Zonureskin map bots: The latest blight on the economy.

[Vahlnir]

And, as I've said, I believe a schism on the part of SE as to whether they really oppose botting, etc.

They've never said they supported it. Quite the opposite, in fact. However, their special task force is laughably small for an MMO of this caliber. So...we pay a sub to get the free to play treatment in some respects. Like this one.

Because there are people within the company who are in league with them.

There is no other answer, neither logically nor if you take one honest look at the history of the Square-Enix MMO's and what it has literally taken to get even a remote punishment on that level of player

Might want to check that tin foil hat of yours. Seems to be on too tight. Not good for circulation.

[KageTokage]

I'd like to think that the actual issue is that the STF being so horribly understaffed means they're incapable of directly investigating every report in person (Which is the easiest way to ID bots due to their erratic behavior), meaning they instead rely on flawed, automated methods of detection that only catch the most blatant of offenders.

That said, I don't see why they couldn't just broaden what behavior the GMs are capable of acting on. Literally all it would take to get rid of most bots would be to simply contact the suspect individual and ban them if they refuse to respond after several attempts.

[Packetdancer]

There is also a limit to how much you can automate in terms of dealing with bots.

You make it so that if someone's location changes by too large an amount at once, they get flagged. Sure, that's easy! But then you discover that if you make that amount too small, you catch real players who have network hiccups. So you make it larger, but the bot authors soon figure out what that threshold amount is and have their teleport-y bots move by just a tiny bit less than that amount.

You want to flag unusually high prices on the marketboard, when the delta between current average price and top price is too high? (I.e. when something gets listed for like 800 million gil, presumably as a way to transfer a giant amount of gil to someone for RMT.) The RMT folks will figure out what the formula is, and break all their sales into things just below that threshold.

You make it so that it tries to spot seemingly 'randomized' names, and you have to make it so that it doesn't also erroneously flag... I don't know, canon-appropriate Sea Wolf Roegadyn names. (And then they probably just start pulling from the random name generators on the Internet and trying name after name until one isn't taken.)

It's like anti-spam software; every new obstacle you put in the path of spammers, they will work out a way around. For folks making real money off of it, their literal livelihood depends on it, and they are most likely willing to put all their waking energy into circumventing whatever's put in place to stop them from doing so. Conversely, Square-Enix cannot afford to spend every waking moment on finding ways to shut them down; they have the entire rest of the game to make as well.

Mind you, I think they could put in a little more effort; they seem to have decided because they can't automate catching everything, it's not worth putting many tools in place to catch anything. And that, I disagree with, especially since it seemingly makes this game one of the easiest target for RMT gil-farming stuff like the trash we presently put up with... as well as the people who use bots to gain unfair advantages. (They definitely need more people for their non-automated task force.)

[MariaArvana]

That said, I don't see why they couldn't just broaden what behavior the GMs are capable of acting on. Literally all it would take to get rid of most bots would be to simply contact the suspect individual and ban them if they refuse to respond after several attempts.

and you could just as easily false flag innocent people by doing this you realize. You can turn off/minimize your chatbox and turn off game sounds, rendering someone incapable of ever realizing they're being contacted. GM thinks unaware person looks sketchy, sends them a few tells, person never responds, bam. innocent person banned.

Which is basically the reason why banning is slow/never. Square's system is setup that it requires an extreme amount of evidence for someone to be banned, hence why things like stuff said in text chat is so easy to get banned over compared to botting, due to text chat being concrete evidence on their server logs. It doesn't matter if it looks like a duck, if it sounds like a duck, Square won't ban them unless they 100% know for sure it is a duck, and if they think even the slightest bit its a goose, they'll continue investigating and won't convict.

[KageTokage]

and you could just as easily false flag innocent people by doing this you realize. You can turn off/minimize your chatbox and turn off game sounds, rendering someone incapable of ever realizing they're being contacted. GM thinks unaware person looks sketchy, sends them a few tells, person never responds, bam. innocent person banned.

Which is basically the reason why banning is slow/never. Square's system is setup that it requires an extreme amount of evidence for someone to be banned, hence why things like stuff said in text chat is so easy to get banned over compared to botting, due to text chat being concrete evidence on their server logs. It doesn't matter if it looks like a duck, if it sounds like a duck, Square won't ban them unless they 100% know for sure it is a duck, and if they think even the slightest bit its a goose, they'll continue investigating and won't convict.

GMs have ways of getting someone's attention outside of just sending tells, like temporarily shoving them into the gaol if they're being unresponsive.

Doing such would also likely cause an actual bot to glitch out and start trying to spam teleport/return to escape which would just make them even more obvious.

A professional team isn't going to allow for false positives, but there's obviously a lot they could be doing better then they are now when the average bot operates for unreasonably long periods of time and exhibits patterns of behavior that are far too consistent for any actual player to duplicate. They're clearly being way too lenient when there's seemingly no way to consistently get a bot of any manner banned, not even ones that are stuck in place repeating a cycle of movement/actions and not going anywhere for hours on end.

