Crime & Punishment

I advice you guys to keep that up. It seems that's Gevlon's one and only purpose in life. Everyone needs something to hold on in life. Gevlon's stage is EvE Online.

You didn't limit your data to losses caused my highsec wars, in highsec, so you can't use the analysis you've done to support a conclusion that highsec wars have anything to do with it.

Still stats 101.

e: Though to be clear, your assertion that the data you are analyzing is an indicator of 'activity' is flawed as previously mentioned, so even if you did this, you're just concluding chocolate ice-cream causes murder.

do you realize how horribly you misinterpreted the correlation coefficient in your blog? hilarious.

The correlation coefficient is a measure of the degree to which the movement of two variables are associated, not the degree to which the dependent variable is explained by the independent variable.

You are confusion correlation coefficient, which is [cov(x,y)]/ SxSy

and a linear regression coefficient, which is a slope coefficient on the independent variable (period 1 kills-losses) on the dependent variable (period 2 kills-losses). The slope coefficient would indeed explain how much a net loss in period 1 influences period 2 net losses. It is calculated using the Ordinary Least Squares method, which results in a linear equation of:

Y = a +b*x + e

Y being period 2 kills-losses, X being period 1 kills-losses, b being the regression coefficient you wanted (% of period 2 kills-losses explained by period 1 kills-losses)



Yes, you are missing many, many factors including real life events, wars, in-game events, economic effects, recruitment/alliance switching, deployments, doctrines, logi/scouting/dictor use, and many others.

Your model, and mistaken as it is, is heavily influenced by exogenous factors. Your correlation coefficients don't prove anything, in fact anyone with a brain could have told you that you would have found a somewhat significant result. active players are likely to have been active 6 months ago. However, you cannot calculate the effect of losing a 1bil freighter based on your 'calculations'

Your regression would be incredibly oversimplified, had you correctly performed it. 6 months of pvp activity is influenced by many more factors than a single 1 billion isk freighter loss in a prior 6 month period. Any explanatory power you might have found if you had done this properly, would be a result of the fact that you have excluded countless exogenous variables. I believe if you properly conducted a linear regression, your F statistic and adjusted R^2 would reflect that.




You used the fact that variance is greater than the sample mean (which actually just means that the data is highly varied) to explain that you could make the following calculation:

ganked ship * variance / mean * correlation coefficient = effect on next period kill-loss

What equation is that exactly? I actually lol'd at that. You calculated a correlation coefficient, not a linear regression. Regardless, what the hell is that equation? I'm still giggling. You need the slope coefficient to predict the effect of a current period loss in the next period. You did not perform a regression. The correlation coefficient does not work like this. This equation is not a thing


The higher the correlation coefficient that excel calculated for you (dont blindly use excel formulas unless you understand them to avoid embarrassment like this), the closer the two are to a perfect linear relationship. Logically, that sounds reasonable for your chosen variables.

What you tried to interpret that as, is the regression (or slope) coefficient. The slope coefficient is the extent to which the change in the expected value of period 2 kills-losses is explained by period 1 kills-losses. That is what you were going for, but they are very different in both calculation and interpretation.

Your calculation of the correlation coefficient explains precisely nothing. It has no weight over what you were trying to prove. Try again.

PS: de-whoring kills is ********. Basically only shows small gang and solo kills as a result. de-whored kills - losses will be VASTLY different than just kills-losses

TLRD: A correlation coefficient is not a regression coefficient. Try again with a linear regression and post the F test, T statistics, R^2, durbin watson statistic. You cannot predict the effect of a period 1 freighter gank on period 2 kills-losses using a correlation coefficient. Using excel formulas without understanding the statistic provides great laughs for everyone. Try googling regression vs correlation coefficient next time

To quote my stats prof

"Stats: The great archane art of saying anything you want with math."

Then he went on to talk about the Abe Lincoln internet thing.

I really think y'all are getting trolled by this Gevlon character. No one can possibly be that stupid.




... right?

I would like to think that.... but the extent he has went to... dedicating a blog... multiple forum pages... a failcrusade in hisec against nullsec sov corps....

He is trolling himself...

It can still be fun but you do have to just skim read. "blahditty blah goons dying blahditty blah ragequit, blahditty blah I'm nearly poor from paying for wardecs". But you're right, it's gets boring quick.

He say this kinda think all the time "OMIGOSH! Amazeballs post coming" *posting of the same stuff he's posted in every post for the last year with a graph in a slightly different colour*

I have done some advanced maths to show how Gevlon is going broke.

Gevlon's Wallet - grrr spending = Gevlon's smaller wallet

Now I don't have no fancy-pants excel like all those high-tier bloggers out there, but I have made a graph of Gevlon's ISK.

Fig 1. - http://i.imgur.com/hqfGjeV.png

Part A - throwing money at marmite / lemmings
Part B - dropped 20bil to Chribba for his Titan thingy
Part C - the blog-rampantly-as-my-funds-empty-phase.

thank you all for your time, please RT and spread the word #gevlonfinances

These are the tactics of the lie.