Bot count, the sequel: 52%

Well the only way around that is to have more people do the surveys, although for some strange reason those running bots seem reluctant or uninterested
on doing so?
Personally my home sim has 10 bots and around 0-4 people. And as long as the neighbours bot numbers don't increase effecting my operations, I don't care anymore, traffic for her means traffic for the sim and my rentals. I'm tempted to run my own.
LL have a better idea but won't publish the results of course, all they have to do is count the people who don't log out after 12-24 hours.
It wouldn't take rocket science to gather that most of those are bots.
I don't think LL will do anything til the figure becomes outlandish enough to cause real damage when the media pick it up.
So seeing LL are condoning bots use in traffic I reckon everyone should just join the bot arms race and keep piling bots online till they do something or their usage is negated by everyone using them.

I surveyed a band centered on my home, which is in no way special except that's where I was when I started the survey. I surveyed a band four sims wide for the entire width of the continent precisely to eliminate individual sim choice from the equation.

I'm not a scientist and, unfortunately, neither are you. The second sweep would have been much more useful if it had used a different set of sims - like people have been saying - more raw data.

Also, you can't say that you did either of them at certain concurrency levels because the first one took 10 hours and there is no high or low concurrency that lasts for 10 hours. Presumably the second one took the same amount of time. Not only that, but a snapshot of avs can't take 10 hours. What is the situation in the first sims several hours later, for instance? Or what was the situation in the later sims at the time you started? But you've presented your data as a snapshot.

I do not use unresponsiveness as a criteria.

Fair enough, but without that knowledge, even the start was suspect.

Anya, good job, thanx

The other way I've seen recently with CAPTCHA is it says a series of letters or word through your speakers that you then type in.

I'm not counting bots. I'm counting bots and campers. I don't care if you're running a bot client or not. I care if you're acting like a bot.

My objective isn't to use a survey to suggest bots should be banned. I don't think they should be banned. I run one myself.

If an avatar is logged in for significantly longer than a person is at the keyboard they're doing a traffic bots job, so they are a bot.
People can't run 19 alts camping on sleek and say they're not bots because they chat to people on them sometimes.
AFK 24/7 RPing sleeping? Bullcrap, because if a avatar walks into the bedroom with a shotgun and shoots out the ceiling then demands money, they fail the RP test.

That's only one criteria, and it's not a very good criteria, so it can't be used in isolation. People who run small numbers of bots often try to make them look real. Some people use 2- and 3-year-old accounts as bots.

/me eats Meade's cookie

Loool Anya, good for you haha!!! First you got bashed with your 44% post and now you strike even harder with 52%! Hehe, well done.

Ideally, a survey would be an instantaneous snap, but realistically it takes hours to acquire the data. The second survey went faster because there were fewer avatars and I remembered where to find some of them, but it still took hours.

I only said the first survey captured the peak and the second captured the minimum. Given the survey method and the times I quoted, I obviously mean that the peak happened during the first survey and the min happened during the second.

I won't accept "it can't be done" and I won't accept "it's not perfect so it's therefore meaningless". It's an informal survey that still has more scientific rigor than most of its critics, and it's the best that I'm able to offer for free. It is what it is and nothing more.

same here

Here's what I last said about determining if an avatar is a bot:

/327/cc/293027/4.html#post2223537

The short of it is that I don't use unresponsiveness as a criteria.

I can easily find 219 sims that have zero bots and do a repeated test and come up with the same results. So my results would say 100% real people and 0% bots. Thats the main problem with this test.

The amount of bots is not as relevant as how they are used.

I didn't say that "it's not perfect so it's therefore meaningless". I said it's meaningless as it stands. The fact that it's not perfect goes without saying.

I'm afraid I don't see anything scientific about it. And what are you offering for free? The titles you chose for the threads seem to imply various percentages of the SL population are bots, so you seem to be offering specific conclusions, which cannot be extrapolated from your data. Your initial posts seem to confirm that. I see such things as being responsible for false rumours - people believing what they read without giving it any thought. Before you know it, people will be saying that it's been shown that ~50% of SL's population are bots, when no such thing has been shown.

You said that you were simply trying to see which way SL is going, but your initial posts in the threads didn't say anything like that - they were written as, here's the data, therefore the SL population consists of nn% bots.

