Scorpienne Posted June 19, 2012 Author Share Posted June 19, 2012 Jeff, check my excel coding, willya? B3 = cell with phi value LowPhi = whatever horrible number we'll call the servers with phi less than or equal to 1. I'm thinkin' 250 but I'm willing to listen to other ideas. Light = 500 Standard = 1500 Heavy = 2375 Very Heavy = 3000 XXX = error code for an equation that just malfunctioned =IF(B3<=1,LowPhi,IF(B3<2,((2-B3)*Light+(B3-1)*Standard),IF(B3<3,((3-B3)*Standard+(B3-2)*Heavy),IF(B3<4,((4-B3)*Heavy+(B3-3)*VeryHeavy),"XXX")))) Link to comment Share on other sites More sharing options...
Scorpienne Posted June 19, 2012 Author Share Posted June 19, 2012 (edited) Okay, check the spreadsheet for the new tab called "JeffTest". That's the numbers from the 18th, done Jeffstyle, compared to the numbers done the old way, and a column with the difference. https://docs.google.com/spreadsheet/ccc?key=0Aregkvys5QFodFJ2OWN5U0hwaVFBYWdqUUh1WmdZUFE#gid=32 The green cells are showing the servers with a phi between 1 and 2 that all had 250 people added since Jeffstyle doesn't need to accumulate people at lower phi status because they're accounted for in the higher phi status. It actually shows a *greater* number of people than my original estimates for 6/18 because of this. The Jeffstyle total concurrent logins number is 75,866 vs 66,966 for the other way. About a 12% difference. Hmm. When I look at *only* the servers with a phi greater than one (so I can continue to avoid the low phi server problem), I get 30,116 logins Jeffstyle compared to 21,908 logins Paigestyle. That's about a 27% difference. Hmmm.... super interesting. Also, look at this... Phi | Paige | Jeff Phi >1 | 30,116 | 21,908 Phi ≤ 1 | 45,750 | 45,088 That Phi ≤ 1 is for 183 servers. So instead of 45k people, that could be 0 to ~91,500 people. I have just found out that my uncertainty is greater than the number of things I can count for certain. I think I'm going to take a walk. Paige Paige Edited June 19, 2012 by Scorpienne Link to comment Share on other sites More sharing options...
Leeoben Posted June 19, 2012 Share Posted June 19, 2012 your spreadsheet formula is correct and works, however by using the top ends of each status you are calculating the maximum average number of players that were on, instead of using the median for each bracket to determine the average average (if that even makes sense) which is why you are getting numbers so much higher than what you had previously. sorry for the delayed reply, and poorer grammar, my firewall at work only lets me on this site about 50% of the time so i have to post quickly when i can get here. Link to comment Share on other sites More sharing options...
Scorpienne Posted June 19, 2012 Author Share Posted June 19, 2012 Yah, yah, don't worry about brevity or grammar too much. It's all good. :-) Okay, updated the spreadsheet using the category midpoints. (I really want to use Crystal Ball to start doing probability distributions, but I do not have that sort of time.) Jeff | Paige Phi >1 | 18,356 | 21,908 Phi ≤ 1 | 45,750 | 45,088 Hm... okay so my numbers are 17% above yours for the phi > 1 servers using the category midpoints. Hunh. Interesting. Gotta think about how to write that up and how it compares to the other method for a while. Also, I got a request via email asking to seperate out the phi > 1 servers from the phi ≤ 1 servers so people can make their own estimation of the uncertainty. Going to think about that too. I think that we'll see a continued drain of the origin servers and a continued swell of the destination servers, and soon the difference between origin and destion will = the difference between phi > 1 and phi ≤ 1 and we can look at them that way. Paige Link to comment Share on other sites More sharing options...
