Bot count, the sequel: 52%

I don't see any point to these threads, especially as they don't show anything that means anything. It would be interesting to know how many bots there are on the grid, on a daily average over a week or so, but this sort of sweep isn't going to give any such indication.

If the purpose is to plug away at LL to get rid of traffic bots, then this isn't going to do anything. What's needed is plugging away at the Places tab search, which is the only reason for traffic bots. Not long ago, LL were very close to doing away with traffic in that search, so it might not be too difficult to prod them into doing something. Rather than spend 10 hours on doing a sweep that produces nothing of real value, it would be better to spend the time at office hours, plugging away at the Places tab, and even for organising it with multiple people there.

Well done Anya,

this confirms exactly what I see in the sims surrounding the private island sim that I live on, (so it is not confined to the Mainland).

Forget the naysayers, I suspect a hidden agenda there, and of course, they produce no data whatsoever, and it is also apparent that they don't know the first thing about statistical analysis, sample sizes, and margins of error.

Even if you tested every single avatar in every sim in SL, 24/7 for a year, to see who was a bot and who was not, PD would still find something to moan about (there is something seriously wrong with that guy, who jumps on all avatar/bot raw data gathered, very suspicious).

Keep up the good work, and thanks for sparing your valuable time in gathering this data for us.

Vi

PS It did make me smile when you counted the number of avatars on this sweep,

591 avatars, 308 of them bots and campers
52% bots, overall

identified the number of bots, and gave us the (correct) 52% figure, and some idiot still asked '52% of what?'. I hope you include idiots like that in your totals.

One thing that seems odd to me is that I've been working on an automated system to detect and count bots. We have this network of 16sqm parcels in over 1000 sims. Anyway, the project isn't working yet and maybe I'll never put in enough effort to finish it. Either way, the weird thing is that I went through about 30 or 40 sims testing this out and I saw about 10 avatars and presumably none of them were bots. How are you finding these very busy sims with thousands of people in? Whenever I move around the mainland I find it's usually 1 or 2 people per sim with very rare busy sims.

hey!


i have seen people stay for 1 -> max. 5 minutes in my shop on mainland, grab the item, than they are off. so my traffic is near to 0
well i have no bots running ...

KK I will catch up with you later

Excuse me....


HAHAHAHAHA.

OK, I'm done.

You may find it suspicious, but your suspicions are unfounded. I have shouted more than most people here to get rid of the one thing that makes traffic bots useful. The reason I nay-say these two threads is because they don't show anything worthwhile. I would really like to know how many traffic bots there are on the grid, just out of interest, but these sweeps don't provide anything worthwhile. If the population was spread homogenously throughout the grid, the sweeps would be meaningful, but it isn't, and the sample is a lot less than 1% of sims - it's just too small for a lumpy distribution of avatars.

Then too, if your advertising is working the way it should, people are going to do exactly that. TP in, get what they want, and leave.

Maybe we should be selling our "low traffic" numbers, instead of the large ones

Thanks hun!

I hope you do put the time in to finish that and run it. It would be very interesting to have some reasonable figures. But even then, it would only be interesting. The main thrust to remove traffic bots should be directly at LL, and should be aimed at changing the Places tab search - or things will get worse.

OK so you choose to personally call me an idiot? Thanks for insulting me personally.

I asked the questions because what I see are skewed figures that are complete bollocks.

What exactly is the test that someone is a bot or a camper? could I have failed it in the past (IE could I a real person have been called a bot or a camper in the past)

I live on mainland, travel to various locations, and I would state that most of the AVs I meet are real people, they are generally talkative, so in my experience far less than 52% of people are bots.

The sample is not representative (less than 1% of sims are checked and only on mainland), the criteria for being a bot is an interpretation but it's presented as FACT, when it clearly isn't.

Oh and my hidden agenda? You tell me. I own the land a club is on, I own the land Dani's mainstore is on, and I own some of the land I call home. I own no bots or campers and will not entertain them on my land.

So come back to me when you have cold hard facts. Then maybe I'll listen.

Rha, if you dispute the methods or results, then counter by doing your own survey, giving us more data points.

Given the method, there were only two subjective aspects:
1) selecting the region to scan
2) detecting bot vs. non-bot

My guess is that regardless of inclination, if honest people perform the same survey, they'd come to very similar results concerning item 2. I mean, it's pretty obvious in most cases.

I'm tempted to do a survey myself.

Science is never completely objective. That's why it's important for others to reproduce results.

Actually, it doesn't matter. If someone is trying to find somewhere to meet people and goes to five locations in a row that appear to be popular but where everyone is silent, it doesn't matter _why_. Even if they were all real people in IM conversations, the negative experience stands.

Sorry you are confusing science with statistics.

Science is completely objective, that's why experiments in identical conditions can be repeated time after time and the results will be the same.

Statistics aren't completely objective

Actually it does matter. People could be labelled a bot when they are not.
What happens if bots are banned? these people could get ARd.

as I mentioned in another post, I would not do such a survey or whatever it would be called, as I have no way to ensure my interpretations would be honest and true

I have no sure fire way to know that an avatar is a person or a bot, or if I am getting a good cross of the "population"

Nor do I have the interest, or the desire to run such a unverifiable conglomeration of data that in my opinion would be more misleading than it would be worth.

