Traffic Alternatives - L$2000 Reward

Traffic is currently 1 point per avatar per minute. Many places add a lot of bot avatars to their parcel to artificially increase their numbers. Many others pay people to stay on the land for long periods of time. Both of these make the current traffic calculations a lot worse than they could be. Unique TPs are not only as easy to fake as the traffic, but the constant teleporting in and out of the sim by different bots increases the load on the servers. Kitty thought of a way to calculate traffic that would not allow people to have more than 1 of their own avatars count for traffic at any given time. Many business owners are logging in 20 or more avatars to do nothing but sit in a box on the land to fake the numbers, and we're trying to find a way to stop that and still keep a traffic based search.

I actually understand what you mean now, for some reason I was applying it in my head to general search, not the places function.

Wouldn't traffic 1=1 without the per minute, for length of stay also make bots considerably less relevant ? As far as the TP issue is concerned, I don't see why the Lindens don't simply auto disconnect at a tp threshold and lock it out for a set period of time, ala 5 minute warnings.....or if not by number of tp's within an interval then by repeat tp's to location within an interval by an account.

I can see how PIOF is filterable and a measurement, but you are then excluding alts, dependent non-PIOF's and such, who also spend lindens or perform the actions you expecting people to be doing at a place you are looking for, which makes the number considerably less meaningful from an owners standpoint who is selling anything, as you are excluding a significant demographic of consumers, performers, pedestrians, etc...

At best is only telling you one of a single PIOF's accounts were/are there....and might not even tell you that if it is monitoring the AV who is not actually manned.......or losing the number if a PIOF series of accounts is used by a couple or family.....

Counting teleports instead of minutes gives the same result for someone that shows up, gets disgusted by what they find and leaves as it does for someone that finds a wonderful place to spend their day. Someone could have hundreds of bogus accounts tp in to their land every day. Or they could use bogus keywords in their ads to trick people into TPing in.

We would be counting PIOF accounts as a representative sample of the population. It's a large enough sample to be valid. It's analagous to the way TV ratings are based on a small percentage of viewers. We would have to use PIOF accounts because that's the only way to keep someone from running dozens of avatars at the same time and having them all count. We would not be taking anything away from non-PIOF accounts. They are still going to be the majority of the population. Places trying to attract customers are still going to want to reach those customers.

Then you are still including bogus keywords and tricked people numbers in your sample, and still excluding non PIOF accounts from it.....which does not make it an accurate sample for any interest. TV ratings for example aim for multiple pools to measure by. I am not saying you are taking anything away from non-PIOF's.....but they are significant for your purposes across the spectrum of multiple places for multiple purposes that you could be looking for in searching a place. Does a map full of green dots, with a much lower traffic total mean bots or non-PIOF's ? Does a place with low search-places traffic make you not look at the map to see the dots to ask the question ?

Even within your PIOF pool you have a one in x number of accts per person chance of being incorrect against the uncertain number of 1 acct per person that will be correct, while possibly including the bot anyway, and not including actual traffic.

Thinking you are just getting back to traffic not being accurately useful with a different set of measurements to make it not a useful number in a different way.....

You're missing the main idea. The selection of PIOF accounts that can be tied to a unique identity are a representative proportion of the SL population. The goal is to keep individuals from having more than 1 account being counted for traffic at the same time. This keeps the numbers from being faked. If the PIOF accounts make up 25% (that's not the real number, just a hypothetical) of the population, each will represent 4 people. This is the same way that other surveys can produce statistically valid results without asking everyone in the country their opinion.

I don't understand what you mean about a "one in x number of accts per person chance of being incorrect". If someone has more than one of their accounts logged in at once, only one will be counted for traffic. This keeps people from leaving a bunch of their own accounts logged into their own land to fake the traffic numbers. The avatars that do get counted will still end up being a statistically valid sample.

I've used several different credit cards on my account, some with my maiden name, some with my married name, some with my husband's or my mother's name even. The accounts page (or browser) still fills in my name on the account and accepts a credit card with someone else's name on it. Maybe it's a glitch, maybe there are four or five people using my account, maybe there is no way to verify who actually owns my account?

