Traffic Bots Against the TOS of LL?

I Totally agree with you Lets hope that it doesn't come to that but as you suggest that if it continues the way it is, it will and that would be a bad day for everyone here IMHO

The bottom line is that, if SL continues to grow and these other grids grow.... the overall policing will be just like it is on the web, simply because nobody will have any overall control. Your vision of the future is far too limited. It won't be controlled by grid owners. Grids like SL will just be bits of the whole 3-D thing.

Great, back from holiday and a brand new bot thread to enjoy. Same players, but a much more sensible discussion.

About traffic bots all is said I guess. Only way to remove them, is either use the TOS to forbid running them, or to stop using traffic. Probably the latter is the best, as the first will bring back camping as a search ranking tool.

About Paying Picks I saw some weird statements I want to react to.
- When people in these forums start swapping picks, it is exactly the same as when I pay you to have my pick. You put my store in your profile for a reward. Be it in lindens, or in a return pick.
- When someone has an active SL social life, so a lot of friends, and he asks all of them to put his shop in their profile, it is as much "cheating" as my payments. Because it has nothing to do with the relevancy, but with the amount of friends the person has.

Using only parcel description for search would not be a great solution either, though I can imagine the idea behind it. It is like in the old days of search, where only text on the first page of the site, or later the meta tags, were involved. The more a system evaluates, the more parameters are being used. At the moment, they are probably using the best one they have at this moment.
Do not forget, picks are only a part of the ranking. Ranking is a mix of parameters, and the better you research your options, the better you can rank.

To clarify my position:
Bots I do not use (though I might change that if I see it fit) but I understand people like Phil perfectly well. For me, they are no problem at all. Don't like them though, but if the competition is using them, and they are allowed, you have no choice if you want to compete. Linden Labs allows them, competitors use them, so how can I be wrong using them too? No one has so far been able to show me that.
Picks Camping I do use, though I only pay around 40 picks from the about 100 I had last time I checked (before my holiday). Again, it is one way to get your ranking optimized, you do not force a thing upon anyone, so I do not see a problem.

As long as a business delivers as promised, and does not harass others, I don't see why they should not either run bots or pay for picks. So most add-farming (harassing) is wrong, as is keyword spamming (not delivering what you promised). But that is my view on things. You might think different.

So, what does LL have to change?
- Stop using traffic as a metric for search.
- Optimize Search All (more results per page, for example). The Search All itself works just fine, and Google will probably keep working on optimizing the system. Maybe give Picks less weight, parcel description a bit more, but for the moment Search All is the best search method possible.
- Educate people how to search, as many people are not aware of the strength of the new search.

I'm really not sure why using only the parcel description wouldn't work at least as well as adding in all the other ranking factors. What I mean is that at this point, IBLs generated from Picks are really just synthetic--it's not like they're revealing some underlying semantics of user preference (like "voting" or anything so naive).

I'm not as rabidly anti-Picks as I am anti-Traffic (because Picks don't inherently degrade user experience, unlike the lowercase traffic load that generates the Traffic metric). But they don't seem to be anything more than an elaborately cumbersome way to spread out responsibility for describing the content of the parcel.

Is the virtue of Picks the fact that they can be bought, so--like camping--generate some paltry cash flow for newbies? I really don't see what they add to the quality of Search results that wouldn't be achieved with more expansive parcel descriptions and possibly better (more theoretically founded) text matching.

Sensible Qie,

And upto a certain degree I agree with you. If used right, parcel description does have the best card. That is why parcel name and parcel description are the 2 most important parameters, as far as I can see.

Google as found long time ago, that it is impossible to put all money on one card though. So they use a more and more complex combination of parameters to make sure you find what you are looking for. The appliance that LL purchased uses a lot of those parameters as well. The more they spread, the less easy it gets to influence the system.

So dropping Picks because they are not as relevant as parcel description will never happen and would not be a good solution. What might happen though, is that the formula to calculate the ranking changes, to put more or less weight to certain parameters.

To be fair about the pick swapping in this forum...

