Jump to content

Scorpienne

Members
  • Posts

    886
  • Joined

Reputation

10 Good

Personal Information

  • Location
    Norcross (Atlanta) Georgia
  • Homepage
    http://www.thethirteenthlegion.torportal.com
  • Interests
    roleplaying games (tabletop and Xbox), board games, card games. Family and job > games, though.
  • Occupation
    Marine Biologist - I clean up hazardous waste sites
  1. I've been called a lot of horrible things in my day, but as a practicing scientist that spends *every working day* trying to determine meaning from data, that is absolutely the *worst* and most hateful thing that has been leveled at me. Ever. I take it as one of the core principles of my very self that I am completely intellectually honest. I have, as a good scientist should, done the following: 1) Made my data public for alternative reanalysis for anyone who chooses to do so. 2) Explained my methods in detail, including their limitations. 3) Incorporated useful suggestions to improve the analysis. If you can do better, then I enthusiastically invite you to do so. If you can only accuse me of intellectual dishonesty, then you are what you are. Paige
  2. Although we've had a long crazy trip here, we're back to square one. The new population caps throw off all the math. I can't do this anymore until and unless I get a lot more data on what server status = how many people. If you're interested in helping, please take this survey HERE a few times a week, on the imperial and republic sides of your server, and at different server loads (light, medium, heavy, very heavy, full). The data, as always, will be housed HERE so that everyone can see it and do their own analysis if they choose. Paige
  3. Rarely have I been lambasted so thoroughly for doing math. I've published and linked my methods. I have always been willing to consider anyone else's method that seems better. (In fact, the current method is definitely a joint logical effort of several contributors.) So, no worries then men. Carry on. I shall not burden the APAC server issue with data further. You are quite passionate about the subject, and are not interested in sweet reason. Paige
  4. TL;DR = Dalborra and Gav Daragon are fine. Master Dar'Nala is lower population and should be merged with some other PvP server. This is how the pop of the three APAC servers compares to the average server of their types, and versus SWTOR overall. These numbers represent the portion of players left compared to 6/27/2012. So for instance SWTOR overall has about 20% less logins than a couple months ago. Master Dar'Nala is only about 58% of what it was a couple of months ago - it's lost about 42% of it's population. (Link to how I got these numbers HERE.) % of 6/27 population All SWTOR servers 80% All PvE 80% Dalborra 82% All RP PvE 99% Gav Daragon 92% All PvP 65% Master Dar'Nala 58% Dalborra is actually doing a tiny bit *better* than other PvE servers. Gav Daragon is doing a bit worse than other RP-PvE servers, because it's "only" at 92% of it's post transfer load, but that' means it's doing great! Master Dar'Nala has lost a lot of people... more than even other PvP servers. Paige
  5. Original post updated. The bright spot is that the pop decline this week was *less* than it was last week. :-) The population *balance* between the types of servers (PvP, PvE, etc.) and the areas/languages (APAC, US EST, German, etc.) are all about steady. Date Logins Change from prev week Change from 6/27/2012 6/27/2012 32,834 6/29/2012 34,310 4.5% 4.5% 7/2/2012 35,383 3.1% 7.8% 7/10/2012 35,160 -0.6% 7.1% 7/16/2012 35,853 2.0% 9.2% 7/23/2012 35,827 -0.1% 9.1% 7/30/2012 34,490 -3.7% 5.0% 8/5/2012 34,173 -0.9% 4.1% 8/13/2012 33,872 -0.9% 3.2% 8/21/2012 33,579 -0.9% 2.3% 8/27/2012 32,311 -3.8% -1.6% 9/4/2012 30,133 -6.7% -8.2% 9/10/2012 28,253 -6.2% -14.0% 9/17/2012 26,998 -4.4% -17.8% Overall population is about 82% of what it was on 6/27. RP-PvE is 102% of the 6/27 number (2% growth). PvP is at 68% of the 6/27 number (22% loss). The other server types are about in line with the overall population trend. Paige
  6. I'm sorry, Yogol, I don't understand your question... Here's the math... LINK Go to the tab called "summary stats". I automated the math in the first table so you can see exactly what the math is doing (it's not just numbers copied out of excel). Paige
  7. Yogol, please, I beg you, read every word of the original post on page 1. I think it explains everything. Once you read it, I'll be more than happy to answer any specific questions that you have. Paige
  8. I have no data about subs. :-) If you think that 5% of people are active online as a 24 hour 2 week average, then that's about 560,000 subs. If you think that 3% of people are active online as a 24 hour 2 week average, that's about 932,000 subs. If you think that 10% of people are active online as a 24 hour 2 week average, that's about 280,00 subs. Your guess is just as good as mine. Paige
