Bot count, the sequel: 52%

Oooo... that's maybe a problem. A lot of bots are clustered at weird altitudes, so to catch those, you'd pretty much have to send up a probe sensor. I'm not sure anymore what's in the prim on the 16sq.m. parcels... whether that prim could be the probe, or if its inventory contains a remote-scriptable seed prim, or... well, it's solvable, but not completely straightforward.

I fully agree that some of the Ops methods of data collecting is reckless and inaccurate, and the way Private messages work these days with no flashing warning on new messages. IM's will get ignored a lot .. Or answered late . Depending on what the resident is doing at the time, like here posting in the forums, reading blogs, working in photoshop, or busy with some killer cyber sex going on in the back ground.. hahaha

Hmm, I guess you're right. The project is doomed to failure. Maybe I can do it with a bot instead. At least they can see much further.

If I'm not mistaken, she said that she doesn't count club people because she can't tell one way or the other.

However, she does count mannequins as bots which, whilst they are usually bots, are not the sort of bots that people are generally against, because they usually serve a useful purpose for users. To be fair, though, she doesn't make any judgemental statements about bots, except that she gives the threads titles that can be read as being judgemental.

There are too many grey areas in this topic to be confident of any collected data. The idea of defining the criteria for determining whether or not an av is a bot sounds good, but the definition wouldn't suit everyone's thinking. Camping is especially grey. Is a camper a normal person who is just making a bit of money while s/he sleeps, watches TV or goes out to lunch, in which case it's not a bot. Or is it one of mulptiple avs that a person has just for camping, in which case it is a bot. The idea that some activities mean that a normal av is behaving as a bot, and therefore should be counted as a bot, doesn't hold up in my view. Alts are also grey. Is an alt a bot just because a someone has logged it in for a purpose, but it stands around doing nothing much or most of the time? I say no, but some would say yes.

This thread has certainly highlighted why it would be difficult for LL to automatically log bots out, as some people have suggested in the past

To the person who wants to see a bot farm:
Go to to the NW corner of the Seymour sim and fly up to 4080m. When you get there, you'll see a box with bots inside it, and a normal looking av standing on the top. That one isn't a bot but the others are. It's not a classic example because the bots are white clouds, but it is a traffic bot farm.

There's a very large volume of space to cover in a sim - 256x256x4096. It's quite a task to write a programme that gathers the data. Is there a way to get data (names and co-ordinates) beyond 96m?

You're absolutely correct: she should have said 52% campers and bots, or perhaps more accurately, 52% inactive avatars.

If that is the case than that would also create huge errors in the data. Clubs etc are one of the biggest drawers of crowds in SL. So mis counting them while counting the large colelction of bots above stores would of coursse make it appear that there are many more bots then there are becuase she has discounted the 20 or so avies dancing as a club as not relevant data.


Like i said. This study isnt even close to accurate. As someone who has studied statistics if you put this data into an assignment of some discription you would be laughed at and reieve a nice big F.

This kind of study needs to at least admit biases when it is being conducted in a manner where the colelctor of data needs to make a judgement call as to whether the data is relevant or not and whether it is in one categoryor another.

As to whether 50% of the avies logged in are bots? Well who know to be honest. But for those who are tking this data as anythign close to accurate look again at the methods used to collect it.

To be honest... the only thing it tells us is the OP thinks that 50%of the avies she saw and didnt discount because they were at a club (if that is what she did. last time i read one of these threads i am surre she said she comes back to clubs to check who is there) but either way, all it shows is 50% of the avies the OP saw and decided were bots were bots....

Sorry Phil, but you're on thin ice here. This scan was in direct response to a point you made, that the most accurate time to do a scan is during highest concurrency. Maybe it was less informative than a more random scan, but instead it shone light on the concurrency issue.

You can't criticize her for something and then criticize her for eliminating the very problem you raised, from her first data point. Note that this data point simply replaces the first one, rather than adding to it.

