How to get 5 real choices when selecting a new Linden Home (Premium)

[Chloe Bunny]

15 hours ago, Marisa Starbrook said:

The few times I’ve had Premium Plus, I haven’t been enthralled by the new themes. I’ve wanted to love them, but ultimately I’ve used the perk of selecting a home and then downgrading my account to Premium. There are a lot of primo spots that have been locked in Maintenance for months. Instead of playing game of homes, I’ve switched to be an “in maintenance” house hunter. 

Me too, I've upgraded for that perk but then also for the free uploads which is a boon only if I have a lot of new builds to make otherwise not. I see a lot of lovely places listed as Linden Home that I think are beautiful and sometimes tempted, right now not although I am a sucker for railway homes - the ranch railway is looking great 

[Raspberry Crystal]

2 hours ago, elleevelyn said:

there is the list of AllHomes and there is the list of AvailableHomes, not necessarily the same thing as some homes can be broken and marked as NotAvailable, and some whole regions also marked as NotAvailable

the likely explanation (from everyone's experience) is that the AvailableHomes list is relatively quite small. Topped up on a needs-must basis  with some recently abandoned Homes and some from the not recently abandoned AllHomes list

it could be that the re-population is triggered when the claimed level reaches some predetermined point. Same as how caches work

say the AvailablelHomes list was 100.  And say at 80% claimed then re-populate. with 80 new selections added to the 20 remaining. Problem is that can get stuck on 21. A catatonic state. Everyone playing is abandoning the same 21 Homes over and over. it could be triggered at 60 rather than 80. It could be 50. Dunno. But it can get stuck, go catatonic, when everybody is abandoning within the same period. the abandons already on the Available list staying on the list

we could think maybe set the trigger to quite low. Like at 20%. Problem is that if we don't flush the whole cache then we got people playing for the same 80% of undesirables, and when ire-populate, only 20 new additions are added. Can end up with 100% undesirables, which only changes when 21 people give up and keep the undesirable Home they don't want

having a short AvailableHomes cache list is not a issue in itself. The issue is how and when iit is re-populated - full fllush on trigger is better than top-up.  Still left with how to fill the cache from the AllHomes list excluding the NotAvailables. Most efficient way is feistel network or knuth shuffle if the widest range of cache selection is the goal

ps add. I not be surprised at all if it uses a cache. Webservers like caches. From all of the experiences that residents have had, then Somebody Linden coded it up as a top-up cache rather than a full flush cache. Because top-up is typically how caches work. Like our SL viewer cache. when it gets full then remove some from the top to make room for new stuff

I found that post easier to understand, thank you! The idea of a cache sounds very plausible, and it does seem as if the process gets stuck with less favoured homes churning around the pool. Now that there are so many styles to choose from there will be less motivation for people to choose a location they are not keen on, as they can switch over to a better performing cache of homes in an alternate home style.
I wonder if the system is manually 'reset' from time to time? ie 'given a hard flush'!

[elleevelyn]

3 minutes ago, Raspberry Crystal said:

I wonder if the system is manually 'reset' from time to time? ie 'given a hard flush'!

could be yes. Region servers can be manually restarted when they get stressed. So is no reason the Linden Homes server couldn't be manually reset

[xViXeNx71]

5 hours ago, Raspberry Crystal said:

Now that there are so many styles to choose from there will be less motivation for people to choose a location they are not keen on, as they can switch over to a better performing cache of homes in an alternate home style.

This!

When a theme is high in demand, and only has a couple of available homes, people will just take whatever they can get, even if it isn't the best location. But for a theme with 2550 available homes, where you can see many empty waterfront parcels, people are not gonna settle for a landlocked one.

