bot count: 44%

I agree

my first impression when I read the OP was.... here we go again, another bot thread to stir the masses

I do not feel that the info is factual enough.
Even LL admits they have no way to determine what is a bot and what is not, factually

and my conclusion is that the OP has done at best, a guess as to how many they think are bots

so now, I must ask.. what is the real point of this thread? What was it meant to accompolish other than rile up some folks and get the old bots wars going again.

my personal opinion about bots is .,... so what?

If I had the means to run some bots in hopes of becoming more known and getting more business I would do it. After all, it is just another tool as it were. They are not illegal, if they were LL would have found a way by now to determine a bot and thus eliminate it

As in any sample you have to take the statistics with a pinch of salt but Anya's project does suggest that bots are more widespread than the 10% figure we see thrown up.

Possibly, but it's hardly conclusive. By her criteria, I am bot. Right at this moment I'm sat doing nothing. I sincerely doubt that she observed each of these 712 avatars in her sample for an extended period of time individually.

Conclusive, I suppose not. Casts some serious doubt on LL figure of 10% and gives grounds for developing a stricter method and actually going through with using it - yes.

To the casual observer you would also appear to be a bot right now. So 100% of the avatars at the Cartel hangout are bots, by Anya's figures.

I think the OP means that 44% of the number of AV's in the small collection of sims she counted was 44% bots in her belief


Hardly a number I would use as solid or even close evidence that 44% is close to total grid bot population at any given time.

I tend to think it is more on the 10% that LL tells us.


After all, what % of the grid is 219 sims
and then, what would it do to the numbers if were to pick a different 219 sims and had say a 5% bot count according to how I would determine an avatar was a bot....

would change the numbers a bit

I agree with that. LL's 10% is probably a gross underestimate, imo, even if it means 10% of the total logged-in avatars in a 24 hour period. Percentages are meaningless anyway unless it's a percentage of avs during a 24 hour period.

Possibly, but it looked like it was being presented in the same way that a Linden estimated ~10% recently.

yep
I never put much on % they can be skewed too easily

Well, you haven't done a sweep, have you?

The samples aren't tiny. 700 avatars is a statistically useful number. I suspect you'd have to carefully select your sims to come up with vastly different numbers.

The objective isn't to come up with a perfect number. What would you do with a perfect number? The objective is to show that the Lindens' number (10-15%) is probably very wrong, and that the number of bots is a very substantial portion of the touted concurrency numbers.

I don't accept "it can't be done", or "only a very expensive study will do".

actually being high concurrency is approx 70,000

700 is a very small number to use

700 is what % of 70,000...1%? (rusty on math) correct me if I am wrong


so you (if my math is correct) would be basing your study on a 1% piece of the high concurrency of logged in avatars

Sometimes I sit in my room and stare quietly at the stars moving across the walls for hours (lost in IM's). Sometimes I stay logged in over night, tucked in my bed, with little Z's coming out of my mouth.

I'm not anti-bot. In fact I wrote one for collecting remote traffic data.

I believe LL really doesn't know how many bots there are because they haven't made a point of finding out. So I did.

Why is it important to me? I want to know where SL is headed. The bot number is just a way to quantify what I see when I wander the grid: SL is becoming less social.

You seem to be missing the point. You did a sweep of a tiny number of sims, during a specific period of time. If your conclusion for the grid is correct, then all but ~1000 avs during the low concurrency period are bots, which I don't believe. Percentages are meaningless for a small timeslice of the day. You can use your figures to arrive at an estimate of the number of bots on the grid. That would make much more sense than a percentage. But then you'd be saying that 38,800 of the 40,000 logged in avs during low concurrency are bots, which flies in the face of common sense.

You need a much bigger sample than 219 sims, and you'd need to sample during the whole of a 24 hour period to get any meaningful data.

If what you are doing is trying to show that the recently stated estimate of 10% is too low, then you are preaching to the converted. I don't think anyone accepts the 10% figure.

I use the low concurrency figure for my personal estimate. Since bots are generally 24/7, they are pretty much all included in the low concurrency figure of ~40,000. And since there are a lot of real people avs in that 40,000, the number of bots on the grid is a lot less than 40,000. My personal estimate is somewhere between 20,000 and 30,000 bots, but that's just my estimate, and it can't be represented as a percentage because the percentage changes as concurrency changes during a 24 hour period.