Phil, YOU can't say what she can or can't say unless you know what the facts are for the second scan. She didn't say, so we don't know. Since we don't know, that adds to uncertainty. Hopefully she'll answer.


It doesn't matter whether it's a snapshot or the accumulation of a number of samples. What matters is the conditions during the sample: what's the range of concurrency numbers.

I worded it badly. I meant that, if I'm not mistaken, she doesn't count any avs in clubs as bots, but she does count them in the total number of avs. That's if I'm not mistaken.

I agree with you about the data, as you know. In my last post but one, when directing someone to my bot farm, I considered writing a bit about the normal av that stands on top of the box. It stands there almost 24/7 and would certainly be counted as a bot, but it's me - my alt. It's logged in in the normal viewer and I keep an eye on the sim with it - even while I'm asleep at night. I also use it for doing things in the sim sometimes. There is nothing automated about it - it's just me as an alt, and it's not a bot, but it would be counted as one according to Anya's criteria. The 6 models in my store would also be counted as bots, but they are used by customers for a useful purpose, and are not bots in the sense that we usually mean, and yet they would be counted as bots.

Because there are a number of very popular sims that have active participants 24/7. Your method would provide numbers, but we wouldn't know what the numbers mean: how many of these sims are busy with active residents, and how many are bot farms? How many are both?

Also, how much of the online population is in nearly vacant sims versus populous ones? I couldn't guess, but based on looking at the map, it might be a sizeable minority, and more than should be ignored.

It might be useful to do this to collect a list of sims to investigate.

I don;t think I made that particular point, but I did talk about "48% of what", and pointed out that 48% of high concurrency is quite different to 48% of low concurrency, so percantages don't really mean anything. Yes it did shed some more light. It showed that, during the different concurrency periods when the two sweeps were made, the number of bots stayed pretty much the same, which tends to show that bots are pretty much 24/7.

I made a reasonable assumption about the length of time the second sweep took (I did say that I assumed it). She did answer it though. The second sweep took a little less than the first.

I came to that realisation myself - in a later post. I realised that, whilst it would be best to take actual snapshots at many times over many days, doing a sweep like Anya did can itself be done over many days, because the avs are changing all the time through Anya's 10 hour period, so it doesn't matter if it's done over, say, 100 hour period - they will still be changing, and it's only the non-changing ones that are of interest - something like that

Have you read her method? If not, you're being reckless. Her method has nothing to do with IMs.

The main valid criticisms are that (a) it's just one data point, and can't be assumed to reflect the whole grid, or even all of mainland, and (b) it lumps campers and bots into the same category. Which is fine, but it's important to be clear and not say "52% bots".

Personally I find this data intersting. If the OP is going to sweep another set of sims next weekend I would absolutely like to see it posted here so we can compare.
To anyone that questions the ways she chose the sims or the criteria she used to identify a bot... you are either not reading the posts or you don't retain what you read. Also... it seemed plain to me that when it was difficult to determine bot vs real person she favored real person.

Sorry, they'd give you the F for assigning a grade without having all the relevant facts.

She counted the population in clubs as PEOPLE, not bots/campers. Therefore, this simplification would cause bots to be undercounted.

There are no stats involved yet, other than calculating a percentage. Stats comes in when we have multiple data points.

I doubt that the number for the whole grid is even nearly 50%. That's why we need more data points. But if it takes 10 hours to get such a data point, we won't be getting many, and need a more efficient way.

\

I'm sorry, Rha! I in no way meant to insult your intelligence!

In this study, campers and bots are considered the same, which makes it considerably easier. That may not be what some folks want to count, but it is what the OP counted.

Your point is valid that it is very difficult to tell bots from campers, except in the obvious case of bot farms where no attempt was made to hide the fact that they're bots. However, this study ignores that distinction.

Here are the criteria she used: /327/cc/293027/4.html#post2223537/327/cc/293027/4.html#post2223537
I think that most people would come up with pretty similar numbers, following that kind of method, assuming they're being honest.

