Data Analysis #3: The arrival of Statshark answers some old questions

[Bruce_R1]

@CorruptShadows9-live

You can go through replays and see what vehicle each person was playing, at what time, what spawn they chose, everything, you just simply had to spend time looking

So the thing is, when K2 Krabiwe tried that last year, just for ground RB, Gaijin took steps to block bulk downloading of replays to prevent anyone else trying exactly what you’re describing. And when Gszabi tried a bulk replay scrape last year to try to figure out average game length, the data came out so messy he had to admit it was beyond his ability.

So yes, in a theoretical universe of infinite monkeys what you’re saying could be true, but the best dataminers in the game community worldwide found in practice it was impossible until Statshark came along. Maybe they were all just really stupid, hard to say. But these aggregate stats you’re looking at in this post weren’t available to anyone in any practical sense outside of the Gaijin inner circle before this month.

I do think it’s interesting to see how Gaijin will react as Statshark continues to get attention. Whenever someone tried to base an argument for a game change on Thunderskill stats, they could always dismiss it as “those aren’t the real stats, and we won’t tell you the real stats.” Not seemingly an option anymore.

“If you went through [service records] and looked at this for every player, you could piece together a total for all gamemodes”… Yes, check my previous pieces, I would do stuff like that with sample sets, I know the level of compute involved pretty well. What I and no one else ever knew we had access to though, and that you can’t get reliably through either scraping any amount of either service records or replays (except for air RB) is the BR of the match played. Absolutely crucial piece of data.

[CorruptShadows9-live]

You can get match BR through replays, in a match that is a full uptier, 8 players will be a full 1 BR higher than the rest. I dont think replays will show the crews, but you can also tell thru SP costs (a Heavy will cost 160 in a full downtier and 100 in a full uptier)

I never knew about the two bulk replays, so apologies on that part

I know statshark has a bot in WT, idk the name of it, but its what they use when you link your account to it. They likely use that bot for stat gatherinf too, or have a seperate one

[Bruce_R1]

No need to apologize, I know you’re defending them against any concern they’re doing something sketchy, I respect that.

At the time last year Krabiwe did his bulk replay pull I thought Gaijin blocked replay bulk downloads to protect their IP. But if this doesn’t bother them, maybe it was more to prevent everyone doing it and crashing the server after K2 Krabiwe proved it was possible.

[reverse_slope]

This is cool to see!

I got curious about the ground sim stats. I pulled data from the vehicle cards on https://statshark.net/globalstats and got these stats. Not sure how to best account for the “player-battle” or other double counting.

Here is games played by monthly dataset

Spoiler

For BR 5.0 the most played vehicle is the Tiger H1 making up 17.7% of BR 5.0 games played and 38.1% of germany’s BR 5.0 games played. For BR 6.0 the most played vehicle is the Tiger II (P) making up 9% of games played and 20% of germany’s BR 6.0 games played.

Here is games played by country and vehicle rank.

Spoiler

I’m guessing one death leavers and planes make up the difference in the higher ranks.

And here is a table

Spoiler

The data I got in .csv https://pastebin.com/raw/b7KJ6Pci

I’m still not sure about these numbers. The total ratio of ground kills to deaths is 0.935. This suggests that 93% of ground deaths are from other ground vehicles, leaving only 7% of deaths from planes or leaving vehicle.

This doesn’t seem right, either something’s wrong or this confirms plane players suck mega ass???llolololololol

[SpectreXXI]

So this is why Super Prop BR is dead, yeah, it’s pretty much self-explainable that start from the Super Prop to the Vietnam War Jet are empty as hell.

[Bruce_R1]

So in another thread there was a claim that I was hiding Naval AB’s secret success by leaving out the March figures, which were collected across the major update launch with significant naval AB changes. Major updates will also increase traffic, so I considered this less useful than the months clearly either side of that update, but some questioned this. And there was also a concern that the use of relative numbers, which is what any data analyst would do in this case, were hiding something sketchy. So hey, here’s the absolute stats for all three months we have, too.

