I first signed up for ThunderSkill a year ago dude. Please stop being the stereotypical Russian main.
I THOUGHT I signed up, but didn’t. I then contacted the guy who “runs” it and he (allegedly) tried to help me sign up and couldn’t even figure it out himself, then gave up.
3 of the top 10 patreon supporters literally have [NULL] as their name on the front page.
There are dozens of vehicles with “0%” or “100%” win rates which is obviously completely impossible and a bug.
Almost every single one of the vehicles has the wrong BR since nobody’s updated them in years.
“People can totally sign up man!” Weird then that the T-26 reserve tank has 24 battles on record in the last month. 24. I’ve personally killed more T-26s than that in the last week, and I’m not even playing lower than BR 3.7, just scrap ones people bring out sometimes at the end of a match.
The overlay heatmap site treats NULLs on Thunderskill as “0% win rate” which is a week 1 remedial compsci student error. If you actually look into it, this alone lines up with almost all the dark red cells on the whole graph.
Thunderskill is Laughable. Garbage.
And those are just BUGS, not “bad designs” of which there are many more. Such as the heavy non-randomized selection bias others mentioned above, or the fact that the site does not limit the data to only the vehicle’s natural stated BR. So, for example, a puma armored car from BR 3.0 being taken out at 10.0 to rush a cap and then J-out into a plane, gets merrily counted toward the win rates and the puma’s stats at 3.0.
INCONSISTENTLY letting some people in (by region? or what?) is even worse than letting nobody in. That messes up the sample even more.
ahahahaha this is possibly the most embarrassing thing I’ve seen from Thunderskill yet so far, just noticed. The most played vehicle in AB from Thunderskill members yesterday was a $125 discontinued marketplace vehicle from 7 years ago 🤣😂
So weird that brand new players aren’t magically on TS already to record matches with the T-26 so you can see how effectively you’re farming new players.
Explaining one reason why a dataset is so unnaturally skewed, and somehow thinking that is an argument in favor of the dataset being MORE, not less, valid.🤦
Nitpicking an obvious weakspot in the data set and pretending this means everything else is invalid because there is a lack of a sample size for reserve tanks so we can’t look at the performance of a T-80 BVM at top tier.
Yes how dare I mention a specific weak spot while listing out 10 weak spots…???
Did you not notice that there were many many other items in the list of weakspots? Not that this one doesn’t matter anyway for top tier, it absolutely does:
“so we can’t look at the performance of a T-80 BVM at top tier.”
Half the people at top tiers are new players who bought squad vehicles and stuff to get into top tiers instantly. The win rate of a T-80 BVM is heavily affected by what T-80UK’s do, for example (which I can’t even find on thunderskill? Is it in cyrillic or something?)
And despite all squadron users heavily affecting the BVMs winrate it still pulls in a 60% winrate, one of the lowest winrates it has had since release.
The point is that you have NO IDEA whether it pulls in a 60% win rate or not.
Found it by the way: https://thunderskill.com/en/vehicle/ussr_t_80uk the uk has a 51% win rate. From the same matches, in the same nation. Lol. How is that supposed to make any sense whatsoever?
I’ll give you one very plausible reason: Because the people driving the actual top tier tech tree vehicles, who also signed up to thunderskill, and are still playing today, are grizzled veterans (mostly from the Russian audience, since the maker of the site primarily advertised to them, not western forums) who probably go out in 4 man squads and decide almost every game they’re in, while the majority of NORMAL games not blessed with their 1337 squad are full of a bunch of squad vehicle wallet warriors and other scrubs who suck and lose constantly.
So the actual win rate is probably more likely to be closer to the ones seen on the scrub vehicles like the T-80UK, of 51%, while 60% is achieved by a minority of veterans in teams with headsets on teamspeak and shit. And since Thunderskill is like 1% or less or something of active players, those squads can be in a tiny tiny % potentially of all matches played at that BR
but I didnt ask if others didnt know???
The point is that you have NO IDEA whether it pulls in a 60% win rate or not.
Neither do you, at least TS shows me 12.000 examples that support it, which weighs a lot heavier than a single person, you, just saying it’s wrong.
And yes, you found the T-80UK I already linked in the previous post for you, good job.
Just because you refuse to make an attempt to understand statistics and they don’t make sense to you at face value doesn’t mean it makes no sense at all, it just doesn’t make sense to you, that’s entirely different.
And I’ve already explained why the whole ‘try hard, sweat lord and grizzled veteran’ argument is bogus.
And then consider the fact that for every person who signs up, it also includes their entire squadron.
So if you want to suggest that it’s mostly sweaty players or whatever, which still wouldn’t invalidate the statistics, it’s negated by the fact they make up less than 1% of the database as it adds up to 127 other players as the majority of players are part of a squadron, and possibly even a whole bunch of squadrons as it also has records of previous squadrons.If you are a sweaty player and look yourself up, you add another 68 players that are in your squadron, it also adds your previous squadron which in this case consists of 4 players so an extra 72 players for looking you up.
If I look up RazerVon as well, who, coincidentally, is in the same squadron, it doesn’t add extra players from that squadron, but it does add his previous squadrons, which is 72 + 93 + 189 + 45 + 64 for another 463 players… so by just looking up 2 players, there are an extra 535 players added to the database, so it doesn’t matter if you are the sweatiest of sweaty sweatlords as you make up less than 0.2% of the data
Yes, almost as if that’s why my very first comment in the thread said “We don’t know what the win rates are, full stop.” “We” being a word that includes myself.
“Just because you refuse to make an attempt to understand statistics”
I’ve taught university statistics, math, AND research methods, but yes, tell me more about how I don’t understand statistics. Actually, please don’t. Instead, focus on the actual arguments and not ad hominems.
“And then consider the fact that for every person who signs up, it also includes their entire squadron.”
Yes… and the other people in a sweatlord’s squadron are also sweatlords. That was the main BASIS of my hypothesis, not a counterargument to it, what are you rambling about?
1 Like
Gaijin’s strongest shills on the prowl
This reasonably well summarizes the sample bias issue of Thunderskill.
In the case of the Obj 279 and other older rare event vehicles its especially the case that the sample is both small and heavily weighted to older experienced players.
Not only is 2% not enough. Also think that people who look at statistics will generally be people who are concerned about improving themselves. It will probably be a population well above average. They will play carefully and exploit the best tanks.
The problem with TS is not one of sample size - it’s more than enough for that. It’s a problem of sample composition, since it isn’t taken randomly from the wider general population, and is therefore unlikely to be representative of said population.
Or, as aLiVe said,