Unless I've missed some posts, it only happened right at the beginning of the All search, when people were getting the hang of it, and before the whole thing was thought through and discussed at length. I haven't seen any swap posts since then. Even so, the reason for it was specifically to improve the places' rankings. If a person swaps picks for the ranking benefit (or even asks a friend to add the place for that purpose), s/he would by hypocritical to come down against picks buying. If a person swapped in the early days, but after thinking the whole thing through, finds against picks buying, then the swapped picks must be removed for the person to not be a hypocrite.

I've no idea if that applies to anyone. I'm just pointing out that what was done in the early days of the All search doesn't mean that those people can't be against picks buying now, as long as they clean up what they did themselves. What they cannot do without being hypocrites, is swap picks, or ask friends to add the places to their picks, and still be against picks buying. It cannot be argued that a little is ok, but a lot is not.


Bots are different. It can't be argued that using bots functionally (e.g. models) is in any way wrong. But, apart from cutting 16m plots to place them on, there is no way of using them on the place's parcel without improving the traffic number and the place's rankings. And yet the traffic they produce is in no way related to popularity and shouldn't count, but it does count. Each one of them adds 1440 unjustified traffic points per day, and cause the place to be higher in the rankings than it would naturally be.

It's reasonable to assume that business owners who use bots/alts as models and such know very well that they will improve their ranking because of them, so they are not completely 'clean'. They could cut small plots for them if they wanted to be clean, but they dont. Instead they accept the unjustified ranking improvements. Nobody who is against traffic bots can be in favour of models and such. If a little is ok, then a lot is ok. The degree of use cannot be argued. Unjustified ranking improvements is either right or wrong, regardless of the degree.

But the odd thing is that the anti-bots don't mind bots/alts being used as models and such. They claim they serve a useful purpose, so they are ok, even though rankings are unjustifiably improved, and they do to competitors and searchers what the anti-bots are so against. This doesn't make any sense to me. Either artificial ranking improvements are right or they are wrong. It can't be had both ways. It can't be a matter of degree. It can't be argued that it's ok to jump over a few competitors in the rankings as long as you don't jump over a lot of them.

So the anti-bots really must be agaist the use of bots as models and such, or they don't have a credible argument. And yet using functional bots must be ok, even though they improve rankings by causing the place to jump over some competitors. It's a dilema, and the only way to deal with it is to get rid of traffic rankings altogether. Bring back the badger!

Official Schwag for this thread!

Although I certainly agree with the conclusion that Traffic needs to go the way of the buggywhip, I think I can see a clear path through the "functional bots are okay" conundrum. The thing is, if somebody else's bots perform some useful service, then someone pro-Traffic might rationally be able to live with the fact that they're also skewing the scores: to them, the practical benefit simply outweighs the (social) cost.

It's a somewhat different matter when, as you describe, the pro-Traffic person has their very own functional bots on their store parcel. Still, the offense of such a practice might be overlooked, especially in business areas where the Traffic metric is so completely skewed that nobody in the market even tries to compete without running bot boxes. Yeah, it's still theoretically wrong, but practically it may just not make enough difference to even bother with carving out a separate parcel in cases for which such limited Traffic inflation still leaves the ranking buried deep in the Search results.

It's a dilema, to my way of thinking. I do understand what you wrote, Qie. What I can't accept is that the antibots agree that it's ok to use model bots and such, even though they have the same effect on the rankings as pure traffic bots do. I don't see any difference between jumping over one or two competitors and jumping over a lot of them. Each competitor that is jumped over moves down one place, and that place might make a noticeable difference to sales. When a place uses bots it jumps over many competitors, but it's no more significant - each of them moves down just one place, which might make a difference.

An aside:
I mentioned earlier that I'm now against the banning of traffic bots and camping because it still wouldn't make the results ok, and I mentioned clubs with shops on the same parcels as an example. Functional bots are another example of why the results wouldn't be 'correct'. I recently started using 4 of my bots as animation demo models, and customers do use them for that. They are functional, serve a useful purpose, and are actually used by customers for that purpose. But those 4 would give me a 5760 traffic points start on most competitors, and I intend to use another 2 for demoing a different anim set. That would be an 8640 points start. So, imo, banning traffic bots and camping would possibly improve the rankings a bit, but not a lot. My functional ones are genuine but, if places aren't allowed to use pure traffic bots, they'll come up with all sorts of functional ones, and the results won't be improved much at all.