Scorpienne Posted June 20, 2012 Author Share Posted June 20, 2012 Okies... spreadsheet updated with today's numbers. https://docs.google.com/spreadsheet/ccc?key=0Aregkvys5QFodFJ2OWN5U0hwaVFBYWdqUUh1WmdZUFE#gid=33 Please note that there's a sheet for today's date AND one with today's date plus "Jeffstyle" in the title. The latter is the alternate method discussed above, and so isn't directly comparable with previous estimations. The more I think about the more I like the "Jeffstyle". Once we get to *only* post transfer data, I might go with that calc method only, but I'm still kicking that around. So todays estimate of total concurrent logins is 65-69k, depending on the analysis method. The kicker is that between 68-72% of that (about 47k logins) is on low phi servers. That means that I really don't know if that's 0 people, 47k people, or 90k people; that part of the analysis is really uncertain. I really only have a good analysis that there's 18k-22k concurrent logins on servers that have a phi of 1 or greater. Hopefully, as more of the population condenses onto the destination servers, we'll have better calcs. What's crazy is that the population on The Fatman has markedly decreased! I don't know what that means for sure, but speculation is that people are not playing the alts that they'd had there. Expectedly, The Bastion, Prophecy of the Five, Tomb of Fredon Nadd, The Ebon Hawk, Drooga's Pleasure Barge, Canderous Ordo, The Shadowlands, The Harbinger, The Red Eclipse, Jar'Kai Sword, Mantle of the Force, Darth Nihilus, The Progenitor, Begeren Colony, and Jung Ma have posted huge gains. Corellian Run, The Jedi Tower, Vanjervalis Chain, T3-M4, Battle Meditation, and Nightmare Lands have posted modest gains. It's the 2 week average thing that's diluting out the recent transfers to these servers. Paige Link to comment Share on other sites More sharing options...
Leeoben Posted June 20, 2012 Share Posted June 20, 2012 I'm glad you like my method and are considering using it! Even if you don't end up using it I'm glad I was at least able to help bring a different perspective to the numbers. As for the issue with the low phi servers, sadly this isn't something that will be going away anytime soon, unless bioware closes some of the servers. If it was me this is how I would handle their population going forward... Since you have a data set from the day before transfers started (granted that has all low phi servers around 250, but its about the best you can do) you can take that total pop and use it as a baseline. Once two weeks have past ( to allow enough time for all servers to register their status post transfer) run the numbers for the destination servers only. Then take that number subtract it from the baseline and divide the remainder among the origin servers, and use that number as the new "low phi" number. The one main flaw with this is there will probably be a spike in population with the transfers, and it's likely the baseline will be a lower total population then what is currently playing. Even a rise of only about 5,000 players could throw off the low phi number by almost 50 per server. But even stating that, I think having an error window that is likely to be 5-10k is a lot better than what we have currently where it could be almost 50k! Link to comment Share on other sites More sharing options...
Evironrage Posted June 21, 2012 Share Posted June 21, 2012 (edited) Funny how many do not want to see the writing on the wall. There is nothing wrong with a game having ~100k active members. These numbers tell the tale. Close ~85-90% of the servers now and accept this game is not going to garner more than that. SWTOR has a place in the market, but it is not the new king, as is evidenced by the lack of active players. If BW can keep the lights on with these numbers and add new content periodically this game will have a decent life-span. If it is not advantageous to do so this game has another 2yrs at most. My bet is on the latter, but hope remains. Edited June 21, 2012 by Evironrage Link to comment Share on other sites More sharing options...
Kourage Posted June 21, 2012 Share Posted June 21, 2012 What's crazy is that the population on The Fatman has markedly decreased! I don't know what that means for sure, but speculation is that people are not playing the alts that they'd had there. I don't know about that. At 4:30pm I logged onto The Fatman today and I saw 244 on Fleet. I went to see if I could switch to Fleet 2 and I was already on Fleet 2. That's a lot of people, especially for before 5pm! But then I wen to Quesh and there were 10 people. Nobody likes Quesh! On another note, can't somebody log in and manually count the pops each night? At least on a couple of servers just to see if this other method is even close to being correct? Link to comment Share on other sites More sharing options...