No. There were actually 18 bots that were turned on exclusively late at night which makes sense. If you run a store and you are smart, you kill some of your bots during the peak periods when most people are shopping so that they don't arrive to a laggy store.

if you want to put one in my store, to have one in that sim, I am ok with it

either send the object or come over and place it

tis ok, I am a bit edgy anyway, meds are kicking my ass, so maybe my best option is to not say anything on the subject or any subject for that matter LOL

not upset with you (or anyone) just not thinking very well either...
and I tend to word things badly to begin with, so I know some of my posts came out more harshly than intended


just because I put no value on her bot counting survey things, doesn't mean others don't, and I do not fault her, or them... I just have my own opinion about it, just like everyone else does

The distinction between 'bots' and campers is largely one of semantics.

A human who parks their one avatar on a comping pad and then leaves it there while doing something else is effectively running one bot.

Why the fuss about bots and camping bots?
1. They exist solely to dishonestly game the Traffic metric
2. In a largely deserted grid, the collection of dots on the map or the high traffic number if Search screams "people!!!" at the uninformed.

(2) is the most insidious aspect.
Now users will intuitively be drawn to what are apparently popular places - and be sorely disappointed.

You cannot of course definitively categorize agents in a sim as bots/campers or residents, but you can make guesses statistically, and if your sample is large enough the errors will average out.

There are a number of very popular sims, but ... the number is small! Over time, one could weed out the "popular" vs. "bot-full" sims and exclude them from the bots/camper category by name.

I also think that, having 24 hours of data for several days or a month, you'd see that popular sims have correlations to grid-wide concurrency that bots do not. One could check out such sims manually to see if there are bots, and, again, exclude them by name.

If anyone is interested, here's an example: -------------------
In "bot-full" sims, one could count as bots the "floor" of the agent count over each contiguous 6-hour period. For example, say you are processing the data gathered from 12-1 PM. If you had this data for the preceeding 6 hours at that sim:

7-8: 34
8-9 33
9-10: 50
10-11: 45
11-12: 30
12-1: 33

One would count 30 bots for the sim for the hour from 12-1. Next hour the data might look like:

8-9 33
9-10: 50
10-11: 45
11-12: 30
12-1: 33
1-2: 4

One would count zero bots for the sim for the hour from 1-2. This method would result in a conservative bot count ... but I bet it would still be higher than LL's 10%.
If collecting data via bot, the aim would be to ultimately collect data on all the sims that one could find, islands as well as mainland. It shouldn't take more than 15 seconds for a bot to TP to a sim and record the number of agents there (llGetRegionAgentCount). So, per hour a single bot could record data on about 240 sims, yes? Get a set of 10 bots doing this, and a pretty big database of one-hour points would be assembled rather quickly. THe trick would be getting an exhaustive list of sim names/coordinates to TP to, but that seems like a less difficult problem.

Anyway, enough of that. I just thought the results would be both quicker and (since a larger sample) more robust than doing it by hand.
.

Maybe it is, I'm certainly not a statistician. But it seems to me that the average traffic bot would spend far more time on-line per day than the average human. I know some people who are in-world a lot but I don't know any who are logged in 24/7.

QFT I agree entirely.

I agree with what Anya said in her first post, when she discussed this. AFK campers are hard to distinguish from bots...but we probably don't need to distinguish them. They behave in the same way, sitting around all day sucking up resources (and in this case, the campers are worse than the bots, because they suck lindens out of the economy without adding either value or RL $).

Nuke them all, and let God sort 'em out.

If I can't tell, I assume it's human. I don't discard any data.

In the first data set clubs accounted for about 60 avatars (of 700). If I recall there were two larg-ish clubs in that survey. In the second data set they accounted for perhaps 20 (of 600). I don't think there were any large clubs in that survey.

/me Puts her head in her hands... "Leave the Bots Alone!!" lol j/k I just saw that video of the guy cryin bout Brittany Spears again.. Please continue..