Data Analysis #5: What is the hardest mode (ground/air/naval) to get score in?

How long do War Thunder games last in the different modes on average?

This is more than just trivia. Just measuring how many sessions are played in each mode, which we’ve been thankfully able to do in 2025 thanks to StatShark, doesn’t directly tell us how much TIME is spent in those modes. And when it comes to events, knowing how much score is made in a game doesn’t tell us as much as if we knew how long the average game was, too.

It should be possible to measure how long games last beginning to end by surveying the game durations on the replay site. It’s important to note though, that people leave games early, either because they’ve run out of spawn points or don’t want to play anymore, so any numbers we get through this method have to be seen as how long it took the game to run for the last people to leave it, win or lose.

That cautionary note aside, what are the average numbers for the various AB/RB modes? (Leaving aside sim and PvE for this analysis). I took a sample of 1057 AB/RB games, all played on August 8, and then added to it with additional samples I’d collected previously going back to July 27 to give a cross-tab of at least 250 games (a statistician would tell you you need about 400 randomly selected games to get a good mean/median here) in each of the six modes, for a total of 2,918 games. The games were collected during weekday European evening hours, with the longitudinal day chosen from the period between the summer naval and air events. Sadly I left out sim and the PvE modes as being too different to really compare here, although it was interesting to see their relative numbers in the longitudinal sample, with Heli/Ground/Air PvE all very small but still there, comparable to the naval modes.

Games started per minute during the longitudinal sample for all AB/RB ranged from 60 to 106, or a new game every second basically (which is itself still rather remarkable, at least to me as a measure of worldwide server load). In the longitudinal sample 42% were ground RB, 28% air RB, 25% ground AB, and 5% air AB (with 9 naval games in the 1,057 for less than 1%), which is reasonably consistent with other data we had previously on prevalence of matches. A small number of games less than 90 seconds in length (there were a few of these) were excluded as probable false starts/server glitches

The result was this chart:

Spoiler

Naval modes had the longest average length (both median and mean), but air AB was surprisingly close. Ground RB was pretty much in the middle, with Air RB significantly shorter on average than AB; the situation was reversed in ground though, where AB matches took less time on average.

Mode and median don’t tell the whole story however, there’s also the shape of the curves (aka, the distribution) to consider. These are the curves of the three AB/RB modes in terms of time and absolute games measured in this sample.

Spoiler



In Air RB, the peak is right around the 5 minute mark, while Air AB peaks at 10. This makes sense given that Air RB is concentrated towards higher end missile play, and also that it’s TDM while Air AB is concentrated more in the slower prop BRs and you have respawns and lineups.

In Ground AB, the average game is around 7 minutes long, whereas Ground RB that’d be a short game, with average games at the 10 minute range. This is certainly in part due to RB having unlimited spawns if you have the spawn points for it, whereas AB is capped at 3 ground vehicles.

Both naval modes are actually surprisingly close both in terms of peak (around 11 min) and how their curve trails off with time. There’s an idea out there that naval AB is significantly shorter/faster and thus more efficient, but it’s really not there in the data (it’s about a 10% offset from each other). As we’ve noticed in other respects, the two naval modes are not offering significantly different experiences here, the way the modes are differentiated in ground and air.

Question 1: Which modes are the most popular in terms of hours played?

So if average session lengths varied significantly between modes, that would lead to sessions not being as useful data, simply because a session in a shorter mode was the same as a session in a longer one in Statshark stats. So knowing how many millions of hours were played in each mode can be useful.

We know from Statshark that there were about 264 million player sessions in July and the proportions between the modes are pretty similar month to month. If we multiply mean time/mode by those numbers, and express it as a proportion of both sessions and hours played in that mode (41.6 million total hours by this math), you get this graph.

Spoiler

Ground RB, by dint of longer game length, becomes slightly more dominant: 54% of game time vs 52% of matches. But overall the changes are pretty minor, suggesting users session can still be pretty useful data for evaluating AB and RB; knowing time played doesn’t really significantly alter our evaluation of a mode’s popularity.

Question 2: Which modes are most lucrative in terms of score and finishing events?

So we know average score per game in each mode from Statshark as well. Knowing a mean game length gives us an idea of average score per hour, which can be useful if you’re deciding how long it’ll take to finish one of these vehicle events.

