Question about 2.5.1 and double counting

[sardrimm]

I was not sure where to post this. I searched several forums and decided this was the most appropriate place.

First off... Love the mod! Love the info on your site. I didnt realize why it always took so long on Azshara to get even a PUG going. Azshara is a dying server...

Now I am trying to find a new server to move to and using census+ I am able to evaluate the contenders during times that I play.

Now for my question... I have the verbose mode activated so that at the end of the scan I can get the number of toons that were actually seen on the server at the time census was taken.

On heavily populated servers I can see census crunching through the /who list and taking smaller and smaller bites to get the data it wants. If census cant get a list of 49 names or less it finally is trying to break the who list by the letters a-z.

My main questions revolve around census getting to a line like this:
/who r-"<race>" c-"<class>" 60-60 n-a

1) Rollie, do you realize that this returns characters with an "a" anywhere in their name? (not the first letter which is what I thought it would do).

2) Because of the behavior described in question 1 do you actively prune the list being currently created because character names may show up more then one time during a scan? (example: The name Abe would appear during the /who for the letter A, B, and E).

3) According to the verbose logging it looks like there are certain letters that are skipped when census+ is at this level of detail. I can not remember specifically which ones they are, but they are like X and V and Y or something like that.

4) There are certain times when even the example line given above returns 50 or more /who entries. I would assume that census believes it will pick up the names from that letter using a different letter of the alphabet. (i.e. the name Abe would be picked up in the B scan if the E scan returned more than 49 entries). This is correct behavior because unless the character's name is "A" the name will be picked up in a different portion of the scan. However, if census is skipping letters as detailed in point 3, then it is possible that some names would missed (The name "Ay" would be skipped if census does not search the "Y" character and the "A" character returns more than 49 entries).

I just want to be certain that census is returning the most accurate information.

Thanks!

[Rollie]

1) Rollie, do you realize that this returns characters with an "a" anywhere in their name? (not the first letter which is what I thought it would do).

Yup, I am aware of this. This is actually sort of a good thing as the letters that are scanned were chosen for the high probability that every character would have at least 1 of those characters in their name.

2) Because of the behavior described in question 1 do you actively prune the list being currently created because character names may show up more then one time during a scan? (example: The name Abe would appear during the /who for the letter A, B, and E).

Yes, an name showing up more than once is taken into consideration while making the scan.

3) According to the verbose logging it looks like there are certain letters that are skipped when census+ is at this level of detail. I can not remember specifically which ones they are, but they are like X and V and Y or something like that.

Yup, only a certain set of letters is visited, chosen for the high probability that every name would have at least 1 of the characters.

4) There are certain times when even the example line given above returns 50 or more /who entries. I would assume that census believes it will pick up the names from that letter using a different letter of the alphabet. (i.e. the name Abe would be picked up in the B scan if the E scan returned more than 49 entries). This is correct behavior because unless the character's name is "A" the name will be picked up in a different portion of the scan. However, if census is skipping letters as detailed in point 3, then it is possible that some names would missed (The name "Ay" would be skipped if census does not search the "Y" character and the "A" character returns more than 49 entries).

That is the hope anyway. There are limitations to the system, and this is one of them. I feel pretty confident that only a very tiny and statistically insignificant number of characters could slip through, and this is something I can live with.

I just want to be certain that census is returning the most accurate information.

Thanks!

I think it's doing a good job at returning the data you are looking for, and even if 1 or 2 characters happen to slip through on a single snapshot, that you're still getting a pretty good representation of the demographics for that server.

Don't forget that you can use the http://www.warcraftrealms.com/activity.php graphs to see what kind of population a given server has had.