[Prokofy Neva]

22 hours ago, elleevelyn said:

i agree. Random, meaning uniform distribution, is not fair in some use cases. The allocation of Linden Homes is one of those cases

as you say with uniform distribution is possible to always get the same number. When so we can say that this is a catatonic state for the user on the receiving end. Is not enough for the implementer to say: is random so is fair. Is not fair at all for the individual receiving. Linden Homes should not be a roulette game

with serial distribution of a set of ordinal numbers [1,2,3,4,5, etc] where the last abandoned home is added to the end of the list then when there are say 100 homes available then at 5 pulls a day will take 20 days for that last home to show up. Which is fine when the home is a undesirable, but not so fine when is a desirable. (Desirable being subjective to the individual)

In 1971 Horst Feistel invented a block cipher algorithm (feistel network), which is pretty much used in most encryption methods in some modification or other. The purpose of a block cipher is that  the output can be decoded to get back the original value. it follows that when we encode all inputs/values of a set then no two outputs will be the same, and they can be decoded back to the inputs without loss

given this then a feistel network can produce a "random" arrangement (permutation) of a set of ordinal numbers. So with [1, 2, 3, 4, 5] there are 120 arrangements: 5 * 4 * 3 * 2 * 1 = 120. Which can all be produced given a "seed' in the range [1 .. 120]

taking the cookie idea mentioned earlier then it need only contain the set id (type of home), seed (arrangement id) and the serial index number. Incrementing the index by 1 to produce the next home

looks random to the user but it isn't - is deterministically biased compared to uniform distribution. A thing is that many people like bias when it comes to 'random' meaning that they like it when it comes out [3,1,5,4,2] . Much more than they like [1,5,1,3,3] or [2,2,1,1,2]. And when it comes to Linden Homes then many means most if not all people

has been about 53 years since Horst Feistel invented this algorithm, so is not like is still waiting to be discovered. When a SL resident can implement a arithmetic feistel network in LSL then this is not something beyond the ability of a Linden professional programmer to do

for sure the programmer (more the product manager) has to decide what happens on change

Like when the next available home has been taken by another person. The most common method is to increment the index til it matches a available home. Understanding that when the index exceeds the magnitude, the index rolls over to 1 and starts over

the other change happens when the seed is changed. Which requires a reset. On reset is when the homes abandoned since last reset can be added to the available list, homes claimed removed. Which gives a new magnitude (number of available homes)

 

ps add. Alternatively to do the same thing then create an index list and do a Knuth shuffle on it (similar to llListRandomize). A feistel network doesn't require a pre-shuffled index list, which is the only difference in effect

 

pps add. For those still reading who maybe into this kind of thing

yourNewHome = llFrand(llGetListLength(AvailableHomes));

this is pretty much how it works now. llFrand (equivalent) produces a uniform distribution (like a roulette wheel)

if as a Product Manager, my programmer offered this as a solution for my Linden Homes paying customers then I would give them my stoneface and say "Really!"

and if they stoneface me back I would say "REALLY!" and not give them any cream cake for afternoon tea. Til they come back from reading up on Knuth, Feistel, Galois, Margolis, et al. and go to me "Oh! yeah"

and I will say "Yeah!" and off they will scurry on the promise they will get two cream cakes next afternoon tea. Seeing as how I ate their other cream cake myself due to all the agony I was feeling when I had to say "REALLY!" 😻

annnd if I was the programmer and my product manager said we going with the one liner solution then I put in for a transfer to another team. There are somethings that all the cream cakes in the world can't compensate for

I don't think you're admitting my point, which is that in a very finite world like Second Life, serial processing in a line, with first available coming up, then next available etc. is more productive and fair than any form of randomizing whatsoever.

The 5 pulls are not in isolation. For one, people use 5 pulls per account, and they may have dozens of alts. For two, lots of people are pulling, so it's not in isolation. That's why there are scarcities. But the Lindens can't afford to add more sims that will be empty when the fad dies out.