Note: as a player you have a lot of control on this. You can minimize your time in lobby, change your server parameters to reduce queue time, leave games early rather than dragging them out to get in more games, play modes and BRs where you personally score well, and so on. What this is here is the playerbase average game length (playing all the way to the end) times the playerbase average score, by mode, that’s all. Leave your tips for beating the statistical odds in the comments below :)

Doing that math gives you this graph. (Note: Ground AB recently had its multiplier increased from 0.93 to 1.)

Spoiler

In general, the RB modes on average are still more lucrative than AB, their longer game lengths not offsetting their better multipliers fully. The other thing that I think should be obvious is air modes are currently relatively under-rewarded compared to ground or naval. In general this makes air-based vehicle events, like the F-106 one about to start tomorrow, harder on average, requiring more player hours in total for the same average reward, than ground or naval. This in turn leads to a lot of players more aggressively optimizing their play time for air events (using base bombing strategies, and the like)… in the other modes you really can just play. In air you have to be clever about it if you want to complete the event in the same amount of time. It’s not that in other modes players don’t also “turn it on” to compete events fast. It’s just in air they have to do it more. In this one sense at least, the air modes currently are the hardest to “do well” in.

Previously in this series:

Data Analysis #4: Pushing Past a Simple K/D Using Statshark Data

Data Analysis #3: The Arrival of Statshark Answers Some Old Questions

EDIT: Updated to change the ground AB multiplier to the correct one.

EDIT: Bonus Round: what’s more popular, naval or PvE?

7 Likes

would it be possible to get average match length per BR?

Certainly not easily, but Statshark has surprised us before, so I’m not saying “will never happen” anymore either.

1 Like

It’s interesting that Ground and Naval game modes are so close in modified score per hour. I suspected Naval would be close, but I didn’t expect Ground to be even closer.

Air Arcade probably is, but it’s much harder to judge Air Realistic this way. It’s the only game mode with just one life. You can compare other game modes with multi-spawn systems pretty accurately, but Air Realistic is different. It’s not about whether players want to respawn or not - they simply can’t. So the average score you see here is never based on the full battle length.

It’s no coincidence that the average score for an Air Realistic battle is just 686 - that’s only a bit more than one kill. Many players play Air Realistic very aggressively. I very often see players flying straight at the first enemy they spot and dying within 3-4 minutes. I definitely wouldn’t say the average time alive for every player in Air Realistic is 8 minutes. That means the average modified score per hour for Air Realistic on your graph is underestimated.

But the progression speed in Air Arcade is indeed very poor. A long time ago, I logged all my battles to see how long it took me to complete one stage of a 40k score event in Air Arcade using a rank 4 fighter (target score: 44,444 due to the 0.9 rank 4 multiplier). Here are the results:

Spoiler

For me, it’s absolutely ridiculous that you need around 160 kills just to complete a single 40k stage of the event in Air Arcade. That’s a lot of kills, and I felt completely exhausted after such an intense session.

That’s exactly why I now play events in Air Realistic. Even though I have much less experience in this mode, it’s still faster and definitely much easier to get the required score there.

1 Like

Longer if the game wasn’t infested with one death leavers with no line-ups, surely.

Would be interesting to see how Air Sim compares to Air RB

As it stands right now Air AB event rewards should probably see a 5-7% score multiplier increase to still be competitive, same as Ground AB got late last year. At the current reward values, every air vehicle event they run is acting as a positive DISincentive for people to continue playing AAB, rather than switching to ARB or ASB sooner than they otherwise might like.

Bonus Round! From replays, what is more popular, the Assault (PvE) modes, or the naval modes?

I’ve run a few samples on this and here’s a good example, in the terms of new games started per minute:

Spoiler

In total, in this one 26 minute sample, taken in European weekday afternoon hours, you had 124 Ground Assault, 46 Air Assault and 20 Naval Games (AB and RB combined)started. On average the Assault games in samples lasted around 10 minutes and had about 10 players each, comparable to other modes.

So yes, it is true that despite a lot of work, Gaijin has not yet managed to bring the naval game as a whole up to the level of popularity of either of its ground or air PvE modes. Along with helicopter PvE and ground sim, it remains very much a niche product that most of the player base doesn’t have any engagement with. As a naval player myself, one who hasn’t played Assault PvE after the first couple years I played this game, I’m glad naval exists… however it’s still true that you could do something to radically improve the PvE “Assault” modes today (like, a Battlepass task that could only be done in assault, the same way some can only be done in naval, as one example) and it would likely make more of your playerbase happier than anything you could do on the naval side, too. (The difference there of course is, Assault PvE doesn’t cost much to keep going, but it doesn’t make Gaijin any money either, whereas ships do.)

