WarcraftReamls.com
  FAQFAQ    SearchSearch    MemberlistMemberlist    UsergroupsUsergroups   RegisterRegister 
  ProfileProfile    Log in to check your private messagesLog in to check your private messages    Log inLog in 
Tank/Heal/Melee Dps/Ranged Dps
Goto page Previous  1, 2
 
Post new topic   Reply to topic    WarcraftRealms.com Forum Index -> Census Talks
View previous topic :: View next topic  
Author Message
Kosh



Joined: 01 Jul 2007
Posts: 84
Location: Somewhere on or near Earth
WR Updates: 1,210,548
Kosh WR Profile

PostPosted: Mon May 24, 2010 7:22 pm    Post subject: Reply with quote

aberman wrote:
Kosh wrote:
aberman wrote:
This would not have to be done live in the game at the same time as the "/who" lookup, but it could if programmed to do so.
Uh, how? An addon has no way to access anything outside the game, including Armory web pages.


Sorry I didn't answer this question.

It is clear that Census+ stores the data gathered in-game and maintains it in a file even when out of game. It is possible that when out of game, then user could click on a launching application (.exe file within the Census+ folder) that takes the in-game information which was gathered, access the armory via the web (still out of game), and import the missing information needed based on the players gathered while in-game. Since most people are dual-spec, you would probably use the one that has the "*" by it to indicate that is the spec with the best gear.
I'll have to come back later for the rest of your comments, but this one really caught my eye. There are problems with having an executable that you ask the users to subsequently run, though. First off, not everyone in the player base is running under Windows, so your ".exe" will not be usable to everyone who is otherwise using the add-on. Second, making the process of collecting / submitting census data even more involved will likely decrease total submissions. Third, even for Windows users, not everyone will be comfortable with having an add-on that includes an executable that they are asked to run (with good reason, as account compromises have occurred in this manner).

Personally, I am extremely opposed to using any add-ons that bundle an executable, and would stop using Census+ if that ever took place. Being bundled with some kind of (shell or other) script wouldn't be so bad, since anyone can examine what it does, but then that would probably exclude the majority of Windows users.

Ignoring the other issues with having each user run an Armory scraping program for the moment, doing such would be much less efficient (not to mention potentially being interpreted as a DDoS attack if / when several people happen to kick off the program at the same time). Any realm / faction which is "well covered" by several people will have mostly duplicate characters in each of those people's Lua files, leading to many duplicate pages being scraped on the Armory. I'll also point out, in case you didn't know, that unless you Purge, etc. your data in-game with Census+, the data file keeps growing with everyone you've seen during any run of C+, so unless you limited the look-ups to some arbitrary "last seen" window, you would be increasing the number of Armory queries (potentially quite dramatically) each time you would run the program after exiting the game.

Even though I think scraping the Armory is a bad idea for Census+ for several reasons, it would logically make the most sense for such activity to occur centrally (i.e., by warcraftrealms.com) to avoid so much duplication.

I am curious on where you got your information that "most [characters] are dual-spec". I took a quick visit to the Armory (man, it seems to get boggier every time I visit), and pulled up a few random characters. Yes, I only picked on 80's. Most of the ones I picked were not dual-spec, but I certainly didn't pull a statistically significant sample.
Back to top
View user's profile Send private message
Balgair
Araiceil


Joined: 30 Sep 2005
Posts: 1506
Location: UK
WR Updates: 11,068,258
Balgair WR Profile

PostPosted: Tue May 25, 2010 4:07 am    Post subject: Reply with quote

aberman wrote:
Balgair wrote:
There's also just plain fun specs, I've got a 30/28/3 priest and 23/37/1 warlock, both lvl 70 so ok may not be counted anyway, but they're not intentionally bad specs, just not-so-common ones that are fun to play (dps/soloing build for the priest; I hate shadow, love holy dps!) and perfectly adequate for soloing so I really do play with them. Would be a bit tricky to decide what to do with cases like that, since they're genuine really-used specs; should they be ignored just because they don't fit a pigeonhole determined byan arbritary group of people? Not arguing for or against, just mentioning it as a question Smile


Perhaps you guys are making me realize that these specs are slightly more common than I realize, though I don't know how much so. I bet it would be a useful feature to even be able to categorize the # of destro locks found by spec. As far as the lock thats 23/37/1, its easy, I just count the 37. As far as the priest is concerned, its a little more bothersome, though the simple answer for the purposes of a census is to count the 30 number. Either way, its still easy to categorize you as a healer (points in either holy or disc is more of a judgement call as to whether u will need to bubble more or heal more). The question as to whether you spec to heal the tank or spec to heal the raid is not as much of a concern for a non-lvl 80 that will not be doing end-lvl raiding. It would be more of a concern if it was a mix of disc and shadow (ie, for pvp). Its still possible for a program to distinguish between pve gear and pvp gear.

