Traffic Alternatives - L$2000 Reward

Um, no, we don't.

No one ever looks for someplace in general. And if you do, you may as well choose something random.

I have a gaming place and beach hangout. I sell no products. I often look for places that don't sell products. I do not choose places at random.

I don't think it would be extremely disliked. Personally, I find them mildly annoying, but I understand why they're there. It would be just one-time when you log in. If that's too much make it come at randam every 2nd, 3rd, 4th, or whatever login (the server would have to know that you were nabbed with a CAPTCHA and always give you one after you get the first one until you successfully log in).

8 hours is a long time, but I think bot runners will quickly grow annoyed with having to manually relog their bots 3 times a day.



Yeah, this may not be the best of suggestions. LL would rather have more premiums than less no matter what.



I didn't think that at all.

--Hugsy

I see what you are saying. I was thinking about the CAPTCHA only happening after being logged in a certain amount of time. I missed the CAPTCHA at login. There are ways around that and they can be implemented automatically, but they are probably more complex than most people are willing to go. If I was making decent money in SL and running bots was part of that, I don't think logging them back in after 8 hours would be a problem.


It happens sometimes, and this thread is set up in a way that could easily lead to the same kind of confusion. I just wanted to put that in there for everyone. I do think the current Places search has problems, I just can't think of a better way to do it that actually works better than the current one.

I just plugged "doesn't sell products" into search, and it came up empty.

So I'm pretty sure you don't literally look for places that don't sell products. And if you were to come up with criteria that returned results along that line, you'd be getting rentals, private residences, empty sims, RP sims, business sims, educational sims, and so on. Sounds like a pretty useless collection to me.

So tell me what you do search for, not what you don't search for. That's the only way to come up with an appropriate mechanism, which may or may not be a metric.

What I find that always works is when someone asks the opinion of people in the forums about finding a certain thing. I recall once someone asked for small time animation stores. That's something that would be nearly impossible to figure out using the current search systems. I've seen someone ask for a specific hairstyle or skin that they saw in an ad. After some mulling around the answer would come up. Again something impossible to do with the current search. Why not something like that. Not a google type of solution with a programmable metric but a human solution. It's not like this isn't being done. There's Digg, Reddit, Youtube... I think it's called Web 2.0.

Some of this is already built into the new search just because of where it came from. So if you search for butter -peanut the minus should eliminate peanut butter, or use "peanut butter" or peanut-butter to get just the pasty stuff. OR and AND in all caps like that should work too.

Well, the examples that I gave that I personally run were a game place and a beach hangout. Product searches do nothing to find a game place that doesn't sell games, or a beach hangout.

Cool! This seems to work. I searched for the following and got different results:

golf
golf -club
golf AND pizza <-- btw, no AND = the same results
golf OR pizza

AND and OR as you said have to be upper case.

--Hugsy

Yep, the AND operator is implied if you don't put your own. You can also use parens to broaden or narrow the search. butter (apple AND peanut) will find entries that discuss both apple and peanut butter, while butter (apple OR peanut) would find entries covering either or both.

and yes LL should document this somewhere

Thank you for pointing that out.

I suppose that there was one extra step to improving Search that I didn't consider. That second step is documentation of the first step, so that people can actually use it.

My understanding is that CAPTCHA is very difficult to beat so I assume very few of the current bot runners would try (or even could have enough success). But I'm no expert and could be wrong.

I think that forcing a manual relog after 8 hours would at least cut down the number of bot runners. A good question is how dedicated they are to keeping the bots running. One answer is not much at all - they're obviously lazy since they want cheap and easy traffic. Another is very dedicated since they took time to research and setup the bot.

Now back on topic. I just remembered an idea I thought of a while ago and this is the perfect thread for it.

One reason for traffic is to rank popular places. So I started thinking about ways to determine how popular a place is. I started thinking about how others might do it. "Others" in a very general sense and TV shows and Nielson ratings came to mind. They can't determine what every single person is watching but they can query a subset of TV viewers.

Hence the idea for Second Life™. Count the traffic for people at random. When a person logs in, they will be picked (or not picked) at random. If picked, whereever they go, their traffic will be accounted for as it is now. They will NOT know whether or not they're picked. Once picked (or not picked), relogging won't change that status until the next day.

For TV shows, it's impossible to account for all viewers. For Second Life™, it's actually undesirable to account for all resident's traffic.