1 Like

Maybe you pick more than just 9 Naval battles for your theorycrafting?

This isn’t funny anymore, it’s just sad.

Want me to play naval? (because I do have the time to play it and I am interested in naval ships/history)

Fix the maps. Fix the spawns. I simply have no interest in spawning in full sight of the enemy fleet. Silly 20th century line battles in ships that were not designed to be used in such a way. That’s why I just play coastal battles whenever there is a naval event on.

I spent a lot of time in Navyfield and WOWS.

2 Likes

Do you even check the NRB battles absents - 9 battles per day doesn’t seems odd to you?

Most of that you not even picked novice battles, there are plenty of them.

ps .Do Statshark even parse them for their statistics? Doesn’t seems like they do to me.

cmon man you can do it, i’m waiting for your wall of text

So the argument put forward in the post above is that there are actually a lot more naval battles that aren’t being captured by the replay system, or its search results. This would be very easy for someone to test, just compare the number of naval battles they fought in their game record in a period with the number findable on the replay site with the “with my participation” tag. Maybe next naval event someone should try that and better characterize what the bug is and send it in, if it concerns them.

Naval Realistic as I’ve mentioned previously a few times, does appear to be more of an undercount, relative to Naval Arcade in general replay search results currently. (In previous checks I tried , the NAB total appeared to be missing 10-50% of games played, while NRB general results could be missing as much as 80%). But in the test above, even assuming the missing portion were to increase the result by 150% on that basis, it would have led in the sample above to there being at most 50 naval battles across both modes, vs 46 air assault or 124 ground assault, still leaving both naval modes combined as easily the inferior mode(s) in terms of battles started. So it wouldn’t change the material fact that there’s currently far more PvE assault games than naval games started per unit of time, except in degree.

(As I’ve said elsewhere, this needs to be factored into any discussion of the “success” of naval as a mode… assault PvE costs nothing to run or build and still retains a comparable player base (as does Heli PvE, as discussed elsewhere). There’s a good argument there to be made, honestly, for doing one seasonal task per battlepass in a PvE assault mode rather than naval… you’d likely engage more players.)

The argument newb battles should count against naval totals is less convincing. We have seen in the past people exploiting the “starter PvE” environment in AB naval to drive up their stats for one. And even if we didn’t know of that exploit, to count the battles of humans vs entire bot sides in arcade, we’d also have to count in all the “newb tank” and “newb air” PvE “battles” as well, increasing the advantage of those modes in terms of battles fought over naval still further and leading to skewed ideas in general statistical samples about what the “real battles” are. You’re not supposed to play new battles against all-bot sides after you finish research on a level II vehicle… the fact some people like to dodge that in all of the modes to get silly stats is not relevant to questions of average score or battle length for “real players” or “real games,” which as explained in the OP used larger samples to measure.

Where there was a bug in the past where 3.3-4.3 regular naval battles were being mis-filed in the newb battles section of the replays, that doesn’t appear to be the case anymore (indeed, those could be the battles that are missing from general search results now… maybe instead of actually fixing the misfiling of the starter-level destroyer fights, they just exclude them now from the general search results, that would be my first hypothesis… I suspect they’re still on system to find if you look for your or a specific player’s battles, just the non-specific search result is bugged, which isn’t going to be really that big a deal to anyone… except to this specific discussion). If you take the time to open the player list of any of those newb battles in the screenshots above, you’ll find they’re all legit newb PvE now, not “real battles.”

2 Likes

This post was flagged by the community and is temporarily hidden.

The post above states that somehow 9 results were used for something here, for the second time. As the first chart in the OP clearly shows, the set used for calculating for match time and score (n) was 361 for naval AB and 393 for naval RB, very close to the +/- 5% 19-out-of-20 data threshold you would use for a random sample of a data set of this size.

The point of the 08-Aug column in that chart, which has been clearly misunderstood here, is to show the vast variance in replays per unit of time by mode, which is why I couldn’t rely on just one regular 1,000-replay, or even 4,000-game pull on one day. Given the delta in the replay amounts, you would need 40-80,000 replays to get a consecutive data set including enough naval games with statistical usability. A hostile interpretation of this phony “argument” would be that it’s just another lie, from another person who seems to like to lie here on the forums a lot.