Also, before open registration it was possible to make an alt without your name attached to it at all. All you needed was a verifiable email address and any made up "real" name would do. I have a few (whose passwords I can't remember, heh), and they still show up in search as basic, no payment on file accounts.

Actually no, I did not miss it, but representative is where it becomes flawed, and if you are not/cannot count legitimate non-PIOF traffic, it becomes less representative still, where it comes to many types of listed places. You mentioned earlier, majority of accts are non-PIOF.....so if you exclude your majority, you cannot achieve representative. No current ratings or survey system I know excludes the majority to get representive data usable by the whole.

What I meant about "1 in x"....is that say a person has a 5 accts, only 1 is being counted at a time. That's 4 bots.....someplace......and one person performing the activity you are searching for in places. That means you have a 1 in 5 chance of counting the person, or a 4 in 5 chance of counting his bots. SL is having trouble...he keeps disconnecting, each reconnect, you may get a bot added someplace. That would be how someone creatively scripting would game it, logoff/logins on a timer. One will always get counted right ? How could you tell if it was a person having rl conn problems ? If they were, why would their traffic not count in the representative sample, how many times would it count ? Traffic would shoot up on bad SL days from relogging alone on 1 person-1 account machines.

Now your representative sample is corrupted.

I am not entirely sure, though you may know better than I, the percentages in how representative traffic fluffing is among legitimate locations.....compared to a corrupted representative sampling...but I am guessing you are not gaining much, even losing ground, again, because you are not counting the legitimate non-PIOF's in your true traffic majority. What is the current representative sample like ?

Turn Kitty loose for a month at LL. Singular AV identifier is UUID. You have a number. Is the person a camper, check the UUID on timer, do not recount same UUID, add 1 to count. TP bots.....same, get UUID, compare against previous.........one UUID registration per day, does that make it legitimate traffic ? Might still be a one timer bot, or make people increase their bot stables...

A popup ? I dunno you cannot script a popup reply as it always pops up in the same place and you can script mouseclicks.

I believe that using your family member's credit cards is legally the same as using a stranger's, but they probably won't press charges. Using someone else's name to get a credit card for yourself is definately fraud, even if it's a relative. If you are using credit cards that don't belong to you, that's a matter for you, the person that owns the account, and whatever authorities control those things. The solution suggested in this thread would still knock out nearly all the bogus traffic numbers.

The television ratings use less than 2% of viewers. According to you that's throwing out 98% of the results. That's not the way test samples work. Many national surveys talk to a few thousand people to get data about a population of hundreds of millions. The method I'm talking about for SL would actually be using a much larger sample on a percentage basis.

You're paragraph about a person logging in and out of SL makes no sense to me. We don't want someone to log in 5 avatars at once and have them all count for traffic. That's the problem we have now that needs to be solved. If someone wants to know which of thier avatars is being counted for traffic, they only log in with 1 at a time.

Hmm.. not sure if this has already been suggested or could be gamed, but I'll throw one out there.
Traffic could only be counted while the player is active. If the player doesn't, type, click things, or change the camera they'd be considered inactive.

*nods* I did point that out though .

You could have 4 "unique" accounts, but how many more than that are you going to be able to manage to come up with? The basic idea is to reduce the impact of "alts", not to get rid of it entirely because that's just not possible unfortunately.

Traffic would also be capped, my suggestion was at 3 hours/day for the same parcel so using 4 accounts like you have would increase your traffic by 720 only which is half of what just one camping zombie contributes today. You'd have an edge over your direct competition with similar traffic to your own, but it's not going to allow you to significantly change your ranking.

Open sign-ups didn't introduce "no payment info on file", they just made them visible .

You didn't have to put a credit card on file before open registration, you could register an account by phone as well but as you pointed out that account would be "no payment info" and wouldn't have gotten the L$250 starting money or the L$50/week basic stipend or would qualify to contribute to traffic with the suggestion.