Also, what I still doubt is that many of you have level 80s with a certain gear score that is raiding as one of these dubious specs.

How many of you tank as death knight or feral druid, where a skilled raid leader upon inspecting your current gear set, could not tell whether you are intending to tank or melee dps?

I know that when I go into wowheroes, I can select to organize my dk search by tank or by dps, so they had to figure out a way to tell.


Are you wanting to cover only raiders with this, or everyone though? Warcraftrealms has always been about getting an overview of the whole population of a server, so to have a subset of the site only covering raiders seems a bit off to me, somehow - non-standard soloing/fun specs both exist and are used quite a bit, so should they be excluded just for the sake of narrowing stats down to only raiders? There's existing sites which do that already tbh.

Oh and as regards classing my mentioned priest spec as "healer", well that's a problem, because it's not a healing spec! It's a holy dps spec. Sure it's probably capable enough at healing but I've got dualspec and a proper holy healing spec as my other one... Razz

As to the point about are "most" people dualspec or not, that'd be an interesting one to get stats on tbh. My gut feeling is that "most" hybrid classes will be (tank/dps or healing/dps), but the pure dps classes will be less so, some will obviously want pvp and raiding specs, but others won't see much need for a 2nd one, certainly my 80 rogue has one spec only. My other 3 80s are hybrids thoughand all have dualspec. Small sample, yes, but I'd guess it's a likely one to carry over.
_________________
EU-Draenor:
-- Sagart - Tairbh - Buinne - Balgair - Eilnich - Ruire - Dubh - Laidir - Naomh --
Back to top
View user's profile Send private message
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Tue May 25, 2010 11:34 am    Post subject: Reply with quote

aberman wrote:
I think I need to continue to post for a while before we become clear on our thoughts. I also wouldn't mind chatting with you about this topic on vent.

All pures are automatically easy to determine whether they are melee or ranged dps. Its that simple. They are never anything but that. From that point, you only need to be able to classify a rogue as assassination, combat, or subtlety. My answer is pick the highest number. If they are 21/25/25, count them as an outlier (or maybe half the first 25 and half the 2nd). If they are 35/36/0, count the 36. Its not necessary to have 51 points in a spec to qualify as being that spec. My follow-up question that may be good guidance is that wow is planning to a mastery system in cataclysm. How would they handle it?


I am not available on ventrilo, but if you wish to chat, you can PM me here and request my MSN / Skype details.

Automatically easy? Well, there are two ways of looking at this and I am not sure, which you are after: Option A) Mage is automatically a ranged DPS... or option B: A mage is a ranged DPS, if (s)he has a sensible spec and decent gear.

Option A is indeed automatically easy... Option B, on the other hand is not.

If yoo go with option A, mages, rogues, warlocks and hunters are easily done. DKs are tank / DPS as are warriors, paladins are tank / DPS / healer as are druids and possibly shamans (I know some people who use and play shamans as tanks, though admittably not at level 80, priests are ranged DPS or healers. The end.

Now with option B... some characters of ANY class have a role definition of NONE. Also, hybrid classes (and here I mean with hybrid any class with more than one possible role) are likely to cause even more trouble...
For example, what is correct classification for a paladin who has mixed PvP / tank gear, but only one (good DPS) spec? If the correct classification is = throw out... the question is: How large a portion of a realm's population does get disqualified for whatever reason?

And how should one feel about a realm which looks for example like this (only including 80s): Tanks 1%, healers 2%, melee DPS 5%, ranged DPS 10%, unknown or non-viable 82%??? When you throw in the fact that there are a lot of realms with very low submission rates, the resulting data can only be described as faulty and/or unreliable.
Back to top
View user's profile Send private message
aberman
Guest







PostPosted: Fri May 28, 2010 9:21 am    Post subject: Reply with quote

1974ER wrote:
aberman wrote:
I think I need to continue to post for a while before we become clear on our thoughts. I also wouldn't mind chatting with you about this topic on vent.

