bot count: 44%

I visited 219 sims on one of the newer mainland continents (I think it's the one called Corsica?) and noted how many avatars were present in each (including the ones with no avatars) and how many were bots and campers. I made a four-sim-wide path across the continent, coast-to-coast. I noted cases where I wasn't sure whether or not an avatar was a bot so I could visit again in several hours.

The results are as follows:

219 sims
712 avatars, 312 of them bots and campers
44% bots, overall

For sims with 5 or fewer avatars there was an average of 18% bots and campers (42 bots out of 235 avatars). For sims with more than five avatars there was an average of 57% (270 bots out of 477 avatars). The overall percentage was 44%.

Some observations:

There are a lot more bots than there are campers. It'd be hard to get good numbers since some campers are bots and some bots are disguised as campers, but I'd put the number at about 85% bots to 15% campers.

The number of bots in low-occupancy sims is noticeably higher than it was just three months ago when I did my last multi-sim surveys. Small numbers of bots (fewer than a box-full) are being deployed more today than just three months ago.

I'm convinced clubs use bots but I don't have a way to determine with any certainty how many in a busy club are bots, so the number of bots is likely higher than I've stated. In this survey I found two busy clubs and a couple less-busy but active clubs, and I just had to assume they were all real, live humans, despite getting no response to my greetings and questions except from the host, and despite there being no meaningful chat. The number of club avatars in this survey is perhaps 60.

I expect the overall average for estates will be similar. Managed estates like Dreamland and Caledon surely bring down the average for low-occupancy sims, but fully-owned sims with large bot boxes are more common on estates than on the mainland.

There's more data here:

http://greendots.typepad.com/

44% of what? I'm in a sim right now where 22 of the 23 avs in it are bots, and as the day and night goes by, the ratio will change a lot. The point is, if you do your sweep during the low concurrency period you will get a much higher percentage of bots than during the high concurrency period. A percentage is meaningless.

Another thing that makes the percentage meaningless is the location. You used a small area of the least popular type of sim. Mainland is just a small fraction of the grid, and most people who are logged in are not often there.

Phil, I'm going to provide a bit more info and then I'm going to ignore you. Your position is known and we've debated this at length before.

My survey started Saturday morning Eastern USA time and ran through Saturday late evening. The peak for both USA and Europe were in there.

In my experience (I've done this kind of survey twice before, and have also made hundreds of random visits all over the grid, concentrating on estates) estates have about as many bots as mainland. There are a great many pancake estates with a couple stores, clubs and casinos, all empty except for someone building, someone standing in a field, a couple campers and a box of bots. Just like mainland. There are a lot more mega-boxes of bots on estates than on mainland, but there are fewer bots on low-occupancy estates because managed estates bring down the average.

Of course you'll find single sims that'll tell you anything you want them to tell you. That's why I visit hundreds of sims. If you check the data I posted on the blog you'll find sims that have large groups of humans and sims that have large groups of bots. Individual data points are meaningless.

Ideally I'd have a lot more data, but I'm not paid to do this. Still, I've put far more effort in to making this estimate than you have.

I have a pretty busy club and only 1 bot whom is a group joining bot. He has a job and does it. Hope that doesnt tip the traffic scale (smirks)
We do not have any campers and traffic is over 76,000



True there are a lot of clubs that use bots. with the massive infestation of bots and campers these days, a business owner almost has no choice but to use them to even out the playing field .
Hard to get noticed without

Traffic is only a small portion of traffic measurement now anyway.

With popular places gone now it really amazes me the continued desire to use such measures to increase awareness

Wow Anya!

Thank you for taking the time and effort to do this kind of survey. I for one appreciate a more accurate concurrency representation of SL than we are getting from LL or anyone else talking about bot numbers.

Way to go!

Just out of interest, keeping mind that I've been out of SL for over a year, how are we now able to tell who is a bot and who is a real person?

It's an interesting project Anya, I've checked your site out a few times.

It sounds like you think I'm in favour of bots, in which case you don't know my position at all. In any case, this is just an academic discussion of the number of bots, and not a discussion of whether they are good or bad.

Percentages are meaningless. High concurrency is ~70,000 and low concurrency is ~40,000, so 44% of what? You said that you swept during high concurrency, and concluded that the percentage across the grid is 44%, which means that ~39,000 of the low concurrency avs are bots (remember that bots are 24/7 avs), and that's at a time when most of the world is awake. I'm sure you'll understand that that figure isn't believable.