Leeoben Posted June 21, 2012 Share Posted June 21, 2012 @Evironrage I don't think the point of this thread is to try and sway peoples opinions that the game is performing great, performing as expected, or under performing. The original thread was made to try and give people an idea of which servers were the most popular before transfers were allowed. It turned into a fun experiment for Paige (the thread starter) to track the various fluctuations in population, and now with transfers she is continuing the tracking out of curiosity. People can take the data and interpret it anyway they like, but I am almost certain it is not Paige's intention to try and sway someones opinion on how the game is performing. I don't mean to speak for Paige, but I have been following these threads for a while and I think their intentions are pretty obvious. @Kourage My formula is entirely dependent on the current numbers for status brackets being accurate. If the numbers for standard, heavy, very heavy, etc are accurate my formula is very solid (as demonstrated by the generic variable equations in my post). If those numbers are wrong or have changed then both my numbers and Paige's original numbers are totally wrong. if you would like to offer your services as a realm census taker I believe Paige has a link to a survey where she is collecting data to try and come up with current numbers. Link to comment Share on other sites More sharing options...
jgelling Posted June 21, 2012 Share Posted June 21, 2012 (edited) Aren't the light servers pretty impossible to estimate now? They could be totally dead or they could be just a shade under Standard, but likely the origin servers are going to trend towards dead over time. The problem with light servers seems like it would tend to overestimate the concurrent users by being very generous on the logins from what could be, now, 100 ghost town servers. Still, the data shows that summer and D3 were not good for the game, the uptick is nice but you might have expected more given the mass emails that were sent out to people to encourage them to login and transfer. It'll be interesting to see though if there's any change when 1.3 hits. And has there been any sign of an increase in server pops on the destination servers, ie is Heavy still being reported as the same number as it was before? If it is, that's a pretty troubling sign, because we just merged 90% into the remaining servers and one would imagine there should be an increase in server capacity for that, unless the game is in real trouble? Edited June 21, 2012 by jgelling Link to comment Share on other sites More sharing options...
Scorpienne Posted June 21, 2012 Author Share Posted June 21, 2012 (edited) On another note, can't somebody log in and manually count the pops each night? At least on a couple of servers just to see if this other method is even close to being correct? Bam! Survey Data thus far Graph of analysis of level breakdown and faction balance Graph of analysis of level 50 characters While you guys are nice and everything, I don't have time to survey your servers. I barely have time enough to census my own. :-) If you're volunteering to census one of the destination servers on a regular basis, then that's great! Please feel free to put the data in that form so it can be publicly warehoused. Or, if you've got mad skillz and can program/automate the process of a character logging in and /whoing the population so that real live human beings do not have to give up game time to do it, then that would also be fabulous. @Evironrage I don't think the point of this thread is to try and sway peoples opinions that the game is performing great, performing as expected, or under performing. You have it exactly, sir. I'm not taking a stance on game peformance. I'm just reporting the facts. Everyone will have to analyze them for himself. I am tired of people pulling numbers out of their ears - hopefully we can at least agree on the data, and then have a meaningful discussion about what it signifies. That's how science works. We see X - what does it mean and why did it happen? We all agree that we see X. The fruitful debate is what it means and what caused it. @Kourage If the numbers for standard, heavy, very heavy, etc are accurate my formula is very solid... [snip] If those numbers are wrong or have changed then both my numbers and Paige's original numbers are totally wrong Again, exactly correct. The reason I'm so explicit about saying where my numbers came from is so you can reject this analysis and perform your own with whatever numbers you think are appropriate. Doing enough server censuses to be able to say that a light pop is X people and that a standard pop is Y people is more than I can do. If there are community volunteers to actually do the work, I'm more than happy to store and analyze the data. Paige Edited June 21, 2012 by Scorpienne Link to comment Share on other sites More sharing options...