All pures are automatically easy to determine whether they are melee or ranged dps. Its that simple. They are never anything but that. From that point, you only need to be able to classify a rogue as assassination, combat, or subtlety. My answer is pick the highest number. If they are 21/25/25, count them as an outlier (or maybe half the first 25 and half the 2nd). If they are 35/36/0, count the 36. Its not necessary to have 51 points in a spec to qualify as being that spec. My follow-up question that may be good guidance is that wow is planning to a mastery system in cataclysm. How would they handle it?


I am not available on ventrilo, but if you wish to chat, you can PM me here and request my MSN / Skype details.

Automatically easy? Well, there are two ways of looking at this and I am not sure, which you are after: Option A) Mage is automatically a ranged DPS... or option B: A mage is a ranged DPS, if (s)he has a sensible spec and decent gear.

Option A is indeed automatically easy... Option B, on the other hand is not.

If yoo go with option A, mages, rogues, warlocks and hunters are easily done. DKs are tank / DPS as are warriors, paladins are tank / DPS / healer as are druids and possibly shamans (I know some people who use and play shamans as tanks, though admittably not at level 80, priests are ranged DPS or healers. The end.

Now with option B... some characters of ANY class have a role definition of NONE. Also, hybrid classes (and here I mean with hybrid any class with more than one possible role) are likely to cause even more trouble...
For example, what is correct classification for a paladin who has mixed PvP / tank gear, but only one (good DPS) spec? If the correct classification is = throw out... the question is: How large a portion of a realm's population does get disqualified for whatever reason?

And how should one feel about a realm which looks for example like this (only including 80s): Tanks 1%, healers 2%, melee DPS 5%, ranged DPS 10%, unknown or non-viable 82%??? When you throw in the fact that there are a lot of realms with very low submission rates, the resulting data can only be described as faulty and/or unreliable.


Option A is not only correct, but the method listed in the article that I have attached.

Right now Census+ provides stats like:

Paladin: 15.4%
Death Knight: 13.8%
Druid: 11.4%
Priest: 9.9%
Warrior: 9.3%
Mage: 8.8%
Shaman: 8.4%
Rogue: 7.4%
Warlock: 7.4%

These numbers are taken directly from Hatch's article though the number that I have collected from Rivendare vary slightly.

I want it to provide numbers such as:

Paladin: 15.4%
*Holy: 4.2%
*Protection: 4.9%
*Retribution: 6.3%
Death Knight: 13.8%
*Blood - 6%
*Frost - 3.8%
*Unholy - 4%
Druid - 11.4%
*Balance - 2.9%
*Feral - 4.0%
*Restoration - 4.5%
Priest - 9.9%
*Discipline - 2.7%
*Holy - 3.4%
*Shadow - 3.8%
Warrior - 9.3%
*Arms - 2.8%
*Fury - 2.7%
*Protection - 3.9%
Mage - 8.8%
*Arcane - 4.0%
*Fire - 2.2%
*Frost - 2.6%
Shaman - 8.4%
*Elemental - 2.9%
*Enhancement - 2.0%
*Restoration - 3.5%
Hunter - 8.2%
*Beast Mastery - 2.1%
*Marksmanship - 2.3%
*Survival - 3.7%
Rogue - 7.4%
*Assassination - 4.1%
*Combat - 2.8%
*Subtlety - 0.5%
Warlock - 7.4%
*Affliction - 2.1%
*Demonology - 1.4%
*Destruction - 3.9%

Again, these are numbers taken directly from Hatch's article. This is what I would like Census+ to be able to do. For your convenience, I will repost it. Perhaps it has not been read by some:

http://esc-hatch.blogspot.com/2010/04/cataclysm-class-population-or-world-of.html

Hatch's math is as follows (though not explicitily stated, you can figure it out if you read between the lines):

Tank Percentage - 15.8% (Hatch used 16%, must have rounded)
*Prot Pally - 4.9%
*Death Knight Tank - 5% (Hatch's said 7% total between DK & Bear)
*Bear Tank - 2%
*Prot Warrior - 3.9%

Healing Percentage - 18.3% (Hatch used 18%, must have rounded)
*Holy Paladin - 4.2%
*Resto Druid - 4.5%
*Disc Priest - 2.7%
*Holy Priest - 3.4%
*Resto Shaman - 3.5%