Other people in the thread, who seem to know what they are talking about and who support your efforts as far as it goes, point to the fact that more sweeps are needed, and a much more scientific/organised approach, and I agree with them. I'm not against the population consisting of ~50% bots - I've no desire to undercount bots, as I'd like to get rid of them altogether (traffic bots, that is). I'm only talking about the shortcomings in this sweep, and why the data can't be taken as any sort of indication about the % or number of bots on the grid.

I'm on the Corsica continent. There's loads of activity here, maybe because it's one of the newer continents and most of the land hasn't been abandoned yet. It's not hard to map jump around here and see lots of avatar clusters, including a few bot boxes in the sky. I'm looking at about 20 of them in a box right now 750m up in Halfclaw. It's not hard at all for me to believe that 40-50% or so of the avatars on this continent are bots.

I'm amazed at the level of meanness that some of you are going to in order to discredit Anya's methods. All she's doing is reporting what she's found, which is more than anyone else has bothered to do.

I am wondering if there is a very significant difference between the usage of bots on mainland and on islands. Although I would think people would use fewer bots on mainland due to sharing resources, maybe the large number of people who have their little lovenests on open sims balance out the overall number of bot usage on islands. As we know, islands make up a much larger percentage of the overall land mass than mainland.

Anya invested considerable effort in collecting these data, and there is information here that I certainly didn't know before. I'm grateful.

It was particularly good that the same sims were sampled again in the second sweep; these results put to rest the question of whether the numbers generalize over peak and non-peak hours. We now know they do.

We don't yet know how well they generalize across different samples of sims. That's a broader question. In the absence of more data, I suspect there'd be some variation across samples on different Mainland continents, but my guess is it wouldn't be too dramatic. It's just a hunch, but my guess is that Estates would have very much wider variation, with some sims holding over 50 bots some of the time, some with more modest counts, and most with no bots at all; I won't hazard a guess how they'd average out, but the larger the variance, the larger the sample size needed to get any confidence about the mean. All that guessing is by way of saying that Estates would be more challenging to measure, I think.

This latest sweep also uncovered something I've noticed recently: Some bot farms don't run 24/7, but rather on some other regular schedule. Possibly the owners are interested in getting the traffic boost during hours that are off-peak for their target customers, and then hiding the bots completely so customers can't see them when they arrive. At least in the places I've observed this, it seems to be all or nothing, and doesn't seem to be related to the actual count of other avatars on the subject sims.

Residential estate use: a fully loaded, mostly residential private estate region averages to something like 2 avatars in it 24/7; 4 or 5 at peak typically. A residential openspace is usually less than 1/2 an avatar average - in fact most are stunningly 'dead' with rich oldbies in them who don't log in much. On maybe 30% of these, there is someone that likes to throw a big party now and then, so you'll see maybe 50 avatars all at once on a Saturday night or something, but this is maybe once or twice a month tops. This is all readily and instantly confirmable by a glance at the map, if you truly know what's going on in the regions you are looking at. Given enough time, if you manage estates you can even know who everybody's green dot is, know who are visitors by process of elimination, and even tell how certain businesses are doing from week to week with scarily good accuracy.

Commercial private estate land is a bit different, and really depends on the products. Some regions that seem *shockingly* dead are generating incredible amounts of revenue - why? Because it only takes a moment for people to port in, buy their hit product, and leave. Other products and industries have longer dwell time by nature (say, texture shopping) so... you never really know what you are dealing with there.

And of course - there are the places that seem to always have 95 people in them, but it's solid bots or campers.

I think the amount of bots is very relevant, it shows the size of the bullcrap residents online figure is and the traffic stats for many places.
Perhaps if many people surveyed more sims we would have a more acceptable answer, but of course bot runners aren't interested in doing that for some strange reason.
Besides it doesn't take a rocket scientist just a tour around SL's green dots on the map, to compare the amount of packed people places with the amount of bot running commercial properties and the fact that bots generally run 24/7 to work out somethings not right about the resident numbers.
I think OS actually made the bot problem worse by spereading the population out more causeing more stores to run bots for their islands rather than rely on neighbours walk by traffic.

We all should know by now the currently on-line, the total number of residents and pretty much any other number LL posts are crap. Nothing we can really do about it. Bots or no bots the numbers would still be crap because of alts anyway.