  9. Tl;DR: Thanks! :-D Those sub numbers seem completely plausible (math at end). Heh, first of all, thanks. The last time someone said something about me personally, they were calling me a troll for posting numbers because they thought somehow I was trying to steer folks off the game (which couldn't be further from the truth). I'll try to stay classy. I do get exasperated sometimes. I'm a scientist by trade. I don't care what the the numbers *are*. I'm just trying to find the truest most representative numbers and figure out what they mean. I have no clue about sub numbers. None. Not even a little bit. "more then 500k, less then a million" sounds completely plausible as my own personal guesses go (GUESSES! not DATA, see math at the very end of the post). The range is pretty high. I will say that 10,000 and 10,000,000 do not seem plausible. For those of you just joining us in our at-home audience, here's what I mean (full explanation HERE). Logins are much less total number of players Keep in mind the numbers I'm giving you are the AVERAGE number of people LOGGED IN - it's probable that not every player is logged in all the time. For instance, if I say that there are 600 people, on average, logged into server X, in real life that might mean: 400 Stay at home parents login at noon and logout at 4 pm. 600 School kids login at 4 pm and logout at 8 pm. 800 Working people login at 8 pm and logout at 12 midnight. 800 College students login at 12 midnight and logout at 4 am. 600 Insomniacs login at 4 am and logout at 8 am. 400 3rd shift workers login at 8 am and logout at 12 noon. So that's 600 AVERAGE LOGINS but it's 3,600 SUBSCRIBERS in this particular example. I know of no hard, evidence-based numbers that indicate that X logins = Y subscribers. Some folks have discussed values from 1% to about 25% based on other games, their friends, numbers they pulled out of their ears, and things they read on the internet. I have only my personal anecdotal, non-evidence-based, subjective experience that lead me to think the % logins is on the low end of that. That's not based on anything like math though. Right now, the best math I can do (HERE) indicates that as a 24hr, 2 week average, about 28,000 people are logged in. If you think that's 5% of the population of subs, that's 560,000, which is perfectly in-line with the published numbers. If you think it's more like 3% (because fewer people are online because of GW2 and midterms BUT they are still subscribed) then that's 932,400 subs. I don't have any clue. That's why I think that 500,000 to 1,000,000 subs is completely plausible. If you think that the % of subs actively logged in is higher or lower than 3-5%, then your numbers will be accordingly different. Paige
  10. TL;DR - Completely correct. I took the server status average (from TORSTATUS, who got it from the SWTOR site) and correlated that with actual census measurements of server logins to correlate status X with Y number of logins. We're on the same mathematical path here. :-) I completely agree that the server status page *ONLY* gives you, well, server status. Not number of logins. Therefore, that's why TORSTATUS only reports "phi" values of 1 through 5 where 1 equals light and 5 equals full. If you'll read the methods HERE you'll see we went in and did a bunch of server censuses, we actually counted the # of people on the server. Then I did took all of the light measurements, found the average # of logins. Same for the other measurement levels. It appears that clicking the link is just not something people are going to do. So here, I'll just repost the methods... If you see an error, or a better way to do this, or a different model, or a way to optimize the data collection or reporting, then let's talk about it. I don't give a hoot about *how* it's done - I'm just looking for a way to get the most representative and realistic numbers. This isn't the best way, or the right way, or the only way... but it's the best way we've come up with so far. I say "we" because if you'll look at the original thread, there's a number of people who have come up with various ways to alter the process to make it better. 2. TORSTATUS AVERAGES LAST 2 WEEKS OF SERVER STATUS FROM SWTOR Torstatus (http://www.torstatus.net) is a program that checks the server load status page (http://www.swtor.com/server-status) every 5 minutes 24 hours a day 7 days a week, since the game started, and records the results in a database. The DB then records Light as 1, Standard as 2, Heavy as 3, Very Heavy as 4, and Full as 5. Every hour it calculates the average of all of those recordings for the last 14 days and reports that number as "ø" for each server on the Torstatus webpage. So when you see a phi value on Torstatus, that means that is an average of it's server status for that server taken every 5 minutes for 2 weeks. 3. SERVER STATUS CAN BE CORRELATED TO # OF LOGINS Then a lot of generous folks from the community went to our destination servers and counted the number of logged in characters and compared that to server status (SURVEY HERE, DATA HERE). As of 07/02/2012, we have 94 measurements among the destination servers. So far, our data says that: The average LIGHT destination server has 649 concurrent logins (range is 224-1056) The average STANDARD destination server has 1,419 concurrent logins (range is 660-2756) The average HEAVY destination server has 2,198 concurrent logins (range is 930-3604) The average VERY HEAVY destination server has 2735 concurrent logins (range is 1526 - 3394) The average FULL destination server has 3,638 concurrent logins (range is 3,166-3810) Clearly the data has some variability in it, as the range of server loads to number of concurrent logins can be pretty wide. If you look at his data, you can see that sometimes a heavy destination server has under 1,000 people on it and sometimes it has over 3.000 people on it. Same