Melee DPS - 32.0%
*Ret Pally - 6.3%
*Death Knight - 8.8% (13.8-5)
*Arms Warrior - 2.8%
*Fury Warrior - 2.7%
*Feral Kitty - 2% (4-2)
*Enh Shaman - 2%
*Rogue - 7.4% (only ever melee)

Ranged DPS - 34.0%
*Balance Druid - 2.9%
*Shadow Priest - 3.8%
*Mage - 8.8% (only ever ranged)
*Ele Shaman - 2.9%
*Hunter - 8.2%
*Warlock - 7.4%

Total - 100.1% (some rounding error, but close enough)

Anyway, I can definitely see Hatch's reason and I think his numbers are accurate, at least for the realm/realm(s) that he gathered from. Earlier in this post we saw a level 70 hybrid priest that was actually DPS and specced in holy/disc. While hybrid specs are somewhat common below level 80, I can honestly say that I have never raided with anyone in ICC that had anything like that and while it may be intelligent or fun for that individual to play that way, they would either be quickly booted from the raid if they specced that or I would quit before being saved to a fail raid. Perhaps some guy pulling 2k dps can sneak into my VOA group with that, but for the most part we can safely assume that prot = tank, holy = heals, & mage = ranged dps (at least among lvl 80s).

This is what I think that Census+ CAN and SHOULD be doing. Arranging Census+ and using data this way, a person can roll an alt which will likely have a much easier time finding a raid spot.

Since you have listed:

1% tanks
2% healers
5% melee DPS
10% ranged DPS
82% unknown

as your numbers, I am surprised that one intelligent person would assume that another intelligent person would want a search criteria that would draw these results. The method that is used in Hatch's article is clear and since I bothered to post it and say how thought provoking it was, why would I then bother to choose a search method so blatantly contradictory?

I thought the 35/36/0 and 21/25/25 was just a matter of fact question. Like, should we throw out these numbers if we see them? Not a bait to see if I would want a stupid search engine.

I think we are in agreement....Option A. Any questions?

Probably the one about distinguishing a DK tank from a melee DPS and distinguishing a bear tank from a kitty dps, right?
Back to top
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Fri May 28, 2010 1:29 pm    Post subject: Reply with quote

Since you quoted me extensively, a few notes.

1) English is not my native language. Though I am considered to have well above average knowledge of it, I have been known to misinterpet things.

2) I have never taken the official MENSA tests, but according to some unofficial ones I am likely to have a BELOW average IQ score.

3) I was not trying to bait you. I was trying to understand what you were after and to which extent you were willing throw out data. And the question about 35/36/0 and 21/25/25 specs was indeed just a matter of fact question.

Now that I have some more knowledge, I will continue. Using option A means that some things get easier, but they also introduce a major problem.

Because using option A means that all pure classes will be accurately defined, but not neccessarily correctly, the final results will be unbalanced. "Why?", you might ask. Because using that method will guarantee that ratio of tanks and healers compared to DPS is going to be wrong. Additionally, members of pure classes, which are not actually capable of doing their job well enough, will still be considered so.

"And why is all that happening, then?" Simple, because method A will throw out 0 members of pure DPS classes (while ignoring the fact that some characters would be refused a place in (some) groups), but will exclude a lot of non-optimal tanks and healers. And, in light of some of the comments of other posters, it will also cause errors in tank/DPS/healer ratios within all hybrid classes.

Also, as is now... Censusplus only covers characters that have been seen and submitted within the last 30 days. We already know that getting a 100% coverage of even a single realm is almost impossible. So, if everyone who has not been submitted gets auto-excluded, which in fact will happen, if you only use the census data as your baseline... well then, we already start with a situation, in which the class percentages are incorrect (more or less, depending on coverage)... If we then take those numbers, divide them into even smaller groups (into 4 in case of for example druids), the errors get even worse.

"How come?" Again, simple. We can't know, if the proportions we have covered correspond to the real distribution of classes, not to mention specific specs within those. Some realms have fairly good coverage and the results will be reasonably accurate... but what should and could we tell a person who is thinking about rolling a character on a realm, which has been censused ONCE in the last 30 days? Not much, I would say.

Slightly off-topic shortly: Rollie, if you are reading this... by how much (approximately) would the database expand, if data expired after 45... or even 60 days? 100%, 200%, 400%, more???