I'd have to think that "non-premium payment info" accounts outrank premiums at least 2-to-1. Depending on which of the two numbers LL publishes you use you get either 11% or 17% that's premium. If my 2-to-1 is a decent guess that brings you in the 33-51% range so you'd be using a 1/3 or 1/2 sample which is really pretty good.

Your traffic contribution per parcel will be capped, if it's 3 hours then 180 is the most you'll be able to add to your traffic no matter how you try to shift things around or how many alt bots you try to throw at it. Bots are not an issue, logging on and off makes no difference either.

As far as the "only 1 account counts" fairness goes: imagine you have your main logged on at a club dancing for 2 hours and you take your alt shopping for 2 hours at the same time. How much time did *you* spend on SL: 2 hours or 4 hours? The answer is obviously 2 hours, just because you have multiple accounts logged on at the same time doesn't mean you suddenly spent twice as much time online. Counting only one account at a time adjusts for the disconnect with "cummulative SL time spent" to "actual RL time only".

That's entirely true, but how would you distinguish between legitimate NPIOF interactive humans and non-legitimate NPIOF human zombies/campers/bots/etc? You need to find a way to distinguish one from the other and I can't see one that is within the limits of what we know LL will/won't do.

* they won't "ban" bots/camping directly
* they won't "ban" unlimited alts creation
* they won't get into a race (which means any bot detection is out of the question)
* they won't police beyond what they're currently doing already

Why not collect traffic through the Second Life Viewer instead of through the servers? Then traffic can be collected for OpenSim etc as well and you can filter by IP so people can't run 50 bots from one computer?

The viewer is open-source so you could have it report back whatever you like.

Bots are also light-weight enough that they can run over an open proxy (connecting to SL through another - generally trojan infected - puter) and there are literally thousands of those, so IP wouldn't make a good enough unique filter .

I'm also not sure why you would want/need traffic on an OpenSim?

You can do all those with a bot.

I'm thinking about the 5+ year outlook. By then there will be a lot more traffic to base the ranks on, so I can't see individuals running hundreds of thousands of unique bot hits through proxies. Also, the grid will hopefully be distributed by then with offerings from OpenSim (and/or any others).

So even in the short term, collecting the data through the viewer may work, but it's in the long term where you'll reap the benefits of this more.

I wrote a little blog entry about it actually, if you'd like to see more of my thoughts. http://snurl.com/279f8

The idea does have a great deal of merit in limiting traffic gaming, but I see a couple of things.

It majors on limiting gaming, which it succeeds in doing, but the objective is to measure traffic fairly, and I don't see anything fair for places in cutting off the majority of their visitors for traffic counts (I'm guessing that the majority of residents don't have payment info used).

The idea that only a representative sample is needed is not right, imo, because it could cause traffic figures to unfairly swing wildly day by day. On some days, a good proportion of the people who visit a place will be PIU, and on other days the opposite will be true, and that will be the same for all places (except places like clubs that attract the same people much of the time). Much, maybe most, genuine traffic won't be counted, and the sample this is counted will cause the search results to fluctuate wildly day by day. It's not as though places get thousands of residents through them every day, so a good sample is likely to produce reasonably good averages. Places don't get thousands of people through them every day. The numbers they get are relatively small, which means that sampling is far less accurate.

As a measure to limit gaming, the idea is excellent but, imo, it doesn't stand up as a measure of traffic that is fair to all places.

There you go, Kitty

Thankies .

Measuring fairly and limiting gaming as two sides of the same coin though .

My best guess (based on what LL does release) would be that payment info residents account for at least 1/3-1/2 of active residents. It would be helpful to have an accurate statistic of that though .

I think a sliding historical (a week perhaps) average could solve most of what you pointed out though, especially since it's a rather trivial thing to add, although a day by day accounting could be useful as well.

A sliding average would compensate for "one time" events such as an opening or a one-day sale or other things that draw a - relatively - large crowd and if it's a week it accounts for all days of the week. A week also wouldn't represent a too big barrier for entry for someone who just started out.

On the other hand, I do think it could be useful to actually see the day-to-day fluctuating results as well, if only because it's more dynamic and it could address some of the concerns that smaller venues have in getting exposure.