I’ve divided the monthly stats by days in a month to get equal player-games per day and put the naval modes on a separate axis because at 1% of total games they’re hard to see.

What you can see here is that with March, and the new update (which came on March 18), total player-games increased (black line). On average across all modes, about 14%. Then after the update, they declined again, but overall the game hung onto some of those gains. This is very normal… exactly what a company wants to see from an update. They’ll go up again temporarily. with the June update buzz next month’s stats, I’m sure.

Naval RB (red) mostly tracked this, actually going up 32%, higher than average, but then losing 28% so it’s basically even or a little bit below where it was before, in both relative and absolute terms.

Naval AB (orange) on the other hand, actually rose more slowly than the rest of the game through the update period in March (+13%) and then also lost more ground (26%) compared to the average in April. they’re about 12,000 player games/day off today from where they’d been if they’d kept up with the average on this update. If we estimate the average number of players per game to be 24, those lost 12,000 naval AB player-games would equate to ~500 fewer full Naval AB games a day than where we would be IF Naval AB had tracked with the game overall. (In reality of course, there’s probably about as many games being played, it’s just those missing AB players have been replaced by Gaijin’s naval bots.) It is too early to say naval AB has permanently lost 15-20% of its playerbase just due to this recent update. But it certainly didn’t gain any new players by it.

BONUS ROUND: In case you were wondering which AB BRs were getting pummelled by RB the most, indicating a player switch to the other mode or withdrawal from play in favour of doing literally anything else with your free time, here’s a handy color-coded chart comparing April matches by BR to February, showing AB losing ground to RB after the update at every BR except 6.7-7.0.

Spoiler

In absolute numbers, 2817 more BR 6.7-7.0 player matches were played per day (or about 100 extra games worth of players per day world wide), but at 1.0-6.3 14723 there were fewer player matches per day (about 600 games worth) in April compared to Feb, amounting to an average 22% playerbase decline over just two months in AB if you leave out top tier.

[安洁莉特]

Hello！I wonder if this statistics website can track the win rates of red and blue teams on each map? I’ve always felt that some maps have serious balance issues, but I don’t have any data to support that.

[Bruce_R1]

They have a Discord and a Patreon, I would ask them directly as I’m just a consumer of their content like everyone else. Cheers. :)

[_Poul]

@Bruce_R1

So in another thread there was a claim that I was hiding Naval AB’s secret success by leaving out the March figures, which were collected across the major update launch with significant naval AB changes. Major updates will also increase traffic, so I considered this less useful than the months clearly either side of that update.

I already explained all this to the person you are talking about. He simply refuses to accept the facts and only cherry-picks the data that suits him. You can check out my post here where I tried to explain him, in a simple way, the same conclusions you came to:

Naval mode is finally dead

It’s you who is trying to convince people to your way of looking at things, and you are not even consistent in this. It’s you who claims that the increase in the number of players in March is solely due to aiming change in Naval Arcade, so you clearly ignore all other factors: You ignore that March was a month with a major update. The update was released March 18. New vessels were added to the game, so at least some players surely wanted to research and try them. You should always expect inc…

In my opinion it makes no sense to argue with this person, because you are wasting your time. I don’t think he will ever understand what we are saying here. The problem is that you can always choose some statistics (ignore everything else) and build a theory based on them. Even though this theory has close to zero chances of being correct.

He didn’t want you to include April data, because this data shows something he refuses to accept. He just want you to compare February to March data, which in his mind proves that Naval Arcade aiming change was a huge success. After all, March had more Naval Arcade players than February. He just deliberately ignores all other factors.

[Bruce_R1]

Yeah, well, that loss at the low end is something else, people can think what they want but the numbers are pretty clear. Fully half of the 2.3 games and a third of the 2.7 and 2.0 lobbies just… gone. That’s basically all your coastal-only play. Also over a quarter if you just look at 4.0-6.0 so a lot of cruiser players have left too.