Returning to the theme, distinguishing between options, indeed... and returning to why I spoke about the option B. For the most part, we have been discussing specs... but gear also plays a role. You already stated that you would not want to group with people who have in your opinion bad specs. What about bad gear? Obviously, you don't want to take a freshly dinged 80 warrior with a mix of blue and green levelling gear to main tank ICC 25... but now the next defining question. If someone has a good and obvious tank spec, will (s)he automatically count as tank, even if the ability to tank is limited by the available gear? I am going to assume the answer is yes, because that would at least simplify the matter a bit.

Now for more complex issues... I am not sure how widespread dual-specs are in reality... but... your method A means that a character can only belong to one group at a time. Which brings us to the following question: A druid is wearing a full PvP set, has a good tanking spec and a suboptimal healing spec... Do we automatically put him into the "tanks" group, though the gear suggests a DPS (my understanding is that tank specs generally aren't good for Arenas/BGs, please correct me, if I am utterly wrong)? Even if the gear is ignored, having two available specs causes problems... especially if both are good, sensible specs. And even with gear included, a shadow priest MIGHT not have much difference in gear compared to a healer.

If I have understood what you are after correctly, I really suggest crawling the Armor(y/ies) as a better starting method. Because Armories will contain all existing characters, you will be able to begin with a much larger amount of data (using all existing 80s as the starting point, I guess), which you can then prune to your heart's content. Smile I don't know anything about writing crawlers though, so can't help with that.

This is now officially a huge wall of text, sorry! Sad Also, since typing this has taken a really long time due to various unavoidable distractions, this post might require (major) re-editing... :/

Edit 1: Already spotted some errors and typos. Sad
Back to top
View user's profile Send private message
Kosh



Joined: 01 Jul 2007
Posts: 84
Location: Somewhere on or near Earth
WR Updates: 1,210,548
Kosh WR Profile

PostPosted: Sat May 29, 2010 3:06 pm    Post subject: Reply with quote

aberman wrote:
...
Since you have listed:

1% tanks
2% healers
5% melee DPS
10% ranged DPS
82% unknown

as your numbers, I am surprised that one intelligent person would assume that another intelligent person would want a search criteria that would draw these results. The method that is used in Hatch's article is clear and since I bothered to post it and say how thought provoking it was, why would I then bother to choose a search method so blatantly contradictory?
Well, I thought it was a very reasonable question, given the abundant bias you have expressed throughout this thread towards (presumably end-game) raiding, and several comments about "bad" specs and discarding outliers. In fact, I am surprised that you chose the all-inclusive option "A".

I will reiterate that I am opposed to having this site scrape the various regions' Armories, but given that....

1974ER covered several implications of going the "inclusive" route, so I will just emphasize a couple.

This site only knows about characters that have been "seen" (online) during at least one person's Census scan. For a well-censused server / faction, the numbers will be fairly accurate, but are still only a floor (i.e. lower bound) to the actual number of characters. If none of the census-takers happen to be on while "serious raiding" is occurring, then some / all of those raiders will be missed. This will tend to still be fairly accurate for presenting numbers broken out by race / class, but the more you break down numbers with a margin of error into smaller buckets, the larger the margin of error of those buckets.

It really sounds like the best way to get the kind of information you're looking for would be to find an Armory-crawling site.

Getting back to the implications of the "all-inclusive" option, since all characters 10+ are counted here, you will get, for example. level 10 feral druid. All of the characters with "bad (raiding) specs" will also be put in one of the talent buckets (presumably with a rule deciding a winner in case of ties). The (to me) odd popularity of paying to turn off experience so characters can stay in a particular PvP bracket will also distort the numbers to some extent. From your described use of this additional information (level 80 raiding), it would seem that this would not be of much value.

Oh , from what I've read, in Cataclysm, whichever talent tree has the most points (even if just by one) gets the "mastery" benefits. I don't know how they resolve ties (if they do).

Something else I read about Cataclysm that probably makes you raider-types happy is that they are planning on moving all of the "tanking" talents into one tree for Death Knights.

Regarding "never raiding" with hybrid specs: Since you have indicated that you play on a low-population PvP realm, your experience may be distorted (the people still left on your server may be the most dedicated ones). They may also (gasp!) not be "serious raiders", so haven't tried to join a raid you were in. Either that, or they only raid with their guild or transfered to a less discriminatory server.

