questions about accuracy of data

[brokenbottle]

Can someone tell me how this mod works or rather what it does when a player runs it? There's a debate on the official boards about how accurate the census data is and I'd just like to get a clearer picture of how this thing works so I can make an educated decision myself. I read the faq and see that the more people that supply census data for a given server, the more accurate the census is. Why is that? does it just take a snapshot poll of the characters that are playing at that moment?

[Rollie]

Do you mind posting a link to this discussion?

Basically the way the mod works is like this. The mod crawls /who data, starting with 70 and working down to 1, taking note of every character seen. One scan is called a snapshot. A singular snapshot in itself doesn't really tell you much. Time of day, day of week and other factors can effect how many people are online at a given time.

Now, take more snapshots, and or aggregate with other people's snapshots and you start to get a good feeling for the population demographics for that server.

A single person who plays quite frequently and allows the mod to take automated, regular snapshots, is quite effective in covering a single faction/server all by themselves. Obviously though, the more snapshots taken, the better.

Now, there is also something else worth mentioning. On one hand, it tracks every character ever seen, and aggregates that data together. This doesn't equate into the number of actively playing character, just the total number of characters seen. However, it also saves data on a per snapshot basis, which can give you an indicator of the actual number of players online at that time. Here at WR, you can find that kind of information on the Activity Graphs and it is broken down in several ways.

At any rate, there you have it! As long as a server/faction is getting a healthy number of snapshots (and by healthy, I mean 100+ snapshots per month) then the data provided for that server/faction is going to be fairly accurate.

The stats cannot ever be 100% accurate, but they don't have to be either. Even with limited submitted data, you can easily get to within 95% accuracy without much trouble.

[DM.]

When the mod is running it will do a series of /who commands and whatever it gets back is stored into its .lua file. If the /who results in 50 or more results then it goes deeper and searches by a smaller range of levels, race, class, and even characters if it gets to that. Its records the name of the character, race, class, guild, level, and current date.

Depending on the size of your realm and the amount of players currently online the scan can take anywhere from 3 mins to about 20mins on a max queued realm.

The mod doesn't run 24/7 whenever you are online. You can set it for specific intervals. The default is every 30mins. But this can be changed to every 1min if you want.

Then you have to upload your .lua file here: http://warcraftrealms.com/upload.php
And it will get processed in a few hours after its uploaded, and then stats you see will be updated once or twice per day.

There is no live feed from the realms to the data here. All the stats you see are collected by players in game and uploaded by players.

Of course someone can't be online 24/7 due to sleep, and other issues so they can't capture all the players online that they can't see so thats where the inaccurate stuff is. Also players that come online for a few mins and go offline may not be seen at all.


What's the link to the thread btw? Maybe I can post a bit there


*edit*.... beaten by the Troll....

[brokenbottle]

Thanks for the information guys. I guess the answer is "yes, pretty accurate if we get enough snapshots."

There's not an active thread debating the accuracy of WR per say (I see another thread here referring to Drysc's comments though), it's more that people will cite WR data as a proof that the population of this or that server is declining and the response is usually sharply divided as to whether or not WR is a credible source. The infamous Drysc thread is usually cited as proof that WR is not accurate.

Given the way WR gathers data, it would appear to me that it's a pretty good gauge for established servers, not so great for new servers if the WR participation is low, and maybe not so great for game wide data beyond very broad trends because the quality of the data can vary from server to server. Would this be fair to say?

[DM.]

Have you seen this thread?
http://warcraftrealms.com/forum/viewtopic.php?t=3078

More specifically this chart... http://www.warcraftrealms.com/temp/activity.htm


Warcraft Realms will never be 100% accurate. And yes the answer is 'if we get enough snapshots'. Blizzard will never release any data about their realms data. They didn't even mention that they raised the cap to 4000 players online from a previous 3000 and thats why there aren't any queues during peak times and its not because that players are quitting.

I think that new realms get alot of attention from those hard core Census takers with alts on a dozen realms.

