People really struggle to keep up …
Sry if I put it a bit out of context. I see this as a main problem at the moment in WarThunder.
I fail to see why people need to buy a cheat for WT when Gaijin give players the biggest most effective cheat available, a full down tier. The biggest seal club available and all legal. Get a t34 at 3BR ,get dropped to 2BR and get 20 kills, who needs a cheat?
Funny thing that happens to me all the time. When I get a full downtier, my team dies so quickly that I rarely get to enjoy it.
I perform better in full uptiers … maybe because I’ve adapted so well to the situation and can’t cope with downtiers anymore ^^
I just get clobbered by something from 30 years in the future and quit. Sorry. My ability to suspend disbelief is stretched like a penny to copper wire currently.
I hear a lot of frustration and would give a lot to be able to take it away from you.
1 Like
Like many I like(liked) the game but it is becoming harder and harder to like and I switch off so much more now. I loved the 3 Star Wars films of the 70s and 80s but what do we have now? WT is getting like that.
Doesn’t fix anything, but keep buying into the blind hope.
Also, 8v8 is possible still. 8v8 only doesn’t happen if there’s enough players for 16v16 almost instantly.
This is the video that goes with it
Seemingly just scrapes the replay for the vehicles being used, from there you just have to determine the highest BR vehicles and some assumptions.
That seems… Non-trivial, given that you can’t view a replay from server and derive that data. At a minimum it would seem to involve download of the 100,000+ game files mentioned, scraping the final scoring table and then replaying them all locally to pull the battle log file. You’d need to combine the results data from the final scoring table with a review of the battle log to get all the vehicles used by player, which could also give you stuff like what killed what. You then run against a lookup table of vehicles and BRs to get the BR for each player and the match. Doable, the WW2OL project did something similar with player-provided battle logs.
But a bulk pull of the entire replay server like that (even if just limited to the ground RB games) would definitely have to be seen as hostile by Gaijin as it’s essentially proprietal data in that kind of bulk, and reports are certainly believable they’ve have taken steps to keep it from happening again.
The speed seems a little remarkable, getting results for 4-7 Jan processed and up on a CC site in polished form by the 8th. That and the total storage you’d need for those files says this is likely not your regular home computer.
Ekshully… a replay is about 10Mb, so having 100.000 of them is just 100Gb, I’ve got larger USB sticks than that.
I also see how they did it; you can scrape the replay links off the website, and all you need is the first part, there’s a whole bunch of things in there that indicate vehicles used so if you have another database of vehicle identifiers-to-BR you can easily make the map.
I’d imagine results for 4-7 Jan were scraped from 4-7 Jan, and on day 8 you just do a query against your db of results and presto majesto, visualisation.
It’s not that hard or earthshocking. I was hoping they had figured out the proper format of a replay file so inquisitive minds can get at the guts of it and do fun things like make heatmaps of where people go to die. Or something. I hate saying it but at least WoT had the decency to use a relatively simple format that could be relatively easily reverse engineered :D
Gaijin doesn’t want us to have any actual information though so they can sit here and tell us that our information is wrong without any sort of proof of evidence of that and they expect us to believe it.
2 Likes
It’s about a TB (check your math).
100,000 replays x 32 players, so 3.2 million rows also takes it out of the range of things like Excel. Just saying someone knows what they’re doing, and that they’re using more than mom-and-pop tools. And yeah, the turnaround time is still pretty fast, given that the data would have got to the Russian CC on the 8th, and then they would have cut a whole video on it in only a couple hours. You’d need to figure out just what you had, and what it all meant, unless this was “the big heist” following on a series of smaller scrapes to perfect the extraction method.
Also suggests it wasn’t just a CC taking advantage of something that had just dropped in their emailbox from an anonymous fan that same day, if nothing else it suggests they knew it was coming and didn’t clear it with Gaijin before publishing. But we’ve seen CC’s get to violate ToS lots before this and not get punished for it, too. Or maybe selling 3 days worth of player data to subscribers on your Russian Patreon is just a thing a CC can do now.
Data in bulk like this, it’s kinda dangerous to them. Would you want to buy a US premium tank from them right now, knowing that is the actual win rate?