Regardless of what you seem to think, not everyone is a "serious raider", or even raids at all. Admittedly, the numbers you see on your low-pop server could be biased, though.

Oh, regarding the size of the database covering more than 30 days, as I understand it, characters aren't actually purged as soon as they haven't been seen for 30 days, they just stop being counted in "current" numbers. As I recall, Rollie has only occasionally purged lower-level characters, and then only ones not seen in several months.
Back to top
View user's profile Send private message
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Sat May 29, 2010 4:44 pm    Post subject: Reply with quote

It's very late (I watched the Eurovision Song Contest to the end) and I am very tired, so just a short comment to Kosh. I know that the characters don't expire quickly... What I was wondering about was the effect of extending the time of "characters seen" from 30 to 45, 60 or even 90 days in order to reduce the number of realms that zero out regularly.
Back to top
View user's profile Send private message
Kosh



Joined: 01 Jul 2007
Posts: 84
Location: Somewhere on or near Earth
WR Updates: 1,210,548
Kosh WR Profile

PostPosted: Sat May 29, 2010 6:53 pm    Post subject: Reply with quote

Ah, OK. I would also be curious as to how much the figures would change if you include "recent" aged-off characters.
Back to top
View user's profile Send private message
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Sun May 30, 2010 1:14 am    Post subject: Reply with quote

I know that there would be less zeroing out. As things stand now, I (and probably several other people) do a lot of "firefighting", trying to keep little censused realms from having no info. There are quite a few that drop to zero on fairly regular bases. Extending the data expiry time would mean that there would be more realms that have at least one census available and heavily censused realms would have quite accurate graphs as any changes in actual population would slow down.

Currently, a realm can zero out a maximum of 12 times during a year. With 60 day expiry that number would be halved. However, my main point and reason for asking the question was that I don't have any idea how much more storage space and bandwith such a change would mean. After all, WCR is Rollie's hobby, not something that puts food on the table of his family or pays the bills.

In relation to what Aberman is thinking, I think I will wait for him to post more before making further comments on the topic.
Back to top
View user's profile Send private message
Rollie
Site Admin


Joined: 28 Nov 2004
Posts: 5374
Location: Austin, TX
WR Updates: 480,131
Rollie WR Profile

PostPosted: Tue Jun 01, 2010 2:07 pm    Post subject: Reply with quote

Yeah, I could increase the time window without much trouble, I just wonder if people will feel the stats are less relevant if we include older data. I can perhaps give it a trial run and see what it does to the numbers.

Also for this particular discussion. In order to have that info available in game, then the AddOn would have to gather the data. Couple that with only being able to gather that data on people whom you are near will reduce the amount of data that would be collected.
Back to top
View user's profile Send private message Visit poster's website
Kosh



Joined: 01 Jul 2007
Posts: 84
Location: Somewhere on or near Earth
WR Updates: 1,210,548
Kosh WR Profile

PostPosted: Wed Jun 02, 2010 12:03 am    Post subject: Reply with quote

1974ER wrote:
However, my main point and reason for asking the question was that I don't have any idea how much more storage space and bandwith such a change would mean. After all, WCR is Rollie's hobby, not something that puts food on the table of his family or pays the bills.
That's what has me confused. As noted, the characters last seen > 30 days are already in the database (just currently excluded from site stats), so I am unsure where the "more storage space and bandwidth" enter in. Ignoring any constant "30" labels on pages that would / should be fixed, the basic change would be along the lines of this rough pseudo-SQL:

where char.last_seen_dt > to_date( today() - 61 )


If the character numbers don't tend to change much when you push the window back a little further. it probably isn't worth the effort to update everything. I don't have a good sense of how many people fall off the "active" list because they play less often than every 30 days, but still play at least every 60 days. I know I've done that a few times here and there, but it would be interesting to know how typical that is.

I would think that as long as the span is "fairly" small, that most people wouldn't have a problem with a somewhat wider span (my off-the-cuff guess would be that 90 days would be about the limit for "current" info).

If you do decide to bump up the window, you should probably also bump up the "yellow" / "red" low census warnings to match. After all, if 100 submissions is the "good" point in 30 days, then it follows that you would need 200 in 60 days to still be "good".

