Falling Active Player Numbers

Lol calm down miner

You do realize… the modeling account for variations… also the data interval is by day giving you 700 data points … maybe you should look up forecasting and arima regressions…you literally have no clue what u are talking about lol

Sigh u data people are cringe. Maybe learn to code with a catalyst ?

I agree up to #6. That cutt throat culture is what makes finding a good corp that much greater.

But yes theres only 2 things that keep me playing. Pve and pvp. Pve has been stale since 2009. So pvp is the only reason im still here. But its very easy for me since ive put literally years into skills and isk.

No it doesn’t and no it isn’t.

For a professor you are certainly coming off as someone that doesn’t apply much rigor. However, I have no reason beyond the poor logic of your posts, to think you aren’t telling the truth, so perhaps you can post some evidence, from the EVE Offline site that demonstrates what you are saying?

Chribba (who developed and hosts the site, also confirmed above that there’s some pre-processing of the data:

You are wasting your time. Professors don’t know how to listen.

2 Likes

Happy cake day @Xeux

1 Like

a time series data set over 2 years does not contain 700 data points … and modeling does not account for the variation? LOl what are you smoking… what modeling technique are you using and what are you qualifications?

The modeling is being done by @Chribba , and he has explicitly stated that the various timeframes are displaying different data sets from a sampling perspective.

Why would you argue with the statements of the data provider as to their methodology?

You are under the assumption that the data point is on a daily basis and thus would end up to 2x365 for the two years, in this case that is not accurate. The number of data points are dynamic depending on the timeframe viewed in the graphs, I admit is not logical or perhaps even the best way of doing it, but that’s how it is working.

An example, I use the slider to estimate two years, and I get 585 data points back;

1324621500000,38544 -- Friday, 23 December 2011 06:25:00
1324731000000,32695 -- Saturday, 24 December 2011 12:50:00
1324840500000,38503 -- Sunday, 25 December 2011 19:15:00
1324950000000,40022 -- Tuesday, 27 December 2011 01:40:00

Monday the 26th is completely missed there as the returned results is NOT by a daily basis but rather a couple of math calculations based on how far back and how big of a field one searches on. Now if you were to do the default 36h search, you’d get returned some ~2100 data points, pretty much one per minute (+/- missing data points for downtime, errors and misc).

So the assumption that two years would be the 730 data points would be true if it in fact did one per day, which it’s obviously not doing in this case. As to why this is being done is because the millions of data points in the database and for performance reasons the database groups and takes max() values from that grouping.

In the above case a group consists of data points located in a 1d6h25m timeframe, and that’s why we end up with a ā€œmissingā€ day, and as a result over a two year period we end up with 730/1.25=584 (as mentioned 585) data points.

/c

3 Likes

At this point, it’d be interesting to see any actual qualifications. I’m a chemist, not a statistician, but I’m not trying to claim any authority here. I’m just seeing that you don’t know what you are talking about.

No, modelling doesn’t account for the variation.

Additionally, I’m not using any modelling technique, nor was Pedro. He was simply pointing out correctly, what the data represents and why it isn’t simple to just do comparisons between the different time series.

This is something that any Stats Professor would be able to see correctly, yet you can’t. I’m sure even a 2nd year uni math student would understand it; and I’ll go give a few the data today and see how many of them are smarter than an internet professor.

Ok. What part of reading don’t you understand? In order to answer the question regarding PCU properly. The data must be transformed into a time series and an arima regression must be fitted on the data to conduct forecast modeling. It’s like what folks do with covid cases modeling. Hence then we could see if the change we see is statistically significant based on historical data. None of the so called method here presented uses this method. This is the only scientifically appropriate way of doing this. Lord people can’t read. Modelling doesn’t account for variation lol go get an education that is exactly what modeling is for…

Reading is fine, however you seem to have completely missed the point of Pedro’s post before telling him to try harder.

He was specifically pointing out that there isn’t a direct comparison, yet you rubbished his post like an idiot.

Further to my previous post, a few student at uni have had a good laugh at your level of understanding so far. I’ll post some graphs later. I’ll try to make them easy enough for even a Professor to understand.

1 Like

It’s ok. Modeling doesn’t account for variation dude. Lol good one

What modelling. I’ll just call it now. You are a charlatan. Certainly no Professor.

What modeling? Almost any modelling accounts for data variation… in this case the technique is called ARIMA forecasting. Oh the internet is filled with people who don’t know anything but pretends to lol

OK, I’ll dumb it down.

What modelling is done in the EVE Offline data?

Ok learn to read. ARIMA forecasting. How many times can one type it lol

So you are saying that ARIMA Forecasting is used on EVE Offline before the data is plotted?

Is that the claim our Stats Prof is making?

… you use historical data and it’s variation to predict the future… argh it’s ok no point arguing u r what r u lol