One disadvantage of functional bots compared to trafficbots is that, for some functions, they must be located right there with the customers, so they can't avoid but contribute to lag (texture- and animation-downloading; some physics calculations; etc.) exactly as would another customer. There's some combinatorics at work, too, so I'd keep them well separated from each other if I could, although with H4 those effects seem to damp pretty much instantly.

Yes, I agree. Kilometers in the sky is much better. My 4 are for people who are on their own in the place to either watch or try the animations before deciding whether to buy a bed or not. If/when I add the other 2, they will be quite a long way from the 4, but all of them add a little to the lag.

It's another reason why banning traffic bots would not be a good thing.

Parcel descriptions are way too limited in size, and can't possibly describe many parcels in a such a way that they are even ranked at all for all that they are relevant for. Whole pages are suitable, but not tiny parcel descriptions.

Google wasn't the first engine to factor IBLs into the ranking algorithm, and the reason that IBLs were included is because we all tailored the content of our webpages to suit search rankings. It was just too easy. We even made different pages to suit different engines (that doesn't imply cloaking for those who know about such things). The same would happen in the SL search if it were just a text-based system. Google was the first engine to treat the link text itself as being very important, but it relied on the web's natural linking - something that largely disappeared as Google became more popular. They screwed up the natural linking of the web.

The Google appliance that LL bought uses Google's thinking on how to rank pages, which inludes the importance attached to IBLs and the link text. Whether it suits SL or not, that's what LL bought, and they don't have much in the way of choices in how it ranks pages. It *has* to have links or it won't work at all, and LL chose the most obvious, and easy to implement, links. They mentioned at the start that it wouldn't include external webpages yet, so the idea of it including external pages hasn't escaped them, and there may be some radical changes in the future. I can just picture the outcry in this forum if they do that. People will obviously create sites and pages to improve their SL rankings - and we think the bots threads are bad

[added for Marcel]
I don't believe that LL has any choices regarding the weightings given to the ranking factors, and I am certain that they don't even know what the factors are. They may even have learned about the importance of certain things from this forum.

Then it would be wise for those dependent on it for a livelihood now, make other plans, because you will be diluted by international spam in 3D and sweatshop banks---the equivalent of RL camping pads.

I did suggest in this thread and others that the Text areas of About Land be increased.
It's the obvious thing to do if the text were to be the sole source of ranking.

Most residents will have a simple "My Home" or nothing at all. People wanting to be found in search will probably write volumes.
It's simple and appropriate for inworld search our tiny universe.

When you get right down to it, the way the Places tab results sets are compiled is so incredibly basic that it wants ditching anyway. A search system that only lists exact matches is too ridiculous to keep, regardless of how it ranks the results.

I'm torn on this because I hate the way parcel names and descriptions have developed to be keyword city, I'd like to see a plain text description of a parcel and a parcel name being just that, a name.

Sling, I guess I am confused how this would help. Would you prefer that the items on the parcel no longer be indexed as they are now in the new all search?

My concern is that if you removed that, and removed the relevance added by linking to a particular parcel via picks, and traffic (all of which have their flaws) what would be the incentive to retain larger, listed parcels? It seems to me, there would be an incentive for anyone concerned with high visibility in search to slice their parcels into microparcels so that they could list as many keywords/parcel names as possible, filling the search with dozens of, potentially, redundant listings. No matter how large the description fields, they will always fill up, and there has to be some way to determine the relevance of listed keywords/phrases so word order and concentration will still have to matter if the listings are to be sorted at all.

Even with the current search - for which indexing the contents of the parcel, inbound links, and traffic provide a disincentive - this is still a common and somewhat useful method for showing up in search. Under your system, how would you determine what is "spam" and what is a legitimate reason for cutting off a new parcel for a unique listing?

I'm not saying I think the search we have is great, but if "gaming" search is the problem being addressed, I'm not sure I see how this fixes it. I'm curious to know more about what type of system you are proposing.

Yeah it looks horrible now.

The problem for using a mega-parameter search indexing algorithm for a SL in-world search is that appropriate in-world parameters are as rare as hen's teeth.
Traffic? Gamed to hell and back and therefore useless.
Picks as IBLs? Well on the way to the same uselessness.