It's hard to tell whether the OP wants the extra information in-game, on the website, or both, but at least some of his stated reasons could be met on the site. Yes, having the information in-game would pretty much necessitate having the add-on gather it. This would be both (potentially) more accurate, depending on when you happen to Inspect them, and less accurate, overall, since the characters the census-taker is close enough to "see" is much smaller than the total characters scanned by /who.

A couple of things that occurred to me the first time Inspecting characters was brought up, but forgot to mention at that time: The first, I just touched on; the Inspect radius is very small. Skewing things even further, if the census-taker is a low-level "census-alt" or mail / AH / bank alt, then they will be mostly surrounded with other low-level characters, if they are even standing in a city.

Second, from what I have read regarding how to programmatically "Inspect" another character, it is asynchronous. Once you have a char you've decided to Inspect, you kick off a server request, then wait for an "information now available" event to trigger, before you can fetch what you want. From what I've read, it sounds like there isn't a way to tell what unitid the returned info is actually for, so if another add-on happened to ask about another unitid around the same time as your request, or you figure you'll kick off multiple requests on different unitids on the assumption that they will come back in the same order, you may not get the information you were expecting.

I am also wondering about how much load there could be on the game from kicking off all of these Inspect queries, whether you pick characters that the player happened to mouse-over, actually click on, or some other method. For someone standing in a starting area or solo questing, it shouldn't be too bad. Someone standing at the AH might bog down somewhat, but perhaps not enough to be annoying. People in a "serious raid / PvP", though, probably wouldn't like any periodic frame-rate drops or increased latency from the additional server traffic that would cause (the encounter itself already produces a lot of traffic, and all of the people running "boss mods" are shooting even more information to each other through the server). Of course, since I am not in that category, I am only going on supposition based on posts i've read on the WoW forums.


I'd better stop now, since I see I've produced YAWT (Yet Another Wall of Text?) Wink
Back to top
View user's profile Send private message
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Wed Jun 02, 2010 5:55 am    Post subject: Reply with quote

Rollie... that stat relevancy is two way street... some people will not like older data... but in light that many realms (semi-)regularly zero out, some people would surely appreciate having at least some info instead of none at all.

Kosh... I was not thinking of the impact of individual characters as they stay in the database for a long, long time. I was thinking of the effect of cnesuses run / characters seen / etc, on most wanted pages, the individual realm stats pages, etc. Storing (or transferring) a value of 0 takes less space and bandwith than for example the value of 2541, correct?

Interesting point about the red and yellow values, though your conclusion is in my opinion partially off. IF people prefer new data, the cutoff would not be 200 in 60, but 100 between 60-31 AND 100 between 30-0 days old data or even just 100+ in past 30 days. And if the people don't care about the exact age of data, 100 might still be enough. Especially, since doubling the number of censuses run does not double the accuracy of the data. Not even in the case of going from 1 census to 2, though there it can be close, if the second census has been run at the same time of day on the same day of the week (for example 2 concecutive Fridays at approximately 17:15 or even simultaneously, should a realm receive attention from two different submission makers).

The "joke" here is that there is no possibility of WCR receiving too much data with the current submission numbers. Also, unless my logic is badly faulty, censusing has somewhat diminishing returns. Running one census on two diffent factions can produce a lot more information than running two censuses one after another on a single faction.

I am creating YAWoT too, so taking a break. Very Happy
Back to top
View user's profile Send private message
Kosh



Joined: 01 Jul 2007
Posts: 84
Location: Somewhere on or near Earth
WR Updates: 1,210,548
Kosh WR Profile

PostPosted: Wed Jun 02, 2010 4:45 pm    Post subject: Reply with quote

1974ER wrote:
Kosh... I was not thinking of the impact of individual characters as they stay in the database for a long, long time. I was thinking of the effect of cnesuses run / characters seen / etc, on most wanted pages, the individual realm stats pages, etc. Storing (or transferring) a value of 0 takes less space and bandwith than for example the value of 2541, correct?
Well actually, a normal database design would store such aggregate numbers in an integer or other numeric field, so zero will take the same amount of room as any other number that can fit in the field. Now, you're right that sending a page with "0" on it will take slightly less time than the same page with "2541" on it, but it's only three fewer bytes, and the vast majority of the page will take about the same number of bytes to send. Take a look at the page source for one of those pages to get an idea, but the bulk of those pages don't really change much with different numbers (well, I suppose the guild list might; does it go away if there's zero submissions?).