The problem may be that bot runners will absolutely flood the system with 100s (1000s?) of bots that log in/out. Combine this with CAPTCHA login though and maybe it'll work. Just throwing it out there.

--Hugsy

I know about a very clever way that beats it. I don't want to repeat it here, but I don't think that most bot runners would have the capability, resources, or desire to do it. It could make a very nice business opportunity for a small handful that did though.


This is so good that I almost decided to pay you the $2000. If we had millions of concurrent users and could easily support 100 or so per sim, this could be perfect. To get a good measurement for enough parcels to make it worth while, you'd need to make the sample group such a large percentage of the people that logged in, that I think bot runners would actually be encouraged to log in more bots. They would at least have no motivation to log in fewer. This is still the best idea I've heard so far in the history of SL traffic discussions.

Build on that to make the random logging apply only to Premium accounts.
That would only skew results if it were the case that Premium accounts have a completely different set of interests to non-Premium accounts.


Bot runners would have to create and log in a large number of Premium accounts with no guarantee that their traffic was being measured.
It is true that a Premium on the US$70 annual basis gets the bulk of the money back in L$ stipend. It is also probably true that many bot runners would have issues with coughing up x times US$70 with no idea of how effective the bots would be.

I certainly wouldn't expect a flood of Premium bots on that basis.


It could be argued that the random samples would have to be large in order to be representative. If a Premium bot-runner believed that they had a 50/50 chance (say) of having a bots traffic measured, would that justify the costs? Whatever the sampling, the incentive to run paid-for bots would be reduced.

Enlarge the sample size by using the last 30 days of visitors. Accounting for random premium accounts only, with CAPTCHA login, and over 30 days is quite an interesting concept.

--Hugsy

What we need isn't a single 'catch-all' measure of popularity. The reason I say this is that while traffic may represent popularity in some places, it doesn't necessarily represent others accurately.

For example; entertainment areas are designed for people to go there and waste time. However, a good store is designed so that people can find what they're looking for quickly, therefore a measurement based on time is irrelevant. A better stat for stores might be a measurement of L$ traffic, though regulating that wouldn't be easy at all due to the way vendors, and especially networked vendors, work.

Still my personal favourite for searching is a tag-based system; it's being adopted all over the place these days. For those that don't know a tag-based search is basically where you get to specify a selection of words relevant to your parcel/classified etc., in the case of my proposal this would be limited. Advantages include:
- It's a-lot easier to give measures of relevance in tag-based searching.
- If you restrict the number of tags you can attach to a parcel/classified, then it forces people to choose more relevant tags rather than word-spamming.
- If a person is abusing a tag then it's easy to report it and have that word banned from their listing for being misleading. It could even be achieved by a self-policing mechanism in future.
- Tags can give more useful statistics by showing land-owners/sellers how often a particular keyword resulted in a teleport. This could be used to produce 'tag-clouds' so that searchers can see at a glance the most relevant tags for a search-result.
- Tags can also significantly reduce the cost of indexing, since you only need to process up to (for example) 10 words instead of full-descriptions.

I'd give a link to the JIRA entry I posted for it, but the piece-of-crap JIRA isn't loading for me AS USUAL. If it is available for anyone reading this, just search for "Tag-based search" and it should come-up.

40 Premium acount bots only cost $1.67 US per day at the current limit order sell rate on the Lindex when you account for the stipend. Then you can use or rent out their tier to lower the price further. If only half are counted, then each one that does get counted is worth twice as much, so that would balance out. You would need a population big enough that the sample size would be such a small percentage to make the bot running inconsequential.

mmm Lots of interesting stuff here,so here's my tuppenceworth.
Given that the firm needs revenue from advertising .Customers want value from their add ie:
A pays 1000 for an add and gets more hits
B pays 50 and gets less
I cannot see this system changing nor can it ,the firm needs the revenue to survive.This then I have taken for granted.
One could still have a fairer search system.Scrap traffic ,after all people seem to agree it,s meaningless now.
If search was updated with categories similar to SLX or onrez search would function better.
What if those categories were then ranked in order of buyers popularity.
1 sex gear
2 fashion
3gadjets
4teddy bears
5 darning needles
ect, ect. or whatever order the stats say is correct.Ok.
Now you use transaction history to award hits,more sales, more points,more popular.
Job done.
The higher a place comes within a particular category then its a popular place.
You might still be the most popular place selling darning needles or whatever,to me this seems fairer.
If this seems stupid it's just the ramblings of some tired old tard ,as my sons say.