The only thing you can say is that ~44% of the very small area of the grid that you swept, during the period of time that you swept it, were bots, and that doesn't really mean anything.

There isn't one, particularly for short-term activities like in a club. If a club owner brings in a few bots for each couple-hour host shift I won't catch it.

Most bots are easy to spot, though. They're gathered around a camping sign and wearing a camping group tag, or lined up on benches with text in their lap, or stuffed into boxes above stores, or sitting at a desk in an empty office, or have nothing in their profile except the store they're in, or they're crammed into the peak of the roof or burried in a column, or they're being used as a mannequin, or they have a tag that says they're a bot, or...

Where there's uncertainty I make a note and visit again hours later. You have to check up on the security officer sitting at the desk in the empty office because it just might be the owner in IM or photoshop or whatever. If he's still there ten hours later, though, it doesn't really matter if he's running a bot client or not. He's acting like a bot. And most of them name their group in a helpful way. "Security" is popular with bots. In a club you expect someone at the keyboard, but in an empty office with "rent this space" signs all over the place he's probably a bot.

If I can't tell I assume it's human.

A group of scientists were travelling by train from London to Edinburgh to attend a convention. From the window of the train the mathematician spots a black sheep on the hillside. "Look," said he, "all the sheep in Scotland are black!"

"Actually," piped up the physicist, "all we can say is that one of the sheep in Scotland is black."

"Infact," quoth the astronomer, "all we can really say is that one of the sheep in Scotland is black on one side."

"As much as I hate to disagree," exclaimed the parapsychologist "all we can really say is that one of the sheep in Scotland is black on one side some of the time."

The End.

Can bots be trained to say 'RRRRRRUUUBBGGAAAA', *****TUUUNE*****, ...spanks Conifer's ass....... HOOWWWWLLLSSS?

So to clarify, taking into context the story above, 44% of the avatars present when you did your sweep are what you believed to be bots. It always helps to have context when dealing with statistics.

Yes, but it will cost you L$10,000

We've had this exact conversation before. Pick up a calculator and divide 312 by 712. Then read my post to see what those number are. If it's still unclear I'm just going to suggest you like being difficult.



It'd be meaningless in isolation, but I figure you're capable of looking for yourself. This isn't rocket science. You *can* do this at home.

If you make random visits, or make a sweep of many sims, you will find lots of bots.

I'd rather it had been the physicist who had got it right, but I'll settle for the parapsychologist

It's not like parapsychology is an actual science anyway.

Still, those 312 avatars may not be bots. They're just people you think are bots based on some very suspect criteria. Hardly a scientific way of going about things.

I don't remember having a conversation with you before, but I don't doubt you. And I'm not being difficult at all. It's purely academic and I don't agree with your conclusion. It's perfectly clear though - read Conan's posts for the clarity.

You said 44% are bots, but you haven't yet said 44% of what. It's assumed that you mean 44% of logged in avs, but when? Since bots are 24/7, it can't be 44% during high concurrency because that would mean all but ~1000 avs during low concurrency are bots, which is something that I can't accept.

Of course I'm capable of looking for myself, but I know that, unless I sweep the great majority of the grid, I can't come up with any meaningful data.

What if I sweep a different set of 219 sims and find that the percentage is 23% or 79%? On such tiny samples, the data will vary greatly.

For that sort of price I'd want the bot to be able to type out song lyrics in time to the music too

For 10,000L I'd spank your ass while howling and singing songs to you myself. Never mind having a bot to do it

Bots are suprisingly sophisticated these days. For example, I've managed to trick people on this forum for over two years into thinking that I'm a real person. Brenda; she's a bot too. So are Phil and Trout.

I stood on a sim with 80 other people last night...not a one of them being a bot. Your numbers are skewed.

I don't think a 44% number should be accepted as final, because the sample is at most 1% of the top concurrency. 5% would be a much better sample in my opinion (although, I admit I forgot most of my mathematical statistics by now).

Regardless, it still strongly suggests that the 10% number cited by LL is bogus.

It's a conservative estimate based on observation and common sense. It's not rocket science and it's not voodoo, and nobody's going to die if my numbers aren't perfect.