Scorpienne Posted June 21, 2012 Author Share Posted June 21, 2012 Aren't the light servers pretty impossible to estimate now? They could be totally dead or they could be just a shade under Standard, but likely the origin servers are going to trend towards dead over time. They're certainly very difficult to estimate. This is further complicated by the fact that the torstatus number is a 2 week average SO there's some pre-transfer not-a-ghost-town data points in there. It'll be interesting to see though if there's any change when 1.3 hits. And has there been any sign of an increase in server pops on the destination servers, ie is Heavy still being reported as the same number as it was before? If it is, that's a pretty troubling sign, because we just merged 90% into the remaining servers and one would imagine there should be an increase in server capacity for that, unless the game is in real trouble? I don't know. As I stated, not enough data to link some proverbial new pop level to a server status. Either someone has to put in the work like Mythbusting did originally, or we'll have to wait til my very slow process of server censuses accumulates enough data to be meaningful. paige Link to comment Share on other sites More sharing options...
ZudetGambeous Posted June 21, 2012 Share Posted June 21, 2012 I've been thinking on how to avoid the double counting problem since the transfers.(Almost) Everyone who was previously on a phi <1 server has now transferred to one of the higher population servers and thus they are now being counted twice, once because the light average is still set at 250, and then again by their actual contribution to the higher pop server. After some consideration, I think the best solution is to only tally up the concurrent users for the 23(?) destination servers and any servers that had no transfers. I think it is a more accurate estimate to take the origin servers as having an average population of 0 then it is to have them set at 250. Or at the very least log into 1 of the origin servers a couple times and get an actual average for it and use that for the calculation. Link to comment Share on other sites More sharing options...
jgelling Posted June 21, 2012 Share Posted June 21, 2012 They're certainly very difficult to estimate. This is further complicated by the fact that the torstatus number is a 2 week average SO there's some pre-transfer not-a-ghost-town data points in there. I don't know. As I stated, not enough data to link some proverbial new pop level to a server status. Either someone has to put in the work like Mythbusting did originally, or we'll have to wait til my very slow process of server censuses accumulates enough data to be meaningful. paige Yep, good data doesn't just appear out of nowhere! Great work - it's a fun little thought exercise if nothing else, to get a glimpse of what might be happening behind the curtain. Link to comment Share on other sites More sharing options...
SnakeCL Posted June 21, 2012 Share Posted June 21, 2012 Bam! Survey Interesting thing happened while i was filling out my first survey (Einahr, Jedi Covenant). Server went from Standard to Heavy as I was completing the survey. Not sure if that's any help, but it might assist in ballparking the lower range of "Heavy" for server status. I checked before I started and the server was at "standard", checked again after the form was completed but before it was sent, and server was "heavy" Link to comment Share on other sites More sharing options...
Emeda Posted June 22, 2012 Share Posted June 22, 2012 I don't know about that. At 4:30pm I logged onto The Fatman today and I saw 244 on Fleet. I went to see if I could switch to Fleet 2 and I was already on Fleet 2. That's a lot of people, especially for before 5pm! But then I wen to Quesh and there were 10 people. Nobody likes Quesh! On another note, can't somebody log in and manually count the pops each night? At least on a couple of servers just to see if this other method is even close to being correct? Are you volenteering? I think they would love for some people to help out. The more people helping the better and easier for all the people doing it and the more data the betterchance of it being acurate I know they would love the help Link to comment Share on other sites More sharing options...
Emeda Posted June 22, 2012 Share Posted June 22, 2012 I'm glad you like my method and are considering using it! Even if you don't end up using it I'm glad I was at least able to help bring a different perspective to the numbers. As for the issue with the low phi servers, sadly this isn't something that will be going away anytime soon, unless bioware closes some of the servers. If it was me this is how I would handle their population going forward... Since you have a data set from the day before transfers started (granted that has all low phi servers around 250, but its about the best you can do) you can take that total pop and use it as a baseline. Once two weeks have past ( to allow enough time for all servers to register their status post transfer) run the numbers for the destination servers only. Then take that number subtract it from the baseline and divide the remainder among the origin servers, and use that number as the new "low phi" number. The one main flaw with this is there will probably be a spike in population with the transfers, and it's likely the baseline will be a lower total population then what is currently playing. Even a rise of only about 5,000 players could throw off the low phi number by almost 50 per server. But even stating that, I think having an error window that is likely to be 5-10k is a lot better than what we have currently where it could be almost 50k! Nice to see a diffent look at it. Question for paige. YOu mentioned you were able to get data from TORSTATUS people, could you ask them how they figure it? Might be worth a try. Also it it possible to get like dec data and other data close to calls to see how numbers compair to mentioned sub numbers. Link to comment Share on other sites More sharing options...