The objective should be to provide an accurate estimate of the number of bots, not to disprove LL's estimate. If you want to find to find bots, you will find them. Research that sets out to prove a point usually manages to do so, doesn't make it good research.

http://en.wikipedia.org/wiki/Turing_test
Q: Are you Human?
Adz: zzzZzzzzzZzzzZZZZzzzzz *drool*

Which is plenty.

http://www.surveysystem.com/sscalc.htm

quite often the sample size of an opinion poll or even a scientific trial are less than 700 people, and they typically are supposed to be representative of 300,000,000+ people...
so to discount this for only taking into account 700 avs seems overly dismissive.
personally i believe the 44% figure is closer to reality than the ll 10% figure.

For a simple two-state parameter, 700 is a very large sample size that would allow a rather accurate estimate of the percentage of bots IF AND ONLY IF the sampling is truly representative of the population. It is in the randomness of sampling, rather than the size of the sample, that this figure can be suspected. randomisation needs to be over time of day (week?), sim type, sim avctivity type, RL and SL geographical location, etc etc. Sometimes the complexity of factors that have to be randomised may require larger samples. Political pollsters, using this kind of representative sampling, routinely produce surprisingly accurate measures of voting intentions from samples of one or two thousand from electorates of tens of millions (less than 0.01%).

I repectfully suggest that Anya has the right to define her own objectives.

Hypothesis testing (here "percentage of bots is higher than LL's estimate" is very well established as entirely legitimate research objective. For example, it underlies just about all statistical research in clinical medicine ("survival in patients taking this drug is better than in those taking a placebo", "rates of lung cancer are higher in smokers than non-smokers" etc etc.).

Hence, of course, statistics. The numbers would be easier to interpret with a few more data points, using identical methods but at a different time, and over a different set of sims, to see whether those factors really systematically move the numbers. And knowing a bit more about the distribution of the current numbers would give some confidence intervals.

FWIW, had I done this survey, I too would have only looked at Mainland, but that's because I think bots on the Mainland pose a bigger problem in that they can affect neighbors (depending on how they're situated and managed). On the other hand, if one really wanted to know if LL's bot-share estimate is bogus, then one would certainly want some Estate data, too (and I, too, would tend to expect that bots make up an even higher share on Estates; in my experience, non-commercial Estate sims are usually even emptier than residential Mainland, and the really bot-full sims are on Estates just because they can crank up the avatar limit far beyond a level tolerable for humans).

But even on Mainland, the data are bound to be pretty "lumpy". (For example: to sample a Help Island or not?)

I have one question: Why are we working so hard to prove LL wrong? Most (if not all) of us know LL's estimate of 10% is pure bull. They could have done better to take a poll of IP's and used the connections that share a same IP and come up with a closer, but still wrong, number.

10% is just a number they made up for the corporations to look at and aren't really meant for the common resident.

Phil, doesn't this sound familiar? We've already had this conversation.

Here's where I explain the percentage...

/327/9f/287267/3.html#post2181015

Here's where I explain why I think the number of bots varies during the day (though not nearly as much as the number of humans)...

/327/9f/287267/1.html#post2180618

219 sims from a >30,000 sim total is significantly less than 1%, but that doesn't really matter. If it's a sample from a homogenous whole, then it would be fine, but it isn't. The population across the grid at any one time is somewhat lumpy, so a much higher sample is required, and over a 24 hour period.

As it is, we have statement that the bot count is 44%, which is both meaningless and cannot be true, due to the daily low concurrency figure. All that Anya can say is that, of the avs she found at a particular time of one day, 44% appeared to be bots to her way of recognising bots. As a representation of the grid, it's meaningless.

I went looking for the LL 10% estimate - I thought I'd read it in Zee's recent thread. What I found there was "10 to 15% of user hours", and not 10% of avs.

Because the squeeky wheel gets the grease. It appears that we have to keep pushing this topic until LL gets rid of traffic as a search metric for good. Until they do so, we can't take the pressure off.

If we sit back quietly, they will manage for forget about the issue and deal with thing they consider more important.

Without constant agitating on our part, traffic will remain a gamed search metric.