I think an "ideal" screening process would be:

1. GM receives report from a player citing suspicious behavior.

2. GM goes to observe the suspect player to see if they're actually demonstrating said behavior.

3. If player is deemed suspicious, GM attempts to communicate via tell to see if they get a response (By asking a question of some sort to ensure they're not giving automated responses).

4. If player does not respond after a certain amount of time and/or number of tells, they get shoved into the gaol and given a further window of time to respond to whatever the GMs asked and will be banned if they continue to fail to react in any manner.

I think that would be reasonable and prevent false positives as no actual player is just going to stand around and ignore their chat after being shoved into the gaol.

[ReynTime]

They've never said they supported it. Quite the opposite, in fact. However, their special task force is laughably small for an MMO of this caliber. So...we pay a sub to get the free to play treatment in some respects. Like this one.



Might want to check that tin foil hat of yours. Seems to be on too tight. Not good for circulation.

This happens in other MMOs, why would XIV be corruption-proof? It's not an alien concept.

[StouterTaru]

I've also noticed a steep decline in the amount of treasure map groups present in the party finder. I've been trying to get a group for weeks, now, with little success. During prime time hours, too.

There was a new zone added 7 weeks ago, which caused a huge spike in interest. Two weeks ago Firmament ranking season started, causing a huge drop in interest. With that now over, you'll probably start to see more parties, but it won't be at mid-December levels until they add something new to maps.

[Packetdancer]

I think an "ideal" screening process would be:

Honestly, I suspect one of the problems is that they probably would get a ton of suspicious reports and have few enough people to handle them. Putting on my systems engineering hat, how I'd probably do it to try to minimize the need for human action (given that potentially huge amount of bots) and thus keep the workload manageable would be something like:

1. Define a number of 'red flag' conditions: frequent 'stutter step' movements over a certain amount, mining the same nodes repeatedly in the same order for more than a given length of time, any point at which your location remains below the ground or in the air (minus a mount) for any length of time, has had suspicious behavior flagged by five or more players, etc. Make as many of them as you reasonably can, but remember that none of these red flag conditions on its own would be enough to trigger anything.
2. If a character trips four different red flags in a given week (i.e., reset-to-reset), flag them as suspicious and start recording more in-depth analytics on them. Where they're going, what they're doing, etc. You could apply additional red flag conditions here to winnow it down further, but honestly, this is where I'd go a bit more modern and apply some machine learning: develop a model to recognize suspicious behavior patterns in these analytics logs, and then feed the log through to that model. Create an 'incident' record to track this, containing the initial red flags and the analytical data fed to the ML model.
3. If the model can be fairly certain that it's bot/RMT/gil-farming/cheating behavior, flag the account, and temporarily ban the account (for, say, a week) along with an email containing the incident record. If it's a false positive, the player can email a dispute and the human team can open the incident, review the accumulated data and act accordingly; if it's truly innocent, they can clear the incident from the player's history (mark it a false positive and close it). If you get tempbanned (and it's upheld rather than marked as a false positive) three times, you're permabanned.
4. If the machine learning model can determine with a high degree of confidence that it's genuinely innocent behavior, it marks the incident as a false positive and life goes on.
5. If the machine learning model isn't certain either way, it flags the incident and escalates it for human review... at which point it gets handed over to the bot/cheating/whatever team to investigate by hand.

Constantly running machine learning on player analytics would be prohibitively expensive (not to mention horrible in terms of data storage requirements), which is why you require multiple red flags to escalate an account into being reviewed that way. That would keep the number of cases needing humans to hand-review down to a minimum, while still hopefully catching most of the egregious bots; most of the obvious bots we see would quite quickly trip more than one red flag, and by requiring multiple red flags to automate the process, you'd be less likely to get false positives.

And if someone is constantly getting incident reports but managing to convince folks it's a false positive... when you see 10 false positives on an account or whatever, that can also raise a flag for human review: either the person is cheating but is good at convincing people of their innocence, or the system's getting a recurring false positive somewhere in there and it's a sign you need to fine-tune the various criteria a bit more.

Now, you probably still want a way to catch folks that fall through the cracks, so... I don't know, maybe if 20 different people reported suspicious behavior and the suspect had tripped one other red flag, you could open an incident. Maybe even elevate it straight to human review in that case.

But that wouldn't be a short amount of development time to implement; it would save a massive amount of work in the long run, but at the cost of a considerable development investment in the shorter term. Though I suspect coming up with the machine learning system would actually be a pretty interesting project that would appeal to someone in the company.

(...now I'm actually starting to think this would be a worthwhile tool to write in a general, reusable manner for MMOs to license and make use of. Yeah, they'd need to train different ML models per-game, but if you had a defined way for the system to interact with those models, and a defined analytical log format, that'd be far less of a time investment than developing the entire system would be...)

[Jennah]

Because there are people within the company who are in league with them.
*Additional conspiracy theories*

No one with a higher level than "Summer Intern" is going to jeopardize their paycheck for status in a video game.
It's vaguely, remotely possible that someone at SE is getting (a lot of) US$ from RMT ... but they wouldn't need to deal with bots for that. They can just make gil. And sell it. For money.

[Kes13a]

why is it every time someone expounds on the potential of false flagging I see someone likely going to be banned for botting?