists for some time now…as all non official stats, some errors may exist, but you can get an idea…
https://thunderskill.com/en/vehicles#type=army&role=all&country=all
You can look individual vehicles and players…even yourself.
I was referring specifically to the scraping from 100k replays. I know Thunderskill exists, but we’re talking about completely different sampling methodologies at base.
French has really good vehicles.
Didn’t know this^^ :)
Anyhow…it may help someone who doesn’t know :)
Well, it gives you a flawed idea as far as individual vehicles goes or most stats outside of individual user ones.
Mostly because of how Thunderskill polls for stats, which afaik is:
- People who have signed up (On average higher skilled players who care about their stats)
- People who have been looked up on the site by people who have signed up for the service (more random distribution, but also probably on average more skilled players who killed or otherwise was interesting to the other player)
This creates a non-probabilistic distribution of participants causing a biased sampleset.
Now it can certainly give an idea of what the more experienced players are currently playing, but idk if id pull much more from it.
I use it a lot to see MY OWN stats with vehicles…like which vehicles have the best rates when i use them :)
But YES…numbers are skewed by the aquisition process…but this would also happen with the other stats above…
No official numbers anywhere.
The WWIILogs data project showed how you’d have to do it if you wanted an actual independent evaluation outside service records. Didn’t catch on and staled out sadly. But was generating really good data on this stuff til it died… A much more reliable methodology than the Thunderskill stuff that crowded it out.
1 Like
Nooooooo gaijin gib back russian bias at top tier pls
I don’t know how people see a database of 4.7 million and think it’s not enough and niche.
Thats not what those words say. Like at all.
Is that the one where they pulled every available replay for some time period and sorted that data?
Your conclusion is that those 4.7 million are all somehow above average skilled players or experienced players.