At a minimum, I view this data as being useful in the same way that a single average trade price on the Lindex is useful. In other words, by itself it may not be much, but when combined with additional data, it can be very helpful.

Statistical work often goes like this. You formulate a survey, you get some initial results. You study the results for consistency, you study the methodology for flaws, you allow others to review the methodology for flaws. However, flaws in the methodology don't invalidate the raw data, they just limit the conclusions that can be drawn. The raw data remains a factual data point that may contribute directly to future analyses or may contribute to an improved methodology.

Reproducibility is a critical requirement for this sort of research. In other words, the survey should be described well enough so that another person could conduct it and come up with the same results as the original author, given the same source data (i.e. same sims at the same time).

In this case, we need a reproducible definition for identifying a bot. It doesn't have to be a perfect definition - in other words, it's ok if people disagree about how accurately it identifies bots. What's important is that we agree on applying the definition, so that two people going to the same sim, looking at the same avs, same environment, same chat logs, same whatever, will agree on how to classify the bot according to the definition. If that's not possible, then the variation needs to be understood and factored into the results.

Please specify a better sampling method. It's not helpful to simply say "this is wrong" without saying what's wrong and how to correct.

The good thing here is that we have full visibility on what the methods are and where they were applied.

I doubt it's representative, but it's a data point. We need more. Let's go get them, and make it perfectly clear how the results were obtained, so people have a clue how they might be skewed.

The results here are clearly specific to a region of mainland. That region may or may not be representative; we have no way of knowing. Those facts are evident from the initial posting. That's how science and stats are done: you give your methods, your conditions, and your results. Other people apply the same methods in different conditions and report the conditions and results. If the test methodology is any good, we get relatively objective results.

For folks who think it's terribly hard to detect a bot, I suggest you haven't been looking much. But, for scientific purposes, it would be best to characterize them. If I see a large number of newbie-looking people, many with forumlaic names, standing together in a hidden section, doing nothing, with nothing to do or see, I'm convinced they're bots.

If I see a lone newbie who seems afk, has some numbers in his/her name, and a relatively old birthdate, I'd put that in the questionable category.

BTW, the method used in the OP's scan counts only 'camper bots'. There are lots of other kinds of bots (e.g., scan bots) that wouldn't be counted. My suspicion is that those are a minority, though.

Agreed Rha.

The only people qualified to do a survey are LL.

They have access to data that con conclusively prove whether AVs are bots. They either.

a) Don't have the knowledge/skills to compile this data

b) Can't be bothered.

I suspect it's b)

But that is still subjective. LL actually could get an accurate figure very easily by extrapolating IP data from it's server logs.

While I frequently disagree with PD and admit that he has a vested interest here, and I disagree with him here, I have seen him make admissions against interest (which is all too rare in any argument) and find that he is capable of aguing rationally. I suggest you avoid the ad-hominem attacks.

But I agree with you that the OP's post is informative and useful.

Bots wouldn't be ARable, they'd just be blocked from connecting.

As I've said, the easy way to do it technically (other than CAPTCHA) is for the server to occasionally ask the client program a question, the answer to which requires it to be running the full graphics engine. (Eg, ask it to rebake textures, ask it the color of a particular pixel on the currently displayed frame etc.) Most bots don't and if they did it wouldn't be viable to keep hundreds running on one PC.

This second thread would have been far more meritous if it had been a sweep of a completely different set of sims. If 44% of avs in the first sweep were actually bots, then it's likely that they will still be there during the second sweep, which they were, more or less. The sweep was much too small for the size of the grid and the unevenness of av distribution.

Another point:
Judging by the titles of both threads, the OP appears to be intending a downer on bots. So how did she choose the sims to sweep? Was it that a look at the map showed plenty of avs in one particular row (4 sims wide), with a fair number of stacked dots, so she decided to sweep there? If that's what happened, it's the same as stacking the results. Anyone can pick rows of sims from the map that, judging by the green dots, are likely to contain a fair number of bots. Not only is the sample way too small for an uneven distribution of avs, but it may well have been tainted to start with.

That may well put more of a strain on the infrastructure than bots do. I suspect LL would take the cheap option and ask for residents to AR bots.

Think about it residents run the client at different resolutions, meaning different pixels are different colours on different PCs. How would you know what resolution they were running at.

Rebaking textures I suspect sends back a confirmation code to state it has been done. Once that code is known, bots can be programmed in to send that after a predetermined time to answer the request.

So I just left my AV standing in my workshop for 20 minutes while I took my child to school. I would have "failed" any bot tests, as I woudln't have responded or replied. My AV was AFK in a skybox over land I own. I guess that makes me a bot?


Before you can claim who is or isn't a bot, you need to define bot. Then, you need a clear, consise, objective test for "botness". Walking up to an AV and saying "Are you a bot?" isn't an objective test, now is it?

Perhaps, I'm afk in world while I type here in the forum.

I don't know what percentage of bots are on the grid. I do know that in order to have valid statistics, you need a genuine sample of the population, an objective thesis, an objective test for said thesis, and accurate numbers. Of course, with any Thesis, your statistical "alpha" can be set to anything you choose, just giving the surveyer freedom to make the numbers say whatever they want. One can prove 2=5 and Coke is Pepsi and Green is Orange with statistics. It's all in the alpha, and in the eye of the statistician.

So, what is a bot. What is the objective test to prove an AV is a bot? Let's start there, shall we?