I do love a testable hypothesis.
Off the top, if the changes in player sessions were solely due to the event cycle going between air and ground and naval, then ground AB numbers would also be up with ground RB, but they’re not. (Similarly, we know naval aiming hasn’t had a net positive effect because naval RB is matching its growth and their were no aiming changes.)
The hypothesis is analogous to finding how many “swing voters” there are in an electorate and the %swing is a findable number. In this case, we can take the relative percentage of air and ground (AB+RB) games (ignoring naval for this because it’s an irrelevantly small number), removing the seasonal effect and also the effects of events that would tend to effect all modes equally (like updates and the April 1 event). We can cross reference that with the number of days a vehicle event of that type was running in that month: for air that’s the F-5A, Meteor and Marut events, and for ground the end of the FV4030, and all of the Fiat 6614, T-34 and T86 events, and express those as the number of “ground days” and “air days” in each month. That gives us this:
Spoiler
So, the effect the 6 months so far shows an effect (there was always going to be some swing) and an initial measure based on the 6 data points so far available. Basically low on the left is a month people swung to air while an event was on, and high and right is a month people swung to ground. If the hypothesis that events drive people between modes were true, we should see a very obvious slope up toward the right here, as more ground event led to more ground play.
The effect of individual events appears to be variable and within a fairly small range. The March ground event (Fiat 6614) did not pull people from air, over and above other factors. Similarly the Meteor event in April did not pull people from ground.
Overall, it looks like the “swing player” (like a swing voter but for War Thunder) manages to account for 2-5% of total games played. The other 95-98% of games played would be played in the mode players prefer for other reasons.
Furthermore that effect could be at the lower end (closer to 2% of games compared to 5%) to the degree the number of ground games is actually growing. If you look at the amount of ground games going from 65% in Feb up to 70% in July (when there was neither a ground nor air event), that 5% increase in ground proportions compared to air likely can’t be explained entirely just by a 2-5% event swing.
Another more likely possibility to consider, especially because Ground AB isn’t changing while ground RB is, is changes to the top tier ground game. Ground RB players, as I showed in the OP, are clustered at top tier. If top-tier Ground RB games are getting significantly faster this year (either due to more one-death leaving or CAS ending games faster, so the average high tier game length has dropped by several minutes) that could also lead to an increase in ground RB games on the same playerbase.
The other OTHER possibility is, as mentioned, that ground RB numbers are growing more than the rest due to new non-Steam audiences taking a liking to that mode. I’ll look at that possibility in another graph next.