Leeoben Posted June 22, 2012 Share Posted June 22, 2012 I know this isn't going to get us the complete answer, but I am logging into my origin server (Bondar Crystal) at random times to do population surveys. So far I have done 2 today, first one I got 8 total people 5 emp, 3 rep (I made a new Empire and Republic toon) and second time was 9 people 4 emp, 5 rep. It literally takes 30 seconds to log onto both toons, do a "/who" and get the numbers. If others would be willing to do this on their origin servers it could really help us figure out what to do with the low phi servers moving forward. I figure ill continue to log on at random times through out the week and see if it stays consistent. I am intentionally not trying to hit only peak hours as we are looking for an average value, and if all we get are peak values it could skew the numbers to the high side. Link to comment Share on other sites More sharing options...
SnakeCL Posted June 22, 2012 Share Posted June 22, 2012 After some consideration, I think the best solution is to only tally up the concurrent users for the 23(?) destination servers and any servers that had no transfers. I think it is a more accurate estimate to take the origin servers as having an average population of 0 then it is to have them set at 250. Or at the very least log into 1 of the origin servers a couple times and get an actual average for it and use that for the calculation. I plan on doing that for my origin (Keller's Void), but so far, its looking like an average of maybe 2-3 so you're probably right. Its much closer to 0 than 250. Plus, 0 is the low-end. if we assume all the origin servers are zero, then we have a absolute minimum number, and it can only be higher from there. Link to comment Share on other sites More sharing options...
Scorpienne Posted June 22, 2012 Author Share Posted June 22, 2012 Question for paige. YOu mentioned you were able to get data from TORSTATUS people, could you ask them how they figure it? Might be worth a try. Ah, they explain part of this on their site... [LINK] Q: What does the ø number stand for? A: The chart list on the homepage shows average population index for the last 14 days. This population index is calculated from the official server statuses during a day... where 1 is light, 2 is standard and so on. It's updated once every hour. The program goes to this page: http://www.swtor.com/server-status every 5 minutes, captures the info, and puts it into a database. The raw data looks like this... (times are CEST, not EST) name;"datetime";"population" The Ebon Hawk;"2012-06-20 02:25:01";"heavy" The Ebon Hawk;"2012-06-20 02:30:01";"heavy" The Ebon Hawk;"2012-06-20 02:35:01";"heavy" The Ebon Hawk;"2012-06-20 02:40:01";"heavy" The Ebon Hawk;"2012-06-20 02:45:01";"heavy" The Ebon Hawk;"2012-06-20 02:50:01";"heavy" The Ebon Hawk;"2012-06-20 02:55:01";"heavy" The Ebon Hawk;"2012-06-20 03:00:01";"very heavy" The Ebon Hawk;"2012-06-20 03:05:02";"very heavy" The Ebon Hawk;"2012-06-20 03:10:01";"very heavy" The Ebon Hawk;"2012-06-20 03:15:01";"very heavy" The Ebon Hawk;"2012-06-20 03:20:01";"very heavy" The Ebon Hawk;"2012-06-20 03:25:01";"very heavy" The Ebon Hawk;"2012-06-20 03:30:01";"very heavy" The Ebon Hawk;"2012-06-20 03:35:01";"very heavy" The database substitutes: 1 for light 2 for standard 3 for heavy 4 for very heavy and 5 for full Then the database takes the average of the 4000-ish measurements from the last two weeks and updates the number to the web page every hour. And it does it for every server worldwide. Also it it possible to get like dec data and other data close to calls to see how numbers compair to mentioned sub numbers. Email the torstatus guys - link on their site. You'll have to work it out with them. I'm not sure that # of logins are predictive of the number of subs. The numbers don't really correlate. Paige Link to comment Share on other sites More sharing options...