[Kweedko]

@_Poul_

He didn’t want you to include April data, because this data shows something he refuses to accept. He just want you to compare February to March data, which in his mind proves that Naval Arcade aiming change was a huge success. After all, March had more Naval Arcade players than February. He just deliberately ignores all other factors.

Sure it’s i am the one who asked "February to March " not the Bruce_R1 himself self posted it.

@Bruce_R1

by leaving out the March figures

@Kweedko

Show numbers how much there was battles and players March to April in NAB and NRB, not percentages of players playing one mode over other.

@_Poul_

The problem is that you can always choose some statistics (ignore everything else) and build a theory based on them.

Just like Bruce_R1 doing it, and when question asked he retreat here posted excuses instead of requested data and chart where you can’t see the numbers.

When there is clearly same percentages drop between March and April in ~25-28% both in players and overall games in both NAB and NRB.
Even what he shows is who know what when removed 1.0-1.3 BR where was increase, and 7.3 section. And not showing actual numbers of this so they could be checked and seen, when some BR got way more players and games(3.3-7.7) and other got way few(1.0-3.0 where he actually played). That a literally cherry-picking as you say.

[_Poul]

@Bruce_R1

In absolute numbers, 2817 more BR 6.7-7.0 player matches were played per day (or about 100 extra games worth of players per day world wide), but at 1.0-6.3 14723 there were fewer player matches per day (about 600 games worth) in April compared to Feb (…).

Top BRs on EU/RU servers in Naval Arcade were almost always full of AI bots. I don’t think the aiming change affected that much players who liked to kill AI bots, because they mostly sail in the straight line anyway. So it doesn’t really matter what aiming system the game uses, they were always relatively easy to hit (now it’s even easier to hit specific ship areas, like ammo racks).

Based on your graph, it’s even possible that top BR Naval Arcade players see more real players now. This could explain why some players claim the update attracted many new players. Don’t ask me how new players are supposed to play at top BRs. For me this suggests that many players moved from lower to top BRs (this explanation makes the most sense). It’s also possible that older players returned to Naval, but I definitely wouldn’t return just to play against AI bots at top tiers. You could also do this with any aiming system at any time.

I have to admit, I never liked top BRs, so I didn’t play there much. My knowledge about Naval Arcade is mostly based on thousands of 1.0 - 6.3 BR battles I played. But even by playing at 6.0-6.3 I saw many 7.0 battles and how they looked like. I also spaded a few 7.0 battleships, so it’s not like I never played them, I just didn’t enjoy them. So even if there are more players at top BRs now, it’s not an attractive area for me to play. I completely stopped playing Naval Arcade after the aiming change.

[Necronomica]

@reverse_slope

I’m still not sure about these numbers. The total ratio of ground kills to deaths is 0.935. This suggests that 93% of ground deaths are from other ground vehicles, leaving only 7% of deaths from planes or leaving vehicle.

You also have to think about those ratios being for all BR’s and all matches. There might be some BR’s where planes do a larger proportion of the kills whilst in other BR’s the proportion is lower. Which would make sense for Sim as getting a air-to-ground kill would be harder in the lower BR’s compared to the higher ones due to the extra help technology advancements provides.

I’m actually curious if it’s possible to divide those stats not only by mode (Arcade/Realistic/Simulator) but also by BR to see if there is a big difference between the BR’s or not. Like if air to ground kills is higher in Arcade than Realistic and if it’s higher for BR 10 compared to BR 5 for example.

I’m also a bit confused about the win rate, should that automatically be 0.5? there can’t mathematically be more of one than the other as one side loses when the other wins. Unless a draw is counted as a double loss in which case ~6% of all matches in ground sim ends in a draw, sounds a bit high to me but i don’t play ground sim so i don’t know. :P

[ULQ_LOVER]

Hi, I have a question.