[brokenbottle]

Actually, while they haven't said what the actual cap numbers are and were they have said that upping the caps has contributed to the lack of queues. Of course this elicits replies of "liars!" and "no, it's because the game is dying" on the official forums

What's the definition of "player activity" as used on that chart? Is it just a character logging in?

[DM.]

Player activity on any stats you see here means a character lvl 10 and higher which has been seen at least once in the last 30 days.

Maybe Rollie can explain a bit more about the chart and its numbers

[Rollie]

In reference to the Player Activity charts, that refers to players currently logged into the game, not players seen at some point during any number of shapshots.

For example, player activity would be like:

At 11:30 PM CST Server Time, 46 warrs, 10 druids, etc were seen online

The census however would be something like:

We have seen at any time in the past 30 days, 1000 warrs, 450 druids, etc

[Catrabela]

Have you guys thought of using the armoury for your census data?
Surly that would be the most accurate data for characters on the server and the class make up. You would still need to use your mod for your activity charts.

[xpolockx]

The armory doesn't give a character history at all, so it's not really the same thing. Plus, I don't think Blizzard has provided a way to trawl deeply into their database to be able to pull out truly accurate census information. I guess it could be done hypothetically, but I doubt it'd be worth the work.

[Rollie]

You could do it, and I've thought about doing it, by farming the Arena data. You would have to say that the propensity for players of either faction to play on an arena team, and have equal footing as to ability to play should be the same. As such, you could farm arena team data for battlegroup and server information that should be able to give just as valid survey as what we do here.

Maybe over the holiday weekend I can spend some time prototyping that out.

[xpolockx]

That'd definitely be a good idea, although you'd still need the census mod for lower level players as well as the decent amount of players that never do arena.

[uigrad]

..... You would have to say that the propensity for players of either faction to play on an arena team....

Sounds cool, but my guess is that Horde players tend to have a higher propensity to be interested in pvp.

As far as class makeup, I think that it would really be inaccurate. People don't level rogues for the purpose of raiding, but I know a few people who have leveled tanks just for the purpose of raiding.

I have pretty strong confidence that WR has located all 70s. Escaping all census while leveling 1-70 seems impossible to me. If you want to know the class balance for a server, I think that the biggest questions are

1) How many of the 70s listed are still on active accounts, and still on this server?
2) How often are these 70s played?

Armory could help answer question #1. I know of some 70s from my friends list who are no longer found an armory.

Question #2 seems to be the tough one. I don't play either one of my 70s very much, especially the warlock. If I'm going to raid, I'll be asked to bring my priest. I would really be interested to see activity graphs based on class, and then compared to the standard player graphs based on class.

[MarkButler]

I'm interested in the logic of the data accumulation. Is there a faq or something, for example, how long does it take for a character not to be found by the various censuses on a server to be considered no longer there and be removed from the list.

[xpolockx]

From what I understand, the only characters that have been removed are the 1-10 "throwaway" characters from years ago, and only those to help alleviate the strain on the database. The history is important here, even of characters that are no longer active nor have been in quite some time.

[Balgair]

Some other older and very inactive ones have now been removed, one of mine that I deleted just before TBC to make space for new characters is no longer findable in the database (and it was lvl 23, not 1-10)

[MarkButler]

So all data is kept, but the normal graphs (US Realm Stats) are current-only. Works for me.

This does give some opportunities for new data mining, graphs-over-time. Did servers start with a different population? are populations gradually shifting toward something?

[Rollie]

Granted I haven't really looked hard at the numbers, but the only time I saw real shifts is when TBC came out. I think with Wrath, we'll see the same with many people taking the opportunity to switch up characters, especially DK's, etc.

[xpolockx]

Some other older and very inactive ones have now been removed, one of mine that I deleted just before TBC to make space for new characters is no longer findable in the database (and it was lvl 23, not 1-10)

Interesting, I looked up some old guildmates of mine that I know haven't played since '05 and their characters are still there. They were all 40+ though, so I don't know if that plays into it at all. In general though most of the entries that I've seen are still available, even going pretty far back. I guess some slip through the cracks however

[Rollie]

The cleanup process is removing chars not seen in like 18 months that are under lvl 40