I don't personally think that throwing a "party" for something like a shop opening (or a product launch party for instance) is really "cheating" since it's hard enough to get a significant number of people to show up for a one-time thing as it is and the new limitations would put an acceptable limit on how effective that would be. In the sliding average results the one-day advantage would even out and not give you (much of) an edge, but in the "previous day" results your ranking/exposure could increase for that one single day giving you more exposure for a limited amount of time.

(While it's likely that the above could be extended to trying to get a more permanent "higher than deserved" ranking, the same things that would make camping count less would apply to 24/7 events as well. There's just only so many people you could attract day-after-day and turning your store into a 24/7 party house isn't that appealing to anyone either, few people even go to clubs as it is)

I think that you did bring up a good point, but I don't see it breaking the merit of the suggestion though . At most it's some fine-tuning on whether to pick sliding average over day-to-day or give the option to search both.

The current traffic metric is great for newbies looking for places to make money.

The premise of a single number to indicate the popularity of something is the flaw. What is the metric for a rl store or restaurant? Reducing a complicated evaluation process to a single value serves the ignorant consumer and the deceptive provider. There is no better scheme, just different ones.

Since the practical effect of the traffic metric is to provide money for newbies and sell list positions to people willing to pay them, why not just do that. LL can auction positions as they do for classifieds, and use the revenue to pay stipends to newbies. Avoid all the wasted time and energy, and stop lying about what the traffic number really means.

Kitty.

Your addition would provide another means of gaming, but it would alleviate the wild swings to a great degree. I'm not going to even try to disect/criticise your method just for the sake of it, because it's a very good method of dealing with something that should be dealt with. In fact it's the 4th best method that I've seen - the first 3 are my methods (remove the Places tab, put the application results in the Places tab, announce that traffic gaming is not allowed and is subject to penalties).

A way of gaming it just occured to me (I know you said that nothing can be free from being gamed). If your method were applied right now, I and others would inadvertantly have an advantage right from the start. There are rented out skyboxes over the same land that my store is on, and they mean traffic (they are remnants of my skybox business and decreasing as people leave). It would be subject to PIU of course, but it would be there. It could be done intentionally, of course. An owner of a large enough piece of land could easily rent out places cheaply to PIU people, and do all sorts of things, to assure some additional traffic. Lucky chairs could be used as they are now, to assure plenty of additional traffic - especially at female clothes place, where groups of people wait and wait for letters to come up.

All in all, I think that determined merchants would be able to take steps to ensure being at the top of the rankings. It wouldn't be as simple as it is now, but it would be done by many merchants.

I was hoping you'd find your way to this thread, Phil. I was thinking of sending you a notecard to get your take on this idea. All of the things that you just mentioned for gaming this system all involve one critical component. An avatar that is actually tied to an identified human being. The biggest issue with the current setup is the ability for one person to have dozens of bots logged in at the same time, all counting for traffic. The things that you were talking about involve drawing in an actual crowd to be engaged in the event or interested in the product enough to wait on the lucky chair. The property rental issue can also be mitigated if you add on another of Kitty's ideas - Limit the amount a person can add to a specific place to 3 hours (or some acceptable amount).

Let me know if you've got any other ideas about why this should or shouldn't work.

For anyone just joining this thread: I'm still offering $4000 for great traffic ideas.
The first place winner is Kitty. I summarized and expanded on her idea back in post number 89. She's talked more about it on this page too.

I'm offering $1000 for a significant improvement to the idea, $1000 to anyone that can convince me it won't work, and $2000 for the next completely different idea that can meet the criteria I laid out in post number 1.

That was kinda my point. ; )

I sincerely wish you the best on the idea, I just cannot see it as being representative, ungameable or useful, to, say my partner's biz with the tools you have available to work with.

I just cannot agree, no offense intended to Argos, that the sampling itself is not corrupted by the things you cannot measure or control, and is not in fact representative (and I explained why and did not see anything to change it), though it is the best you can do with what you have here.

Good luck with the idea !