 

[elleevelyn]

4 hours ago, Prokofy Neva said:

I don't think you're admitting my point, which is that in a very finite world like Second Life, serial processing in a line, with first available coming up, then next available etc. is more productive and fair than any form of randomizing whatsoever

addressing your point

when the AvailableHomes list/cache is a subset of AllAvailableHomes then a serial index can make sense and be the most efficient

how might this be done ?

1st pull (any person) gets 1
2nd pull (any person) gets 2
3rd pull (any person) gets 3

and so on til the serial index reaches Magnitude/listlength. On which: Claimed Homes removed from the list/cache. Abandoned Homes on the list not removed. New Magnitude. Serial index reset to 1. And start over

in this case, from the users pov, serial index increment is better than random() as it means they will get different Homes in each of their upto 5 daily pulls - assuming the list has 5 or more Homes on it

at some point Magnitude lowers to some level, on which the AvailableHomes list is re-populated

on re-populate, shuffle the list (feistel or knuth). or more correctly shuffle the list index. The serial index pointing to the list index which points to the Home entry in the list. We have shifted the shuffle from the serial index to the list index

why shuffle ? 2 homes from Region A. 3 homes from Region B. 1 home from Region C, etc. We don't want to present these homes in block order. Caveat: In either case (block order or shuffle order) is still possible for a person to get homes on the same region in their 5 pulls. Depends on how many people are playing at the same time and their combined RateOfPlay pattern. Altho when fewer people are playing at the same time then shuffle order is better for these fewer players. The fewest being 1 player

remaining questions are:

1) when the AvailableHomes list is re-populated should all remaining entries be flushed, the list filled with a new set from the AllAvailableHomes list ? In my view the answer is yes
2) when should the re-population occur ? In my view the answer is when the list lowers to some level, and every X hours regardless of level. Every 24 hours, 48, whichever

picking which homes on which regions to populate the list with is another exercise. The algorithm to do this is a feistel network if we want to cycle AllAvailableHomes thru the AvailablebHomes list with the widest distribution

i not get to deep into how this algorithm works in this case. So I try keep it brief

example: 4 regions with 4 homes on each. 4 blocks of 4 homes. Total 16 homes. The goal (widest distribution) is to place 1 home from each block into a set of 4. 4 sets with a home from each region. Present each set in order

feistel network does this as is fundamentally a block cipher. Arithmetic feistel network deployed when the number of homes on each region and/or the number of regions is not in the binary sequence

note that the distribution goal can be lower. Like 2 homes from any region and 1 home from any of the other 3 regions. The algorithm when coded as a generator stops sooner

will Linden do any of this ? Maybe, maybe not, dunno

my main point in all of my posts on this topic is that "random-looking" is better than "random" from the Linden Home owners pov. Random-looking is the outcome of a serial process applied to an arrangement. [1,2,3,4,... etc] is an arrangement

as humans we may not see [1,2,3,4, ..10] in the same way that we see [6,3,2,5, ..., 8] as this latter "looks random". Algorithmically tho is no difference

[Matthieu Quander]

12 hours ago, elleevelyn said:

addressing your point

when the AvailableHomes list/cache is a subset of AllAvailableHomes then a serial index can make sense and be the most efficient

how might this be done ?

1st pull (any person) gets 1
2nd pull (any person) gets 2
3rd pull (any person) gets 3

and so on til the serial index reaches Magnitude/listlength. On which: Claimed Homes removed from the list/cache. Abandoned Homes on the list not removed. New Magnitude. Serial index reset to 1. And start over

in this case, from the users pov, serial index increment is better than random() as it means they will get different Homes in each of their upto 5 daily pulls - assuming the list has 5 or more Homes on it

at some point Magnitude lowers to some level, on which the AvailableHomes list is re-populated

on re-populate, shuffle the list (feistel or knuth). or more correctly shuffle the list index. The serial index pointing to the list index which points to the Home entry in the list. We have shifted the shuffle from the serial index to the list index

why shuffle ? 2 homes from Region A. 3 homes from Region B. 1 home from Region C, etc. We don't want to present these homes in block order. Caveat: In either case (block order or shuffle order) is still possible for a person to get homes on the same region in their 5 pulls. Depends on how many people are playing at the same time and their combined RateOfPlay pattern. Altho when fewer people are playing at the same time then shuffle order is better for these fewer players. The fewest being 1 player

remaining questions are:

1) when the AvailableHomes list is re-populated should all remaining entries be flushed, the list filled with a new set from the AllAvailableHomes list ? In my view the answer is yes
2) when should the re-population occur ? In my view the answer is when the list lowers to some level, and every X hours regardless of level. Every 24 hours, 48, whichever

picking which homes on which regions to populate the list with is another exercise. The algorithm to do this is a feistel network if we want to cycle AllAvailableHomes thru the AvailablebHomes list with the widest distribution

i not get to deep into how this algorithm works in this case. So I try keep it brief

example: 4 regions with 4 homes on each. 4 blocks of 4 homes. Total 16 homes. The goal (widest distribution) is to place 1 home from each block into a set of 4. 4 sets with a home from each region. Present each set in order

feistel network does this as is fundamentally a block cipher. Arithmetic feistel network deployed when the number of homes on each region and/or the number of regions is not in the binary sequence

note that the distribution goal can be lower. Like 2 homes from any region and 1 home from any of the other 3 regions. The algorithm when coded as a generator stops sooner

will Linden do any of this ? Maybe, maybe not, dunno

my main point in all of my posts on this topic is that "random-looking" is better than "random" from the Linden Home owners pov. Random-looking is the outcome of a serial process applied to an arrangement. [1,2,3,4,... etc] is an arrangement

as humans we may not see [1,2,3,4, ..10] in the same way that we see [6,3,2,5, ..., 8] as this latter "looks random". Algorithmically tho is no difference

That would certainly work, but I don't see the need for all that extra effort.  Simply using AllAvailableHomes, and sorting it by the oldest abandoned date meets the need.  It ensures house hunters get 5 unique homes (assuming the count of AllAvailableHomes >/= 5), and that all homes in the list eventually rotate up to the top of the list.  List shuffling will occur naturally through variations in house retention lengths.  

[Raspberry Crystal]

1 hour ago, Matthieu Quander said:

That would certainly work, but I don't see the need for all that extra effort.  Simply using AllAvailableHomes, and sorting it by the oldest abandoned date meets the need.  It ensures house hunters get 5 unique homes (assuming the count of AllAvailableHomes >/= 5), and that all homes in the list eventually rotate up to the top of the list.  List shuffling will occur naturally through variations in house retention lengths.  

I think that might need a tweak to allow for those occasions when available homes have never been claimed.

[Sid Nagy]

For my business I switched a hundred times or more between themes and houses.
In all themes I had houses over time that I would never have given up if I wanted to live there. Except the Newbrooke type. They only seem to be situated in average and boring places, sim after sim.

I never had the same location so far, not even once.
Isn't this subject a bit of "I want it all, I want it now" too?

Edited January 10 by Sid Nagy

[Raspberry Crystal]

1 hour ago, Sid Nagy said:

For my business I switched a hundred times or more between themes and houses.
In all themes I had houses over time that I would never have given up if I wanted to live there. Except the Newbrooke type. They only seem to be situated in average and boring places, sim after sim.

I never had the same location so far, not even once.
Isn't this subject a bit of "I want it all, I want it now" too?

I don't think so Sid, I'm not sure where you got that from.
How long ago did you do your mega-hopping and what kinds of intervals did you leave between your search days?

[Sid Nagy]

31 minutes ago, Raspberry Crystal said:

How long ago did you do your mega-hopping and what kinds of intervals did you leave between your search days?

I hop between homes a few years already, sometimes several times a week, sometimes even a few times a day.
On average I'm never longer than two weeks on the same spot. I'm totally not interested in location, that helps of course.  I owned several beautiful locations over the years for a short time. But I live elsewhere, so I released them when I no longer needed them.