LL bought a hammer. They have no nails to hit.

Hey guys! We need more things to use as parameters!!!
How about.... parcel size divided by time elapsed since claimed to the power of the time of day divided by the second number you first though of plus the number of vowels in the parcel description minus the number of vowels in the parcel description?
That would make as much sense.

Would I prefer that items on the parcel are not indexed?
-Yes.

Why?
Because that's gamed with plywood boxes as much as is traffic and paid picks.

New religion = Remove factors from search indexing when every dog in the street knows that they are being abused into irrelevance.


If they had a mind to, LL could clean out these abuses.
They don't appear to have a mind to. It's just another thing that as an organisation they really excel at sucking at.


This is the fatal flaw in my proposal for a simple system.
Unless the search algorithm did it automatically for them, LL wouldn't do a dammed thing about someone filling their expanded About Land fields with a single keyword.



The *good* thing about all indexing being done purely on the parcel name and description is that keyword spamming is very apparent right in the search results.
There are no hidden bots, pick, prims, etc. gaming the results. It's all in front of your eyes.

Google has a decent but far from perfect sense of keyword spamming. In theory plywood boxes should only work so far, keep repeating the same terms and you're wasting prims, or you should be if it's configured correctly.

Thanks for taking the time to answer, Sling.

Just out of interest, have you seen this yourself? I don't get out and about so I haven't seen it, and I'm wondering if it's just a myth that began in this forum.

Plywood boxes do have valid uses, and it can't be assumed that they are spam (which doesn't exist in SL). For instance, items that are sold from vendors won't be represented in a parcel's html page. I see it as a good thing to represent them with plywood boxes, otherwise the system short-changes the parcel.

Scene: ShinyShoes' office.

Pragmogul Moneybags: Yes, my CTO agrees that in-world Search is the way to drive customers to Amalgamated Widgets, but that all my current web links don't count here.

ShinyShoes Linden: We use the Google Search Appliance, so your web advertising staff should be very familiar with the concepts of In Bound Links, and how to optimize them.

Prag: But all the money we spent on web Search Engine Optimization doesn't apply here?

Shiny: Oh, sure it does! It's the exact same thing, just a different domain. So, instead of other webpages, IBLs come from things like Profile Picks.

Prag: "Profile Picks?"

Shiny: Yes, that's how a resident can show other residents their close friends and places they like to visit.

Prag: Oh, the Friends List, right, I've heard about that. So I just have to get staff to make lots of friends, then.

Shiny: Uh, no, the Friends List is something else; it doesn't enter into Search ranking. Profile Picks is... well, it's like a list of links, with pictures, that residents use to show off their discoveries.

Prag: Oh, they told me about that: the Showcase! That sounds just right for us.

Shiny: Well, no, not the Showcase. We should talk about that, too, but Profile Picks are different...

Prag: Okay, whatever. Just tell me how I can score lots of Profile Picks.

Shiny: Well, some enterprises pay residents to include their stores as Picks in their Profiles.

Prag: Ah. Paid advertising. Like classified ads, hosted by residents.

Shiny: Umm. No, "Classifieds" is something else...

Prag: Enough! Time is money, and you're wasting my time. Now, can you introduce me to this nice "display advertising" fellow I heard about? I bet he'll talk straight with me.

Yup! Using plywood boxes to index the individual items available from the vendors would be a sensible and useful thing.
However, if such boxes add to ranking, then they absolutely will be abused, and there goes yet another 'good thing' subverted by the people_whose_lives_are_not_high.

Dwell - abused into meaninglessness.
Traffic - ditto
Picks - ditto
Prims - more of the same

It can't be assumed that -
It can't be assumed that an avatar on a parcel is a bot.
It can't be assumed that a pick is not paid for.
It can't be assumed that a prim on a parcel is a bogus ranking prim.


If LL won't stamp on the abuse, then the alternative is to remove the abused factors as ranking indicators, and never introduce a new factor without first ensuring that it can't be abused.





Spam does exist in SL.
Unsolicited electronic messages.
I get group invitations out of the blue - even sent when I'm offline.
I get messages when I pass some parcels without actually entering their boundaries.