Alright, alright… I’ve got larger drives in my desktop ;) Heck, I’ve got 40Tb sitting in a server at work, not that outrageous to use a slice. Anyway…
Well yeah I never claimed people were using mom and pop tools; of course if you’re going to be analyzing a large data set you want to use the proper tools…
Well yeah, if I were to do something like this I wouldn’t be clearing it with Gaijin either - and there’s enough ways to publish it without it being linked to my game account :) Such as leaking it to a CC who, indeed, at times, seem bulletproof :D
Yeah that’s why for instance Wargaming (who aren’t stellar examples of great map design) generally don’t mess with their maps the way Gaijin does because they know damn well that there are a few people suffering from weaponised autism (hi… I’d be one of them) who will go dig into a bajillion replays to generate heatmaps to prove them wrong.
But I guess in the end, it all boils down to this: potatoes are gonna potate. And you’re not allowed to tell them (or demonstrate to them) that they’re potating. Because then they’ll be sad, and they will stop buying premium time and vehicles. And information that lets you get a leg up on your enemies, well, that just won’t do - we can’t have that, because lord forbid you use your brains to play instead of just your W finger.
Data in bulk like this, it’s kinda dangerous to them. Would you want to buy a US premium tank from them right now, knowing that is the actual win rate?
I mean the problem isn’t people not wanting to buy a premium, the problem is the winrate then.
If your whole business model relies on tricking people into buying garbage vehicles and they’re opposed to this type of data because of it, that’s the problem.
It’s not that - they’re selling decent vehicles to a lot of garbage players because some garbage players have this weird fixation with the idea that a premium will somehow be better, and some will be garbage because they bought into high tier without any idea what that’s like.
I have a fair rack of premium vehicles across all nations, and the only one I have that I would say is garbage is the Japanese Type 16 (FPS) - that thing is just absolutely horrendous compared to it’s TT counterparts.
(Okay and the 2S38 because it didn’t make me an omgwtfpwnwingod the second I bought it…)
Anyway, back on topic (sort of); the problem with having data is that it can be used to keep Gaijin honest - while I don’t believe there’s evil conspiracies afoot, I do believe that they do not want an informed playerbase because it’d force them to explain their reasoning behind changes, and if some of those changes boil down to “good players use it to clap bad players, and the bad players will cry, and leave the game, but the bad players are the ones that buy the most premiums”, well… that’s not something they’d want.
Or that some changes just have no reason at all beyond the whim of whoever does map design. Or why a BR change for a vehicle isn’t necessary because we can now verify their claims. Or that perhaps maybe by analyzing the entire team makeup of a few hundred thousand matches shows that the matchmaker doesn’t just lump people together based on BR. I mean, imagine the hilarity that would ensue from that one alone.
(Tangent: I am convinced the MM does a bit more than just look at BR… topic for another day)
Isn’t the FPS the same as the TT version anyways? The main issue is that they’re 9.3 for some reason when they should be 9.0 both because of performance and because of the uptier hell at 10.3
And with the hard nerf of the M735 it’s a joke.
data can be used to keep Gaijin honest
Which is pretty much the entire story, they like to do whatever they want, dismiss everything we say because muh statistics and force it through anyways, with blatant lies if they have to.
It’s like the Type 16 (P), they both get M735. The Type 16 at 9.7 gets the Type 93 round which has double the pen of M735 so the 9.7 version is actually useful all the way up to 11.7. It won’t be a murdermachine but you do get mobility, scouting, and the potential to clap things from the back ;)
Are they going to fix it yet?
Huh, Japanese M735 shows 200-something mm of pen at 10m 0 degrees, at least, did for me in-game earlier when I replied to you :)
As far as M735… it’s a useless round anyway, that’s the reason the (FPS) shouldn’t be at the BR it’s at because it can’t pen anything worth a damn at that BR (especially if pulled into a 10.3 uptier). Same for the (P) version in the tech tree.
Doubt Gaijin will fix it, they’ll just look at stats and go like yep, working as intended - mostly because nobody plays the (FPS) and (P) all that much, and the people that do generally know how to work around it’s limitations somewhat so in the end stats say it should be 9.3… but yeah. Gaijin. Whatcha gonna do…
Huh, Japanese M735 shows 200-something mm of pen at 10m 0 degrees, at least, did for me in-game earlier when I replied to you :)
Yeah, that’s what it does now, it used to be 350mm.
The Type 16 is in no mans land, who wants to uptier their 9.0 vehicles to use the Type 16 P at 9.3 with extremely frequent 10.3 games, and then another step to use the Type 16 at 9.7.
Type 16 P and FPS as well as the Type 74 G should all be 9.0.