Search needs to be decoupled from search.

I would have a checkbox that removes traffic as part of the search.

Then I would use profile "picks" as a way to vote for sites. The more sites pointing to a site the more votes that site has. Having the same site listed multiple times would only count as one vote.

In a way this is how Google works.

The problem with traffic right now (in my opinion) is the fact that it treats everyone the same (1 human contributes the same amount of traffic as a zombie camper or a camp bot) and it's entirely predictable (I want x traffic so I need y "campers" 24/7).

Restricting the pool of contributors to premiums would be too small (there's only 100,000 of those), but payment info used should work fine. Not everyone needs to contribute to search, you just need a representative group that consists of unique individuals.

The advantage of payment info used is that it's tied to a RL identity and everyone only has a finite number of those (your own, but "gamers" could use their spouse's, child's, parents', etc) and every payment info alt would share that identity.

If you only keep track of traffic generated per RL identity instead of avatar you sidestep most of bot/alt issues. You could log 10 payment verified alts/bots on, but only one avatar's traffic would count since all 10 alts share the same RL name and if you assign which avie gets counted at random you don't even know which one would count and which ones wouldn't.

To reduce the benefit of bribing actual humans into camping you can only have a certain percentage of them count which removes the "predictable" aspect. If you have 20 campers they might all count, half might count or none of the may count. It doesn't remove camping as a way to "cheat", but it makes it less guaranteed to work and you have a much smaller pool of potential campers already as well.

You can still cheat, but you can *always* game whatever is there and while the above wouldn't be perfect, it could level the playing field considerably.

(Gift cards are one problem, but you'd need to get 40 of those to get 40 alts/bots and you still wouldn't know how many of those 40 actually contribute traffic at any given time and LL's record of accepting prepaid credit cards is sketchy at best. The only way to counter "abuse" like that is for LL to make it an offense, which means ARs which they'll never agree to. Either way, traffic would still be far more accurate in general than it is now)

(Edited to add that adding other little things like capping the maximum of traffic any avie can contribute to traffic would help things some as well. I don't think most people spend longer than 2-4 hours at one single place so traffic doesn't really need to count beyond that)

---

As far as getting rid of traffic entirely: at the end of the day you need *something* to order Search / Places. The most common retort seems to be "well, just sort by relevancy" but relevance is a concept that's based on some other metric. A isn't more relevant than B in and by itself, you need a number to base relevancy on.

Another thing that frequently comes up is "just sort them randomly". I don't think random is in any way fair though. If store A and B sell the same thing, but A is legitimate, provides quality and is truly popular while B just is small and sells rubbish then A should rank before B. Both because A is clearly the more relevant result and because A really earned their ranking.

(Pre-snarky disclaimer: even though it may be phrased as "fact", everything above is obviously merely my opinon and my personal view on how things should be )

I was surprised to see SL want to alter the traffic system rather than scrapping it alltogether. I vote for no traffic system.

Well, there are two things here: quality of search results and traffic measurements. The traffic measurements can still be useful even if they aren't used to sort search. Parcel owners can use that stuff!

Anyway, no matter what the outcome of the current traffic discussions, LL already has plans to get rid of the current Places tab and replace it with one driven by new engine where traffic already has a reduced role. See Milestone 2 at http://wiki.secondlife.com/wiki/SL_Search

Well, that was my point. The bot and camper problem is almost exclusively situated in the commercial sector. Some clubs might have campers, but those are really less of an issue and (to be honest, not really a problem - clubs have always paid people to hang out in them as long as I've been in SL.) You won't find any of the the non-commerical places using bots or campers.

I also don't think BIAB will be a problem with search results. A crap red shoe is still a crap red shoe even when you see it on the page 3 times. Someone looking at red shoes will just keep scrolling because they've seen that one before. People doing new, original and creative stuff will actually do more business because their stuff has a chance of actually being seen which it doesn't currently as they show up on page 30 of the search results and a shopper will have settled on something else within the first 10 pages.

Again, this can only work for places that sell things. If you did a search for Jazz Club on your system, you would only find places that sold items for jazz clubs.

That's the new Search All page. It does nothing to suggest how much time people are spending at a location, and is already being gamed out the wazoo with paid profile picks.