But I’m going to be charitable here, and assume people, despite their constant abuse of other players and devs and moderators, honestly like naval and want to keep it around. My point is that doing things like putting up fake facts to challenge dedicated volunteer researchers like @HK_Reporter when they tell you your idea has been tried and didn’t work, or challenging checkable statistical claims with such nonsense arguments to inflate naval beyond its actual objectively provable importance to the game, or calling the moderators and devs like @Smin1080p_WT and @Scarper_CM “lazy and unimaginative” just because they don’t agree with you, don’t actually help anyone.

It helps us all to be honest and clear-eyed here about the state the game and the mode are in. “They would have made the models for the mobile app so naval hasn’t cost them anything to build out” or saying “but are you counting newb battles?” are such obviously dumb, see-through arguments. How does it help anything and anyone to make them and look stupid by doing so? Naval is currently, despite massive investment over nearly a decade by the company, still only about 1-2% of this game’s activity. There can be no real question about that. That is investment that could have been spent on improving other modes. That was their choice. While as a fan of naval games I’m personally glad they did it, we should all be cognizant of the downsides.

(Same goes with people who say, “submarines will fix it,” just like “battleships will fix it” btw.)

The interesting issue in these stats for me was that NAB and NRB are still, after two updates with major naval changes, basically the same game length, at all BRs. There were two plausible arguments why the company changed naval aiming drastically in March. One was that they were trying to really start to differentiate two very similar modes to see if there were other player bases. (The other, less charitable explanation is they wanted people who bought store premiums for a lot of money not to get frustrated by having to learn how to aim.) The fact it turns out that naval aiming changes haven’t actually done that, and in this as in most other metrics (game length, score per hour, KpS, KpD, survivability, BR balance, etc.) the modes are still very close to each other in their characteristics should be concerning… either because you want naval to play a bigger role in this game, in which case they really need to start making even more drastic changes to pull them apart (AB and RB in ground and air have VERY different metrics from each other by comparison), or because you want naval to stay the way it is, in which case you’ll probably hate those more drastic changes if and when they come.

I think a very good case is building here now, given the failure of Leviathans or the March changes to really change anything about the mode’s core problems, to collapse the two modes into each other so you at least get up to about 2% of players and activity for ONE naval mode, and maybe make the other one purely PvE/Assault. Gaijin has made similar decisions in the past with similar poor popularity modes… Heli PvP, WWM, etc. Investing a third of your dev activity and marketing into two half-full modes that combined would still only make up 2% of player sessions at peak seems on its face unsustainable for any company and naval fans here should probably keep that in mind. The changes people disliked in March may seem drastic, but really if something doesn’t change something much more dramatic will have to happen here.

Anyway, there is a really good naval-is-dying thread that this all probably better belongs in, but I’m always happy when people making bad easily refutable arguments only bump my stuff up to the top of the forum again for more people to read, so I’m not one to complain.

3 Likes

Going forward let’s keep discussion polite, and constructive.

It was constructive, you deleted feedback on his assumptions. What in it was wrong?

What naval costs? If all of the models already were done for WTMobile. You again picking up facts that suit you theory ignoring those which doesn’t fit, for me it seems like biased not independent overlook. Do you trying to convince us that Naval is dead mode, but you not even playing it constantly, how do you know. You by yourself choose one unpopular mode - NRB and sticked there with boat fights, which even less popular, and now projecting you theories on whole Naval on all of BRs

Next time take one battle, why bother.

You imagining things, that never happened.

Again imagining, you do not know costs, you even do not know their financial results, that could be acquired from the open sources, do you ever try to find them?

Why should they, if money were acquired from the naval modes, and were not reinvested in it for two years, what is basically what you talking about, investing to the other modes with the profits from Naval.

I’d like to say mediocre changes, which could be done by single dev in couple of weeks or maybe month. Stat tweaks cost you almost nothing, as is for change something in damage model like hull compartments, which already have their core done years ago.

What for, to revive NRB? To disappoint NAB players? I bet you will not be statisfied if merge will be on the terms of NAB rules.

Again imagining that it’s third not 10% or even 5%, changelogs shows that naval changes is far from “third” in terms of amount.

It was never good.

Sure i am.

Success or not success it’s you own argument. I’m just pointing at the issues in that theory that you have.

You started from 'What is the hardest mode (ground/air/naval) to get score in?" and you took 9 battles, measured it with 400+.

it’s a big mistake in statistic, did you do that on purpose?