Especially for H/A ratio, was it actually accurate?
Shouldnât be very relevent, since you unsubbed.
6 Likes
Blizzard never released information so we will never know how accurate the figures were. One aspect to keep in mind is that the ratio might be accurate even with measurement errors, as long as the errors are the same proportion of the total population figures. Basically, the ratios would be incorrect if the data gathering methodology was somehow biased towards a specific faction.
Highly unlikely it was remotely accurate. Due to /who not returning a complete list of all the people of that level if there are more than it can display, its entirely possible for the census to never see all the characters (because it chooses randomly what characters it shows you).
You can overcome that issue by collecting the data with higher granularity. You could query level, class, and zone.
1 Like
Even then, with tens of thousands on players per server, its still entirely possible, and even quite likely, that the add-ons never saw everyone.
The same add-on, when it was on live, was WAY off. Like⌠10+% off.
Because in Retail/Refail/Live whatever you want to call it you can simply mine the Armory API to get totally accurate results.
They werenât that precise, ever.
1 Like
The add-on accounted for that.
He explained his methodology.
Most of the people who objected have been incapable of coming up with an alternative methodology that they think might give a better picture.
These things come up for discussion because the concept of estimating the game population is interesting, not because they are 100% accurate.
1 Like
The addon will only collect people on at the time it is executed. One could argue that doing this for a few weeks is more accurate than an armory scan. In a way, the armory doesnât account for how long you are online. I find their methodology to be quite good.
1 Like
Characters in Classic have no armory entries, though for retail that has always been commonly used.
We know. Not sure why you replied with that groundbreaking discovery.
⌠you consider 10% âwayâ off?
Not sure what kind of accraucy you were expecting but Iâd be happy with 75%.
Why do you care OP, you unsubbed didnât you?
To answer the question. They were very accurate which is why Blizzard shut it down.
1 Like
The addon in question uses queries based on level, name, class, etc to dig down until it gets everyone. It will eventually get a scan of everyone on the server, so itâll see a whole population of people who are currently logged in.
There are two problems:
- It counts characters not players
- It only sees characters that are online at the time of the scan
So if you scan every day at different times of the day, you should see a large fraction of the total playerbase⌠BUT
- How many players were missed?
- How many players were counted multiple times because they have multiple characters?
Nobody knows the answer to this.
Thus, the data is really only accurate for comparisons between populations. If I know the alliance are also scanning regularly, I can infer that the alliance to horde ratio should be a reasonable approximation of reality. If Iâm just using my data, I can be pretty sure that class ratios should be fairly accurate.
However I canât say with any real confidence how close the scans are to the total population. Thatâs not something this data can do by itself.
I never used it, but supposedly, it would use /who to see how many of a class was online for a given level bracket. If the query returned a full search, it would refine the search until it wasnât a full search. For example, there may be 100 level 39 warlocks on, but if you search for just level 39 warlocks whose names begin with âRâ etc. youâll eventually pick up the entire 100 warlocks in your count.
Thatâs actually much better than if it shows the same slice of the population every time. At least if it ran a few times it could count up unique names and get an estimate of how many it was missing in a scan. This would give an approximate measure of the true size of the population.
If it showed the same people every time then youâd have no idea of how many were outside your slice. There are ways of narrowing your search that can help this issue.
Anyways, they are likely to be true within a certain amount of error, Iâd say if the measured faction is 55% then itâs probably anywhere from 50%-60%. The numbers arenât great if you want an exact measure but they should be good enough to get an idea of the balance.
Of course, it all depends on the sample size and when the samples were taken. As time goes on the data is likely to be less and less relevant since itâs not getting updated.
Confidence interval dependant on user agenda
It was extremely accurate.
What? He unsubbed?
Uses Google
Oh, he unsubbed.
Yes, you donât need the info anyways. Youâre not going to be playing the game anyways.
1 Like
Well, this is embarrassing.
People donât forget
1 Like