GOTM18 Statistics

[mad-bax]

sometimes there are posts you just wish you could take back. This is one of them. I am just gonna get myself deeper and deeper and deeper into a place I don't want to go


To answer your question I need to use even more jargon I'm afraid. Sorry.

The residuals are normalised, which basically means that a residual of 2 indicates that the data point in this example is around 5000 Jason points from the fitted curve. In the fitted data example, the data set is naturally split into two subsets and the two subsets do indeed represent the winning games and the losing games. The reason for this is that the losing games have a higher residual, they are further away from the fitted line (Y0).

The fitted line is weighted heavily for winning games as there were more of them. So a data point at X=7000 with a residual of -2 represents a Jason score of around 2000. The two subsets of data also have opposite gradients because for a winning game the fewer turns you take, generally the higher score you get, whilst in a losing game the longer you survive the higher score you get.

Because there is no win bonus for a losing game teh only source of variation is the rate at which an individual game accumulates Firaxis points, and this is why the losing game subset is very nearly linear.

I am up to my elbows in sh!$..... but my hands are above my head!

I really am not a statistician. Be nice and let me off the hook now. Please?

EDIT:

I just looked at the charts again, and I did actualy remember to scale the Y axis with the gross residuals (not normalised as stated above). So a data point at X = 7000 with a residual of -5000 represents a Jason score of 2000. Sorry for any

BTW you guessed right... I am an automotive engineer by trade. Why do so many engineers play CIV?

 

[Offa]

Originally posted by mad-bax
sometimes there are posts you just wish you could take back. This is one of them. I am just gonna get myself deeper and deeper and deeper into a place I don't want to go


BTW you guessed right... I am an automotive engineer by trade. Why do so many engineers play CIV?

You shouldn't regrat your post. I think its really cool. Thanks for your reply to my query; I almost understand it now.

I didn't know that loads of the players were engineers, I thought they were all programmers. I am a doctor myself, and have always assumed that I am the only one playing GOTM.

 

[ControlFreak]

Originally posted by mad-bax
I really am not a statistician. Be nice and let me off the hook now. Please?

Dang that one got away. Well, string up another lure and cast...
I think your response was the graph vs the Order of the data (low to high Jason score). I think I understand the nature of its S-curve and how your comments relate to it. My question was about the graph vs Fitted Value. What is the fitted value? You said that the regression used the number of turns (read - end date). So that would make the Fitted Value the predicted score for the date a game ended. If you were below the predicted score, you're residual would be below the line....{as I'm typing my question, I think I'm figuring out the answer}...The prediction is assuming you won since most people did. Thus the group of dots below the group are the losses...:grad:...OK, nevermind. I figured it out. With a linear regression, fitted score is related to end date so the vs. Fitted graph actually is a representation of games by end date. Its interesting to see that there were "losers" thoughout history.

Now my question is what are the few games with high positive residuals? The really early ones (fitted value~2500) are probably early conquest with the Jason bonus? The later ones are really early culture,diplo,space victory? Or do they highlight the extreme milkers. Like could you circle Moonsingers dot as the high point at 4800 or 6000?

Originally posted by mad-bax
BTW you guessed right... I am an automotive engineer by trade. Why do so many engineers play CIV?

I think its the problem solving aspect of the game.

Originally posted by Offa
I am a doctor myself, and have always assumed that I am the only one playing GOTM.

Doctors are just Engineer-wanna-bes that hated Circuits.

Actually I was a Doctor wanna-be for a long time. I even took the MCATs and did decent but never applied. Then I went and got my Masters in Biomedical Engineering before taking a Controls Engineering job in the Automotive industry.