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

Game Discussion General Discussion
32

This is going down a bit of a rabbit hole, but the quote from BVVD in Dec was “Yes, we have plans for playable submarines, but no further details on this yet.” That’s what he says whenever the change is (1-n) years in the future. I haven’t seen anything clarifying since that would trump that from Gaijin. I agree they’re a natural counter to BBs and are still likely to show up eventually. The real limiting factor on that was always the effort required to retool the subsurface areas of existing random battles maps, which are often simply not deep enough at present. But yeah, we’ll still likely see subs before carriers or post-war missile boats (both of which face much bigger obstacles). I just don’t think it’ll be 2025.

Getting back to the subject, I think it’s also important to note how balanced Naval is now at 5.7 and below. I think it’s one of its strongest features as a mode. One way to express this mathematically is the chance of being in the top BR of a match (yay!) minus the chance of being in the bottom BR (boo!). In a perfectly balanced game with 4 BR steps this would always be zero. As you can see from the graph below, naval isn’t doing too badly, up to 5.7, it’s keeping pretty close to the “fair and balanced” line. 2.0 is fun, 2.3 kinda sucks, from there it’s a pretty smooth progression up to 5.7 as the sweet spot, from there you want to skip right to 6.7-7.0. This will of course all change at the top end with BR decompression as I mentioned but the principle then will be the same.

image

EDIT: The other two modes for comparison:

Spoiler

image

image

2 Likes
33

The only thing not freely available that statshark can track is your lineups, everything else is able to be found ingame

34

I thought this whole Community Update article was about 2025, but maybe I’m reading too much into it.

Maybe this was just a random answer for the far future, based on the fact they are getting so many questions about submarines. But you are correct and there is no date specified in this answer. It was just my interpretation that all these questions are related to the 2025 development.

I don’t think they ever said that. They just admit they have goals with how quickly you can complete the tech tree (and therefore the game). This doesn’t mean that all game modes should take the same amount of time to complete the tech tree. But this is again something that is not fully transparent. Their explanation brings more questions than answers. You can check this fragment here (from 15:44):

I saw community calculations on the forum, where it was clear that it’s getting harder and harder to reach the top tier with every major update. This is something I would actually love to hear more from the devs. Where are these RP reductions?

35

So that’s not quite right: obviously it’s publicly available data today, that doesn’t mean anyone knew it was there before this year. Otherwise there wouldn’t have been so many efforts over the years by dataminers and content creators to reverse-engineer stats to answer questions like, “what mode or nation is actually the most popular?” I can’t count the number of arguments I’ve seen on Reddit or this forum that could have been definitively answered if any of this was “freely” available. It seems we now have those answers out in the open, which is awesome. But there was no way until now for a regular person to tell that from looking in the game itself. (Also opt-in sites like Thunderskill would never have been created.)

(They were showing current lineups for a while too, but they’ve stopped doing that due to privacy concerns. So that info is apparently freely available as well, they’re just choosing not to show it anymore because it seems invasive. Which is fine by me.)

36

I would sugest that one reason for the rising popularity of the realistic modes may simply be linked to the kind of premiums sold.

For ground, light nimble vehicles make up a fair share of the expensive premiums. In arcade, they are far less useful than in realistic. Which means, it is usually not feasible to form premium lineups in higher tiers in arcade.

With planes it is even more obvious: In arcade, only attackers have bomb reload, fighters take way too long. You can’t really bomb well with premium fighters in arcade. Need I say more?

1 Like
37

Realistically, they’re just taking the statistics for the amounts of matches played of each mode at each BR, which isn’t freely available on its own but can be found through simple algorithms looking at each match played in a certain time frame

Thunderskill was opt-in because it was a lot older of a site, they weren’t planning to take stats from everyone, Statshark is connected to WT and tracks all accounts

Nevertheless, it couldve been determined prior completely ingame if someone had that much time and looked at every account ever made, the only thing statshark has added is the tine periods

38

You keep saying this was all findable “in game.” There’s currently no way outside of looking at the other players vehicles (with the exception for ground RB and SP, I know) of a regular player even knowing what the BR was of the match they’re in, or the one they played the night before. If the match BR for every single game you played (as one example) was a recorded and findable stream of data that players could have seen before this month, no one who knew ever said anything about it. I just think we should give Statshark a little more credit here for lifting up the hood and letting players see the engine.

(Also Thunderskill/WTDP were just bought out and relaunched by a whole new team last year. And they still didn’t have any naval vehicles. If this was so easy, they could have done it, but they didn’t.)

39

My argument here was more towards this, btw

I was arguing that all of this has been available ingame and that Statshark hasn’t breached any privacy or anything, and Gaijin would have no reason to shut them down

I was never discrediting statshark, I was just saying that all this has been findable, you simply just needed to work

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

Ingame, there is a menu on your profile that shows matches, time played in vehicle types, total hours, etc. If you went through and looked at this for every player, you could piece together a total for all gamemodes and from there determine the popularity of each.

Sure, you might need the external site to view replays, but you can view your vehicle stats and lifetime stats ingame, Statshark has added a rating system and will scrape data from each of your play periods, which is what makes it much more than whats ingame and it does so through computer algorithms

40

On paper, it makes sense in that the time-to-earn in air for the same amount of score assuming equivalent challenge and gameplay favours AB (Respawning planes so don’t need to compete for kills as much compared to RB, small maps + markers + mouse aim for increased and longer range lethality) is compensated for using score multipliers.

Ergo - getting 2-3 kills/match in RB over 12 minutes is fairly plausible even for a mid-skilled player. Not every match, but the effect is apparent.

As a Prop-tier dogfight-oriented ASB player, my progress looks like this looking at some games I saved:
75 minute game, 7276, worth 15K progress (high average)
80 minute game, 5475, worth 11K progress (low average)
63 minute game, total is 7226, worth 14k progress (my best game ever to date)
93 minute game, total is 3,633, worth 7K progress (a fairly pitiful game of low pop on off-peak hours)

Progress is score x 2.3 x 0.9 (rank 3/4).
(For comparison, someone suicide bombing airfields and respawning constantly can hit 15K score within an hour for zero effort provided they’re in a lobby that doesn’t have people intercepting them)

The problem manifests in more abusive approaches to score earning that inflate the score needed each time and also skew game modes. Sim is more or less ruined whenever a popular grind event appears because people start rigging lobbies and refuse to fly in competitive/dangerous ones.

At least on my part, this has an inverse relationship on sim pop - I start playing less and less as I get more and more frustrated with lobbies that feel unplayable due to PvE rigging/unbalanced due to people leaving because they want a high average game rather than low average or even pitiful game) to the point I finally end up taking a break waiting for it to end.

This is to say I wish there were ways to penalize PvE rigging better than spamming Schindlbee with reports and to somehow shut down lobby shopping for pushover opponents.

Until then, I’d honestly be glad if we had RB-level event multiplier in sim to discourage ingenuine tourism.

1 Like
41

Incredible work, this was the stuff I was expecting to see once StatShark launched.

Being able to access incredibly reliable data records for this game will open such a large amount of possibilities.

1 Like
42

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.

43

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

44

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.

1 Like
45

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

games
games1080×649 72.2 KB

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

country
country1081×648 51.3 KB

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

And here is a table

Spoiler

table

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

46

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.

47

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.

navalplayergames
navalplayergames2624×1477 110 KB

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

image

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.

2 Likes
48

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.

49

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. :)

1 Like
50

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:

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.

51

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.