Most of the time I switch theme, that helps too I guess.
Only when I want a house boat on a square parcel, that can take a few quick changes. Sometimes even more than the allowed number a day.

If I want to stay in the house boat theme, I always require a new one within a minute or so after abandoning the old one. That way the old boat is not on the market yet. At least that always worked that way so far for me.
 

Edited January 10 by Sid Nagy

[elleevelyn]

4 hours ago, Matthieu Quander said:

That would certainly work, but I don't see the need for all that extra effort.  Simply using AllAvailableHomes, and sorting it by the oldest abandoned date meets the need.  It ensures house hunters get 5 unique homes (assuming the count of AllAvailableHomes >/= 5), and that all homes in the list eventually rotate up to the top of the list.  List shuffling will occur naturally through variations in house retention lengths.  

is true about the natural occurrence of variation due to house retention lengths (rate of play)

a issue with the sort by abandon date method is that it can be gamed, resulting in people not into gaming wondering why somebody else seems to get nice homes at a frequency rate greater than theirs. Which creates a perception problem among the customer base about fairness

i am not going to get into how this can be gamed. Just say that some people will do it.

from a implementation pov, knuth shuffle is O(n) . quicksort is O(n logn). Is a bit quicker overall to do a shuffle. And given that shuffle or sort is going to happen one tiime, when the list is flushed, then I would go with a shuffle to avoid the gaming issue, moreso than it being a little bit quicker

[Matthieu Quander]

4 hours ago, Raspberry Crystal said:

I think that might need a tweak to allow for those occasions when available homes have never been claimed.

Nah.  They likely get a default date that is well in the past (or easily could).  They would all collect at the top of the list until claimed.  

Edited January 10 by Matthieu Quander

[Matthieu Quander]

26 minutes ago, elleevelyn said:

is true about the natural occurrence of variation due to house retention lengths (rate of play)

a issue with the sort by abandon date method is that it can be gamed, resulting in people not into gaming wondering why somebody else seems to get nice homes at a frequency rate greater than theirs. Which creates a perception problem among the customer base about fairness

i am not going to get into how this can be gamed. Just say that some people will do it.

from a implementation pov, knuth shuffle is O(n) . quicksort is O(n logn). Is a bit quicker overall to do a shuffle. And given that shuffle or sort is going to happen one tiime, when the list is flushed, then I would go with a shuffle to avoid the gaming issue, moreso than it being a little bit quicker

Gaming the system already occurs.  I'm sure some would celebrate its demise, but not all, and it's not the issue that it once was.  I'm simply looking at this from the perspective of how to solve the problem posted in the OP as easily as possible.  Changing which column is used for determining table order would take practically no time at all to implement.  There may be some very valid reasons why it would be a bad idea, but I haven't seen any yet.  

[elleevelyn]

3 hours ago, Matthieu Quander said:

Gaming the system already occurs.  I'm sure some would celebrate its demise, but not all, and it's not the issue that it once was.  ....  There may be some very valid reasons why it would be a bad idea, but I haven't seen any yet.  

gaming a pseudorandom system is about identifying the source (seed) of the pseudorandom number generator, the method use to obtain the seed, and the method used to generate the outputs. There is no other identifying/discovery method that can gain us information to better predict the next number. Is why people give up on doing this

with a sorted list then we can gain information independently of the system, that will indicate to us at what time a home we are interested in is most likely to appear. How we gain the information I am not going say in any detail, but is pretty trivial to do, and when people do work it out then they are going to do it. A lot more people than previously who tried to game a pseudorandom number generator

edit: for completeness

with a shuffle list we can't predict the order of the homes remaining on the list, unlike the sorted list where we know the list is ordered by abandon date

 

Edited January 11 by elleevelyn

[elleevelyn]

3 hours ago, Matthieu Quander said:

Changing which column is used for determining table order would take practically no time at all to implement. 

is pretty simple from a coding pov to do yes. As simple as:

list homes = llListSort(homes. 1, TRUE);

list homes = llListRandomize(homes, 1);

product manager goes thru the impacts these two methods will have on the customer base and picks one

 

Edited January 11 by elleevelyn
1)

[Matthieu Quander]

I'm bumping this because I'm finding it extremely challenging to effectively house hunt.  I've been casually shopping for a new Log Home, and even though there are over 2000 available, I continually get the same parcel twice or even three times in a row.  I never get 5 unique choices anymore.  I can understand why this happens when supplies are low, but this really should not occur otherwise, and it is extremely frustrating.    

The usual tricks, like waiting for someone else to grab the home I just abandoned is futile when a theme isn't in demand.  I understand that I can upgrade to Premium Plus, and select any available home I want, but I should not have to resort to that if my criteria for an acceptable home are fairly broad, and I simply want to randomly shuffle through available inventory over time.  

I fully believe that this impacts the occupancy of some themes, as I expect people give up and move on to other themes, or worse, just downgrade from premium altogether.  I'm not threatening to do that, but as someone who usually enjoys the Game of Homes, I am likely to demote an account or two if it brings me no joy.

Multiple people have offered suggestions on how to resolve the issue, but of course none of us has any insight into how the process actually works, so none of them may be viable.  It would be nice to know if this is the case, or if there is the possibility of this getting resolved at some point.

Are there any moles or Lindens in a position to address this issue?  

 

[diamond Marchant]

1 hour ago, Matthieu Quander said:

Are there any moles or Lindens in a position to address this issue? 

Clearly they could change the algorithm. However, there appears to be a financial incentive to keep things as they are, as the only guaranteed way of getting a preferred home is to upgrade to Premium Plus.

There is a new feedback system ... perhaps they might  respond there.

Too bad the month-long 4-region Belli 5th Anniversary Celebration has no Meet the Lindens/Moles event where questions could be posed. We were encouraged to submit questions to Patch a few months ago... but it remains to be seen if those will be addressed.

[Persephone Emerald]

1 hour ago, diamond Marchant said:

Clearly they could change the algorithm. However, there appears to be a financial incentive to keep things as they are, as the only guaranteed way of getting a preferred home is to upgrade to Premium Plus.

There is a new feedback system ... perhaps they might  respond there.

Too bad the month-long 4-region Belli 5th Anniversary Celebration has no Meet the Lindens/Moles event where questions could be posed. We were encouraged to submit questions to Patch a few months ago... but it remains to be seen if those will be addressed.

Maybe those questions and others will be addressed in June, though I think there are some very significant issues that need to be publically addressed *before* the SL 21st Birthday celebrations. Linden Lab really can't expect those to just be swept under the rug and forgotten.

Most SL users can't afford to go Premium Plus just to get a Linden Home they like. LL must do something to improve the user experience for being able to pick a *new* Linden Home parcel when such are available. Maybe all it would take is for freshly abandoned homes to go to the bottom of the list?

Edited April 15 by Persephone Emerald

[Sparkle Bunny]

3 minutes ago, Persephone Emerald said:

Maybe those questions and others will be addressed in June, though I think there are some very significant issues that need to be publically addressed *before* the SL 21st Birthday celebrations.

I'm pretty sure this is why we don't have any Q&A sessions at B5A. Bluntly, the Lindens have more important things to worry about than Bellisseria right now.

[Matthieu Quander]

11 minutes ago, Persephone Emerald said:

Maybe all it would take is for freshly abandoned homes to go to the bottom of the list?

That is one of the suggested fixes, but it is likely more complicated than that. 

I'm not cynical enough to argue that they are leaving it this way on purpose.  I tend to believe that the current situation is more likely to cost them Premium memberships than it is to result in people upgrading, but that is pure speculation on my part.