1974ER wrote:
Interesting point about the red and yellow values, though your conclusion is in my opinion partially off. IF people prefer new data, the cutoff would not be 200 in 60, but 100 between 60-31 AND 100 between 30-0 days old data or even just 100+ in past 30 days. And if the people don't care about the exact age of data, 100 might still be enough.
Well, having 100 between 0-30 & 100 between 31-60 would be basically what we have now, but that would take extra coding on Rollie's part. Just changing the number of the displayed window to "60" would be more simple (just updating a number, unless there are gotchas to worry about in the code), so that is what I as assuming. Given that assumption, to me, the reliability of the data would tend to diminish if the total is the same while the allowed time span is increased. Look at it this way: if 100 samples in 30 days is what is needed to be considered "good", then 100 samples in 60 days (assuming a fairly even distribution), would then have half as many censuses done in any chosen span of time (say a day or a week), increasing the probability that characters will be missed.

If you use data in the graphs, etc. for characters seen within 60 days, but still display the "seen" numbers for 30 days, then you haven't solved your "zeroed realms" problem, even though there will still be (somewhat) valid information otherwise shown on those pages. If you display the "seen" numbers for 60 days, but internally choose the colors based on 30 days, then new code would need to be written, and the coloring might be confusing (e.g., showing 120, but red, since most of those 120 were submitted more than 30 days ago).

1974ER wrote:
Especially, since doubling the number of censuses run does not double the accuracy of the data. Not even in the case of going from 1 census to 2, though there it can be close, if the second census has been run at the same time of day on the same day of the week (for example 2 concecutive Fridays at approximately 17:15 or even simultaneously, should a realm receive attention from two different submission makers).

The "joke" here is that there is no possibility of WCR receiving too much data with the current submission numbers. Also, unless my logic is badly faulty, censusing has somewhat diminishing returns. Running one census on two diffent factions can produce a lot more information than running two censuses one after another on a single faction.
Now you are right about the number of submissions not having a linear relationship with "accuracy" of the data, but I don't feel like trying to dredge up math to calculate confidence levels for increasing samples of a population (and I'm sure you are all happy I'm not going to, as well Wink ).

I've already generated another WoT, so I'll just briefly mention that, while each new submission will tend to contribute less coverage of the population as a whole, you still need "enough" submissions in a span of time to have a good chance of showing reasonably accurate totals.
Back to top
View user's profile Send private message
1974ER
Epic Censi


Joined: 07 Nov 2008
Posts: 729

WR Updates: 23,375,480
1974ER WR Profile

PostPosted: Thu Jun 03, 2010 3:20 am    Post subject: Reply with quote

Part 1: Alright, so the storing and transmission of larger numbers will have neglible effect on storage space and bandwith required. Smile

Part 2: I agree with your diminishing concern, which is exactly why I brought up the 100 + 100 thing, which is NOT the way I want to head, by the way. I am just saying that going from 100/30 to 200/60 could still cause problems, if all or almost all entries are older than 30 days. And on the other hand 100/60 still MIGHT be enough to give to fair accuracy, if all/almost all entries are 30 or less days old.

The problem here is that the amount of censuses run on any specific one faction is not static. Whatever method we pick, some factions will still be less accurate, even with an identical number of censuses.

Part 3: Hehehe, I am not even sure it is possible, as the WoW engine in itself prevents 100% accuracy. Very Happy Also, the "accuracy" isn't linear, because time also plays a factor. Adding one census at 04:00 increases the overall accuracy less than adding one at 20:00. And adding one at 20:00 on a Monday will probably be slightly less accurate than doing it on a Saturday at the same time.

In terms of accuracy 1+1 does not equal 2 here, but rather 1,xx, where xx = from about 01 to about 99.

I need to leave for now, so I will check back later today...
Back to top
View user's profile Send private message
aberman
Guest







PostPosted: Thu Jun 17, 2010 7:41 am    Post subject: Reply with quote

You guys look like you are doing a great job! It's so cool that you are actually doing this. It looks difficult, but very practical. Let me know if you have any general questions. For instance, it seems like you are concerned about what to do with the data after 30-60 days.
Back to top
Display posts from previous:   
Post new topic   Reply to topic    WarcraftRealms.com Forum Index -> Census Talks All times are GMT - 6 Hours
Goto page Previous  1, 2
Page 2 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
WarcraftRealms.com  


Powered by phpBB © 2001, 2005 phpBB Group