Since subscription data became a private statistic, interested players have tried to track EVE's overall "health" by looking at metrics such as the Average Concurrent Users number, tracked by places such as EVE Offline. Much significance has been put into these numbers--first, when they rose significantly around Retribution, and, second, when they have since fallen, particularly in the past year. There's little doubt that more people played EVE 3 years ago than 5 years ago, and that more people likely played EVE 3 years ago than currently. However, little else is known with any significant degree of confidence, making ACU numbers an interesting if mostly unreliable statistic. Players will, of course, continue to make exaggerations about EVE's health--good or bad--based on this single statistic alone. However, it is an unreliable piece of evidence that should not be trusted unless paired with a large amount of other supporting data (which it rarely ever is). Here are three reasons why, as well as a forth reason which will (probably) come in Phoebe:
I. DDoS strikes skew ACU numbers
DDoS events have been an ongoing issue for CCP. In addition to annoying players, one thing they certainly succeed at doing is skewing the ACU numbers. What appeared to be a DDoS strike happened Saturday during peak EVE time--server population was just over 41k. Then, at least 20k players were disconnected. When I logged back in, server numbers were just climbing back over 20k, creating a noticeable nose dive on EVE offline's ACU count for the 24 hour period on Saturday. If players had more data about server status and player count, beyond just the current users online, we might be able to standardize the ACU count by removing deviation from server interruptions, but we don't (and it wouldn't ultimately remove the other problem that some players on some occasions don't log back in after getting disconnected, whereas they would have kept playing otherwise).
II. Multiple character training on a single account fundamentally changed how many players use alts
Prior to the ability, having an "alt" usually meant having a separate account. A player with 3 characters in skill training at the same time would have 3 accounts and could then log in to each at the same time (and EVE uses so little computer resources that there is little reason not to have multiple accounts open at the same time if you have them). This presented a highly variable set of data to the ACU count, where one player might be registered as having two or three or however many characters logged at the same time. While many people still have alt accounts, there was a fundamental change with the ability to train characters at the same time. Now single players are less likely to be logged into multiple accounts at once, having one account per alt running even if not actively playing on those accounts. We don't have data on how many players have activated multiple character skill training, so we don't ultimately know whether it has made a serious impact on the number of accounts total--personally, I've kept my account total at two, while I've trained many months of skills on separate characters. If I didn't have that option, I likely would have gotten a third account, and I usually have all of my accounts open at the same time.
III. ACU does not track player activity
A basic limitation of the ACU count is that it flat-footedly tracks players online, regardless of what they do in game. 30k players out in space is not much different than 40k players online, 30k in space and 10k afk in stations. ACU time gets inflated by afk players, in station or in space (such as hundreds of pilots sitting on a Titan afk, waiting to see if the enemy fleet will form up). The aim of many changes to EVE over the past 3 years has been to get people logged in, out of the station and into space--from making every ship worth flying, to adding new exploration content, to making more content drivers in space, and so on. Though the ACU count is lower now than it was 3 years ago, other metrics such as ship losses, NPC kills, and jumps, suggest that the reality is that less players are online, but they are engaged in more activity in space compared with their 2011 counterparts.
IV. When unlimited character training is released in November or December, expect the ACU count to drop further as a result, even if activity in EVE goes up overall.
The ability to plug in literally hundreds of days of skill training on an alt will likely cause the ACU to noticeably drop even if, after Phoebe and Rhea, a lot more people are actively playing EVE. Training a new character even just for skill training is hard work, requiring logging in multiple times a week to update the queue. I have plans for some fairly low maintenance alts--plug in skills, wait 3 months, come back to a highly useful character. Skills with missing pre-reqs still cannot be plugged in, but the ability to pile on week long skills on top of week long skills means that alts in training are going to be very low maintenance.
From various metrics I watch--player activity statistics, total market data and activity, new characters created, etc.--player count in EVE looks to be on the uptick. Many quality of life changes to EVE continue to have side-effects on the ACU, typically lowering it. Even the new multi-sell ability saves time--more time for enjoyable activities in game. As far as I'm concerned, an hour playing EVE now is worth quite a bit more than an hour playing EVE three years ago.