All I know is that residents should not repeatedly be given the same home choice over and over when there as an ample supply of alternatives.  This has been an issue since the L1 days, but feels worse currently, most likely because I am shopping a theme that doesn't see much activity.  

[Love Zhaoying]

3 minutes ago, Matthieu Quander said:

All I know is that residents should not repeatedly be given the same home choice over and over when there as an ample supply of alternatives.  This has been an issue since the L1 days, but feels worse currently, most likely because I am shopping a theme that doesn't see much activity.  

Yeah, how hard would it have been to implement a "random" selection (or even a "cycling" selection)?

Maybe there were "reasons" for doing it this way, but unless several people want to play "the game of homes", then nobody wins if nobody wants that one home that keeps being displayed!

Since Bellisseria is one of the "key offerings" in Second Life for Premium Users, you'd think this would be a high priority fix, sometime in the last 5 years.  It's not like nobody filed "feature requests for it" (I'm assuming).

[Matthieu Quander]

5 hours ago, diamond Marchant said:

There is a new feedback system ... perhaps they might  respond there.

Good idea.  Everyone please go upvote, and tell your alts about it too:

https://feedback.secondlife.com/ldpw/p/change-linden-home-selection-process

 

Edited April 15 by Matthieu Quander

[karynmaria]

7 hours ago, Matthieu Quander said:

I'm bumping this because I'm finding it extremely challenging to effectively house hunt.  I've been casually shopping for a new Log Home, and even though there are over 2000 available, I continually get the same parcel twice or even three times in a row.  I never get 5 unique choices anymore. 

 

 

I hear you.  I even get the same parcels days later and even sometimes weeks later.  I enjoy hopping around and hunting for a new place cause I like to try different locations.  Even with the older homes I use to do the same thing.   I feel your frustration.

[Prokofy Neva]

9 hours ago, Matthieu Quander said:

That is one of the suggested fixes, but it is likely more complicated than that. 

I'm not cynical enough to argue that they are leaving it this way on purpose.  I tend to believe that the current situation is more likely to cost them Premium memberships than it is to result in people upgrading, but that is pure speculation on my part.

All I know is that residents should not repeatedly be given the same home choice over and over when there as an ample supply of alternatives.  This has been an issue since the L1 days, but feels worse currently, most likely because I am shopping a theme that doesn't see much activity.  

Yes, surely it's possible to do that? But maybe they have created a "stock list" and don't really give you "the entire list of what is really available". The thought occurred to me as I went through these same frustrations the other night.

I kept finding that the really desirable one could never be captured, even while empty -- the system kept giving me the dogs around it. Days went by, and nobody else seemed to get that gem, either. Is this because the Lindens have a reserve of "really nice" ones that they leave for the Premium Plus people to choose, i.e. all-sides protected, waterfronts, etc.?

This would be reasonable for them to do -- we don't know if they do this and they won't tell us.

But...How hard would it be to ensure that you never get the same one again??? So the reason they won't change this must be due to the fact that they are keeping a set list -- not ALL the choices, not the prime lots, but just a set list that they hope to unload.

Therefore, you are forced to play with several alts, at least to hang on to one more or less tolerable lot with one of the alts while you try with the others.

I've had times when I've settled on the first try because what I was doing with the house didn't require actual "living" and a "feel" for a neighbourhood, but the stamp venue inside or a story I was making with the house as a prop. But most of the time I want it to be tolerable to hang out in, at least a few hours a week.

Which is about all I can tolerate of Belli houses by contrast with my more spacious and in the end more diverse Mainland lots.

See my thread elsewhere here about Sakurasseria for additional thoughts. Also I'd be curious to know what you mean as a good house.

For me, a good house these days has to have at least two sides "protection" -- a Linden road, park, or waterway. Ideally, three or even a lucky four.  Any less, and I feel I'm in the Pete Seeger "Little Boxes" song. 

Edited April 16 by Prokofy Neva