Is it possible to get data regarding Ground RB in terms of:

Number of spawns in tank and number of spawns in plane depending on B.R. and comparison of it? Like at what B.R. what is played the most.

Number of deaths to the air in ground vehicle (doesn’t have to be only SPAA but if that would be possible, it would be nice)

Number of deaths to the ground in air vehicle (again same with SPAA)

That data would allow to end once and for all many discussions.

[_Poul]

Yes, you can check what is most played in a given BR.

Just open this website: StatShark - See All Player, Missile, and Vehicle Statistics
Select the month (e.g. April-may 2025, which means April stats).
Select Realistic and then Tanks.

You will see a table with win rates and below this table vehicles, by default sorted by the number of games. You can sort them however you like. There is also a filter on the right side and a search bar you can use.

If you want to check specific BR, in the win rate table just click on the BR row you want to filter. For example, if you click at 4.0 BR row, you will see only 4.0 BR vehicles below. You can select more BRs than one (using control button you can even select/deselect completely different BRs).

For example, at 4.0 BR in Ground Realistic, the most played vehicle is the T-34 (1942).

You can also see how many air kills and ground kills the vehicle has. For example, the M19A1 SPAA has more ground kills than air kills, which is a bit funny, but not too surprising for someone who played this vehicle. But you can’t check how many times this vehicle was killed by air or ground vehicles, there is only one Deaths stat.

Comparing specific vehicles between each other has to be done manually.

[ULQ_LOVER]

The issue is I don’t see the distinction between ground and air modes, only realistic.

It would be really helpfull to have data from Ground RB for both air and ground otherwise bot kills from air modes are taken into consideration when it comes to air units.

Or maybe I just can’t see something or don’t understand.

[_Poul]

Oh I see what you mean. I don’t think it’s possible to see e.g. plane stats only for Ground Realistic battles.

General statistics show the number of players/battles with specific BR lineups, but you can’t see the details about these lineups. I don’t see a way to check e.g. which plane is the most popular in Ground Realistic.

[ULQ_LOVER]

That is unfortunate

[Bruce_R1]

I’m not going to answer the troll directly, because he’s just a troll and not a very good one, but you can see the numbers he wants just fine if you just switch from a black background to a white background for a minute or two, and 1.0 to 1.3 is included in 1.7 in that chart because even taken together they’re a very small number of games, an order of magnitude smaller than 1.7 and 7.3 is included in 7.0, because if you’re a dumbass who gimps your 7.0 lineup to fly a 7.3 plane, you’re basically a 7.0 player. All games are there. But yeah, never makes a point, only more empty accusations. I mean you can basically see the spittle on this guy’s screen at this point. It’s sad.

For the record I “retreated” here because people in the other thread were begging everyone not to encourage him.

[Bruce_R1]

@Necronomica

I’m also a bit confused about the win rate, should that automatically be 0.5?

The issue there is because it’s only a sum of ground vehicle flyouts, and whether they were in a win or a loss, all the flyouts which were in a plane in ground sim aren’t being counted in the winrate calculation. Basically you’ve proven there if you don’t play planes at all in Ground sim, your average win rate is 0.472, which would make sense because you’re leaving a tool in the shed. If plane flyouts were included, I suspect you’d get the 0.5.

I’m not actually too surprised that planes only do 7% of the killing in ground SB, given the difficulties with ground recognition. Assuming plane spawns are like ground spawns, only slightly higher than deaths, you’re looking at about 2.5 million plane flyouts compared to 43 million tank flyouts, with the tanks killing 41 million enemies and the planes killing 3.5 million in the aggregate over those 3 months. So a plane flyout in ground SB is still about 50% more deadly than a tank flyout on average, with a K/D >1 as opposed to a slightly <1 K/D for the average ground spawn. Not playing the mode myself, that doesn’t seem too crazy, or out of line with RB.