Considerations about earch ranking

I've beem followin the threads on traffic, picks, and search gaming. Those threads are already too big and biased, so I'm starting a new thread instead of repeating the same in all of them.

First, any single factor used to rank the results can be gamed by anyone who knows how it works. That lead to the camping system, camping bots, payed picks etc. The simple answer, ged rid of these factors is not an answer: It just lead to another form of gaming the new factor.

Some have proposed that LL use plot description, instead. Would it be better? All mall/store owners have the same skill to describe appropriately his/hers plot? Not all are proficient in English (I myself am Brazilian, my English is far from good, I have a poor vocabulary and make terrible mistakes). People with more money can pay for others to write a good description. Would it be more fair than the the pay for picks? Anyway, the first search engines in the web were based on the description of the page to rank results. That lead to descriptions with hundreds of "sex" in the description, for example ("sex sex sex sex ..." until the software of the engines were modified to purge this type of cheat.

Gamed or not, these factors (traffic, decription, and picks) do have valuable information. The answer, in my point of view, is not get rid of this or that, because it can be cheated - any factor can be cheated. The more factors involved, the less value in cheating a single one. In other worlds, the amount that a guy is prone to spent in gaming one factor - say, paying campers - must be divided in various factor - pay campers AND pay for picks AND any else. This is what Google does - the more factors, the more confiable is the rank of its search results, even with people ever trying to cheat the Google search.

Just forget, dont blame who is cheating with camp bots or so. Some make those. Other have to make the same because other do. And all of us stay blaming one another, and the things will stay the same.

Just this - and forgive me for my mistakes in English (hope that I've made myself understandable)

The point of search from the resident point of view is to find what you are looking for, whatever that is. So the best search ranking system is the one that brings those results to the top. From the business or club owners perspective, search is a way to bring people to your parcel. Ideally you will bring people who are looking for what you offer. Some owners feel they need to bring the most people, even the ones who were not looking for your stuff exactly.

The conflicts and gaming of the system seems to come from when many owners offer the same things, and only a few places show up in the search results on the first page.

What I think is important in SL is that it shouldn't be possible to game Search in a way where paying increasing amounts of money results in higher placement.

Paying someone to write a description, or even do general SEO, doesn't apply because in that case there is a finite fee to pay and then they do what they can. Paying for picks or bot hosting, on the other hand, has no limit.

The reason is that, as I've mentioned before, it stagnates business innovation by driving the risk too high. SL doesn't need the business filtration that the real world does (because raw materials aren't scarce in SL).

It may not be much of a solution- but we can police ourselves and hope that ethics will catch on- I don't sell sex- so yo won't find "sex" in my store or parcel or group descriptions- you will find every word for ladies fashions in Spanish, English and French that i could think of. Maybe I could do better if i paid for it- but other than asking advice, I will stick to my own efforts. Advise though- I am always willing to look at.

That's it. Is that any insurance that the better description is for the best place for wathever are you looking for? A well writed description, full of relevant keywords can led you to a poor designed sim. In this way, traffic can be more relevant. If you are looking for, say, romantic places, a place with a single linden tree, with a poseball of hug is poor, no matter how good the description seems.

Yumi: I agree that gaming the search is a crap, but it is what that do exist, and will exist ever. Google is constantly fighting it, and ever there appears new ways of cheat. Thats why they are ever changing they ranking softwear.

Yes, absolutely, traffic is of importance to search. The problem is that every time LL changes the search system, they TELL everyone what the new rules are - and everyone runs out to game the rules.

I think that the random listings someone mentioned the other day are still the best way to go. Because, for all you care, that poseball under the apple tree might be just what I'm looking for.

Traffic is of minimal importance to the new search. It is of maximal importance to the old search's Places tab.

LL didn't tell people what the new rules were. It was largely figured out by Phil Deakins, who shared most of it with everyone else, and is discussed in the sticky up top. Phil had a lot of experience with the Google search app, and translated it to SL.

By and large, the plot description and the objects on it matter but the most influential factor is the use of profile picks, which has spawned its own contoversy. I suggest reading the sticky.

If you're just talking about camping, or paying for picks, then I'm probably in agreement. But the way you've worded that sounds like you're talking about classifieds, too. If you're talking about classifieds, then I don't agree with this. A business that grows to a large size often needs to vastly increase their advertising budget in order to keep the business going.

In a strange sense, rental properties do deal with a scarce raw material: people! We have to expend great time, energy, or money to keep replenishing this raw material, for no matter now nice your property is, a few months is basically a lifetime in SL, and people simply seek a change after a while.

I grew my rental properties through a lot of hard work, money, and frustration, into a 6-sim community, and feel that I've earned every penny of my classifieds budget. It would be pretty ridiculous for me to have no way to pay money to advertise bigger than a newbie trying to rent out their new 512.

That Deakins sticky is from 2006, as is every post in it. I was thinking more of the blog posts from THIS year.

http://blog.secondlife.com/2008/04/28/second-life-showcase-popular-places-and-the-future-of-traffic/
"Places was moved to a web-based feature, this tab will be returned to its previous interface, which includes “Sort by alphabetical” or “Sort by traffic” options. This change to Search>Places will appear in 1.20 RC9..."

http://blog.secondlife.com/2008/03/26/tips-to-improve-your-search-ranking/
"Traffic Matters… kind of – Traffic is a number for each parcel which is based on the amount of Residents who visited, and the time spent on that parcel out of their total time inworld that day. It’s calculated using a complex algorithm. Read more about traffic in the Knowledge Base (KB). Traffic is still used to help determine relevance with the new Search, but not as much as in the old version.

Picks Matter - Having links in other Resident’s Picks tab counts, but creating bots/zombies stuffed with top picks doesn’t help. The more active Residents Picks that reference your parcel/profile the more relevant your listings will be in the overall search."


Traffic is important, and it is not exactly "minimal".

I joined in December 2006, Phil joined in January 2007. The new search is only a few months old and the sticky is only as old as April 2008. What are you talking about? /327/b1/252152/1.html

As for LL telling us how it works - those are only vague terms needed to tell people how to optimize for search. It doesn't tell you that traffic will get you at most 12 inbound links, which is what makes the effect minimal versus an infinite amount of profile picks.

I'm talking about the need for increased coffee quantities at work. My bad.



Well, this is true. But if I knew that getting more picks would increase my traffic, just not at what levels, I'd just keep doing it until it worked. I'm tempted to think that the ones gaming the system are not going at it with the most scientific of minds - they're just aware that more picks brings better rankings, among other factors. My thought would be that, since search is so important in SL, it would be better not to reveal all of what's used at all.

Back to the parent post. This is really an interesting idea I haven't seen brought up elsewhere and I think it could be a workable solution. There must be dozens of non-related metrics we can use to rank search results. We have traffic and picks, we can also do parcel size, unique visits, money spent on the parcel, number of transactions. we can monitor chat logs for mentions of the location and so on. I'm not a very imaginative man but I bet those smart guys at LL can come up with a few more just by trawling all these endless tedious threads about search. Anyway, if LL mixes ALL those parameters into the search using some non-obvious formula, you know, 3 times the number of picks plus the square root of the money spent plus the cube of the unique visitors etc etc then it will be very hard to game and probably pretty reliable.

Sure, it's security by obscurity but it seems like the only viable option to me. Once people start to work out the parameters LL can just tweak them a little in some semi random way. Well, I really like this idea and I think it could be very effective.

I'm not sure they don't tweak them as it stands. Certainly at one stage I seemed to notice my traffic would drop in the same pattern and then return to the higher level, almost as if the weighting given to certain parts of the description was weighted differently on different days.