kind of range (or more) for the other server statuses. Part of this is because people log on and log off while we are /who-ing the population, and part of this is because different servers might have different population levels and/or Bioware may alter those population levels. (So standard might be 1,500 on one server and 1,000 on a different server, or 1,500 today and 1,000 a month ago - we don't know.) 4. USING TORSTATUS AND CORRELATION OF SERVER STATUS-TO-LOGINS TO ESTIMATE LOGINS ON ALL DESTINATION SERVERS We can use the Torstatus phil values and our estimates of the correlation of server load status to concurrent logins to estimate the number of concurrent logins in all of the destination servers. If Φ is exactly equal to a whole number, then use the average value for that status level. If Φ is between two whole numbers multiply the decimal times the higher status average population multiply 1 - the decimal times the lower status average population Φ = 2.8 20% at status 2 (0.20 x 1,419 = 284) 80% at status 3 (0.80 x 2,198 = 1,758) Estimated concurrent logins = 284 + 1,758 = 2042 people logged in Φ = 1.3 70% at status 1 (0.70 x 649 = 454) 30% at status 2 (0.3 x 1,419 = 426) Estimated concurrent logins = 454 + 426 = 880 people logged in If the decimal of Φ is equal to 0.50 then the server spends 50% of its time at each status. Math is here - you're welcome to look at it yourself and see if you can find a better way to analyze this data. You'll also see tabs for previous analyses as well as summary tables for total and average populations by server type and region and a tab with transfer info. https://docs.google.com/spreadsheet/ccc?key=0Aregkvys5QFodDlvU05lQ3o2S3ZrYU9SYjRwM29OSnc#gid=12 That means if you take the numbers from Torstatus representing the average population over the last two weeks for the destination servers you end up with the list above. 5. CAVEATS - LOOKING AT 24 HR AVERAGE, LOGINS DOESN'T PREDICT SUBSCRIBERS, AND EXCLUDING ORIGIN SERVERS 5A 24 Hour average doesn't equal prime-time logins. The catch here is realizing that this estimate is based on a 14 day average status, so that's dead slack times and peak times. The theoretical average population UNDERESTIMATES the peak population and OVERESTIMATES the off-peak population. You might login at prime time with a queue and realize your server has 3x? 4x? 5x? as many players as the list above indicates. That's to be expected. 5B Logins are much less total number of players Keep in mind this is the average number of people logged in - it's probable that not every player is logged in all the time. For instance, if I say that there are 600 people, on average, logged into server X, in real life that might mean: 400 Stay at home parents login at noon and logout at 4. 600 School kids login at 4 and logout at 8. 800 Working people login at 8 and logout at 12 midnight. 800 College students login at 12 midnight and logout at 4. 600 Insomniacs login at 4 and logout at 8. 400 3rd shift workers login at 8 and logout at 12 noon. So that's 600 AVERAGE LOGINS but it's 3,600 SUBSCRIBERS in this particular example. 5C This analysis excludes 191 origin servers. All of the 191 origin servers have a phi value on Torstatus of below 1.0. I do not know how to assess the population on these servers AND there could be a *lot* of people on these servers that aren't counted as part of this analysis. If there were 500 people on each of these servers, that might be another 95,000 logins... or it could be nobody. I just don't know.
  11. Heh! Actually it visits every five minutes, but yeah, that's pretty much it... Paige
  12. 100%. Sure, the game has lost some population, but I think the actual numbers show that it's doing a heck of a lot better than some folks on the forums seem to think it is. Paige
  13. I belive people's eyes! I really appreciate you volunteering to collect data. Let's get actual data and compare real numbers? That's the best way to do this. :-D Thank you! HERE is a survey form. Once you fill it out, the data are posted HERE for everyone to look at and draw their own opinions. TORSTATUS is run by a nice European fella named Thomas. I've chatted with him via email a few times. You can grab Thomas' email and more info about him and what TORSTATUS does and how on that site. http://www.torstatus.net/about It looks to me like the site does what he says it does in the way he explains that it does it. If you don't believe TORSTATUS, please let me know what things about it you think are off. I'm more than happy to help you check any math. Paige
  14. Hm. I must not be explaining this well. Or maybe I don't understand you. Is the basic logical gap here is that you believe that the average number of players on a server over the last two weeks is a meaningless number? If so, cool. :-) Don't use that number for anything. Or is the basic logical gap that you believe my estimate of the average number of players on a server over the last two weeks is incorrect? If so, cool. :-) Show me a better way to get at that number? My original methods are discussed (at length) here. If you don't/can't/won't read that wall of text, that's cool too. Just let me know what your method is and I'll give it a try. Paige
  15. Didja read the methods? LINK THis whole analysis is based on 90+ census measurements (/who by level band) of the destination servers. Check out point 3 of that post. Paige
×
×
  • Create New...