I standby my comment. The amount is not as important as what they are used for. Bots of all kinds are out there. Chat bots for clubs, traffic farm bots, bots for modeling clothes, bots for land barons, bots for data collecting, camping bots and so on.

What they are used for is the key for their demise. If they can decide what is a legit use and what isn't then they can make it a TOS/Community Standards violation. The only reason I can think of why bots are not a TOS/Community Standards violation is LL likes the added concurrency and the boosted resident numbers.. I still here them bragging about the 15mill+ number even though only 1.3 mill log in every 60 days and of those at least half are bots and alts.

We are just beta testers for the platform they will sell and support in the future. The latest blog post is pretty evident of it as with several past posts from Phil Rosedale about making it a business/education platform.

Overall, I think this a good start for further research, but is too limited to show anything more right now than how many bots are on that continent. Much more research would be needed. It probably varies greatly.

And for anyone who wants to include campers, what about the people who hang around lucky chairs waiting for their lucky letter to come up? Places that have a lot of lucky chairs with good prizes can attract and keep a very large number of people. If they're too busy looking for their letter to socialize, but are obviously active avatars, is that the same as being campers?

Personally, I think that only campers who are bots should even be considered. Just because someone is using green dots to try to find somewhere to socialize doesn't mean that there's anything wrong with those green dots turning out to be campers. You just have to find out through experience where people are socializing and go to those locations.

And if anyone can point out a nice location for finding hidden bots (inside a wall, for example) I'd love to go check them out. I once thought I'd found an example of hidden bots because I couldn't find anyone around but saw lots of dots, but it turned out everyone was camping on a sky platform (and most of them were actually "there" not afk). Just curious to see it. You can IM me if you don't want to post it here.

Overall, I think that even if you didn't like the way the study was done, if you have at least read some of this then it has brought your attention to the topic, offered you ideas about how you could look into it yourself, and encouraged us all to try to define what a bot is - all of which will be beneficial for any further studies.

Intriguing data, Anya. I hesitate to say this in this thread (and apologize if it's been brought up before), but wouldn't it be much easier to collect a much larger mass of usable data using a ... bot?

The bot could be given a large list of regions to visit, and be wearing an attachment that uses the llGetRegionAgentCount command. The bot could visit each region at least once an hour, and keep data points for each region for the most recent 6 hours. Regions that report 6 consecutive samples of greater than 5 avatars would have their stats reported to a web service that would store the data in an RRD database for later review by Anya or whoever. (I suggest RRDtool because it readily produces graphs that can be easily glanced at to see if the sim traffic patterns appear to be suspect; e.g. if there appears to be a nonzero "floor" to the number of avatars in the region).

Sure, a few sims might be counted as having bots that have none ... e.g. an ETD, Simone or Dazzle sale; a highly successful performance venue. But that is comparatively rare. Plus, nothing would preclude the person running the study from visiting sims and excluding successful performance venues (for example) from the bot count. Over time, the counts would get more and more accurate, as the researcher became more familiar with the shape of the data (perhaps the best trigger to suspect a sim of having bots isn't 5 avatars, perhaps it's 10), and also more knowledgable about the sims that have high real traffic.
.

If LL were to release their estimate, what would be the minimum bot percentage that they could report and not have people tell them that they are wrong?

I'm curious why so many people totally ignore LL's 10-15% estimate. They are the only ones with the kind of data needed to really get a wide-angle look at the numbers. I know it's quite popular to bash LL, but I've always felt comfortable with their other published stats - even if LL does tend to spin their actual meanings.

I think that's it likely that in some areas traffic bots are 50% of the population, but I think that this is balanced out by other spots that have no bots. There are an awful lot of educational, corporate, residential, and personal sims that have no need for traffic bots.

We also need to keep in mind that LLs estimate is for User Hours not population. So I suppose if 100 people logged in for 1 hour each during a 24 hour period then that would rack up the same number of user hours as about 4 full-time traffic bots. I wonder how many people log in every day as opposed to how many people are online at any given time? If there are say 300,000 unique log ins every day then their total User Hours could easily tip the scales of this sort of measurement.

Of course I am also curious why LL don't seem to care that there are a large number of traffic bots sucking up our resources. Seems pretty obvious that something needs to be done, and also obvious that removing Traffic numbers would a good start. Do we really need an arbitrary and easily manipulated number to tell us where we can supposedly find the 'good stuff'.