SnakeCL Posted June 22, 2012 Share Posted June 22, 2012 I think we might have to poll the destination servers individually, since its looking like they have unique pop amounts set to their server status. (we need more volunteers! LOL) For instance, last night, (and feel free to correct me on this, if its actually applicable) Jedi Covenant had 1748 Republic characters logged-in. (if we assume pop is 1:1, then that's 3496 users for that server at that time). Server was only "Heavy" at the time of that reading though, but that's almost 500 users above the accepted value for "Very heavy ---> Full". The rest of the census forms I did seem to corroborate a higher pop cap for Jedi Covenant, or am I imagining things? Link to comment Share on other sites More sharing options...
sumuji Posted June 22, 2012 Share Posted June 22, 2012 Nice work but.... why? I don't get why anyone but Bioware or their competition would care. Everytime I see a thread like this I ask myself the same thing. It doesn't even feed curiosity because it's subject to so many different things that a player going /who wouldn't have access to try to come up with a active sub number. Why they're all hung up on numbers that shouldn't matter if you're on a healthy server is another thing to wonder. Unless you stand to make money off of the game.............It's just odd. Link to comment Share on other sites More sharing options...
Leeoben Posted June 22, 2012 Share Posted June 22, 2012 Everytime I see a thread like this I ask myself the same thing. It doesn't even feed curiosity because it's subject to so many different things that a player going /who wouldn't have access to try to come up with a active sub number. Why they're all hung up on numbers that shouldn't matter if you're on a healthy server is another thing to wonder. Unless you stand to make money off of the game.............It's just odd. I don't know how many times this has been said, but I will say it again... THE POINT OF THIS THREAD IS NOT TO DETERMINE THE NUMBER OF SUBS! The original thread was made back when there were serious population issues and people were looking for options on where they could reroll. So Paige came up with a system to try and determine the average number of users on at any given time. At no point was she trying to determine the actual number of subs the game had. Obviously people always want to try and find a link between the two, and im sure with enough raw data it would be possible, but as you stated there are just so many unknowns that this becomes almost impossible. Now that the transfers have happened its more of a continuing what was started thing. It's interesting to some people, myself included, to others it's not. if you don't find it interesting and don't have anything to add to the discussion you are free to ignore the thread. Link to comment Share on other sites More sharing options...
Scorpienne Posted June 22, 2012 Author Share Posted June 22, 2012 I think we might have to poll the destination servers individually, since its looking like they have unique pop amounts set to their server status. (we need more volunteers! LOL) For instance, last night, (and feel free to correct me on this, if its actually applicable) Jedi Covenant had 1748 Republic characters logged-in. (if we assume pop is 1:1, then that's 3496 users for that server at that time). Server was only "Heavy" at the time of that reading though, but that's almost 500 users above the accepted value for "Very heavy ---> Full". The rest of the census forms I did seem to corroborate a higher pop cap for Jedi Covenant, or am I imagining things? You are completely correct, sir! BW flat-out said they were altering the population caps and I have no clue what that's going to do to pop numbers. It makes this whole analysis much more uncertain. Between the low-phi servers and altered pop caps, the analysis has gotten so uncertain, I'm not sure it's worthwhile anymore. What do you guys think? Paige Link to comment Share on other sites More sharing options...
SnakeCL Posted June 22, 2012 Share Posted June 22, 2012 (edited) You are completely correct, sir! BW flat-out said they were altering the population caps and I have no clue what that's going to do to pop numbers. It makes this whole analysis much more uncertain. Between the low-phi servers and altered pop caps, the analysis has gotten so uncertain, I'm not sure it's worthwhile anymore. What do you guys think? Paige Check the data page for the input form. Seems like a LOT of people are adding pops for the big servers. With enough help, we might be able to nail down the new "tipping points". All listed as light. Fatman 896 Jedi Covenant 807 Canderous Ordo 780 Shadowlands 695 Just noticed Jedi Covenant hit "Standard" so I dropped in and polled it. Tipping point between Light and Standard for Jedi Covenant is between 807 and 1056 apparently. Edited June 22, 2012 by SnakeCL addition Link to comment Share on other sites More sharing options...
Recommended Posts