Mafia/NOTW Statistics

[Zack]

I made a new and improved version.

Spoiler :

• I deleted The Thing 2, since it died early and is an outlier anyways.

• I added a "Moving Average" trendline. The period is three, meaning it forms points by averaging the last three data points, making it a good indicator of trends over time.

• Colors -- the red line contains the actual data points, the blue line is the linear trendline, and the green line is the moving average trendline.

• Added the r² value. A value near 1 is a good linear fit. As you can see, this graph has an extremely weak linear association. The linear trendline is only good for showing the big picture trends, and even then it's debatable.

I'm now working on one with game posts.

 

[Jarrema]

yeah, thats good.
also interesting could be correlation between mafia and NOTW participation.
As for R-squared: I would not expect it to be very high for regression without any additional predictors.
I would prefer some "longer" trends for moving average. Like 25% of whole dataset (is it called different, when you do that? Lowless or something?)

 

[Zack]

As for R-squared: I would not expect it to be very high for regression without any additional predictors.

Like what? I'm thinking I should just remove the LSRL.

I would prefer some "longer" trends for moving average. Like 25% of whole dataset (is it called different, when you do that? Lowless or something?)

That would be too big for any meaningful trend. 25% of the dataset would be almost a whole year.

 

[Zack]

Posts in a game (excluding posts by the host). LOTR is excluded because it's a ridiculous outlier.

Spoiler :

Sign-ups vs. Posts.

Spoiler :

Sign-ups (now without LOTR).

Spoiler :

 

[Takhisis]

Can you do a count of non-Takhisis posts? Then a non-pizza post list, too.

 

[Zack]

I'm actually working on some models that use data gathered from a game thread after Day 3 to predict the total number of posts for the whole game. Some of them work scarily well.

 

[Takhisis]

But how would you include replacement players?

 

[Zack]

What? Not for individual players, for the whole game. Also working on predicting the number of day phases the game will last, which is harder.

 

[Takhisis]

My mistake. How would you model the way in which the avilability of replacements and the willingness of the hsot to use them -or WoGs- affect the duration of the game?

 

[JohannaK]

This is awesome.

 

[Jarrema]

Like what? I'm thinking I should just remove the LSRL.

Maybe try for dummy variable, like: hosted by Zack vs. by anyone else. And others like that (of course, if you feel like doing something like that)

 

[edse]

I would rather see time on the x-axis instead of number of games.

 

[Zack]

I would rather see time on the x-axis instead of number of games.

LOOK AT THESE BEAUTIES:

Sign-Ups vs. Month

Spoiler :

Sign-Up Capacity vs. Month

Spoiler :

Sign-Ups over Time

Spoiler :

Player Posts over Time

Spoiler :

 

[Double A]

It's a sign of the devil's handiwork, surely.

 

[choxorn]

So, I was thinking to myself a few days ago, "You know, updating the database can take a long time and be tedious, can't it? If only there was a way could like, get a machine to do it for me." Followed by "Wait, choxorn, you took some computer science classes this year, maybe you can write a program that updates the database for you?" And, several hours later, it was somewhat shoddily written, and it appears to do a heckuva job of both updating player statistics and the "Best Of" section, much faster than I can and with much less effort from me. I don't as of yet have a way to make it do anything else, but those were the only two parts that took much time- keeping track of winning streaks, updating the list of games, and setting the totals of games played and wins by each side all take barely any time or effort at all. It would probably be more effort to actually write code to do that for me than to just do it myself.

TL;DR version: I made a compter program to do most of this work for me, which means I can update much faster than before.

I might still be a bit procrastination prone, though, as it does still take some time. Like, right now, I'm about to have to deal with finals, so I really lack both the time and the energy to add more games right now, but look for updates in about two weeks or so.

 

[Zack]

So you manually input the data, and then it's automatically sorted? Sounds like a huge improvement. I was disgusted that you kept track of all of it in a bunch of Word documents.

 

[Takhisis]

Lol, Word documents.

.odt is the format of the future.

 

[choxorn]

Basically that. The word documents were the only way I could come up with for how to do it before.

In retrospect, that was terrible.

 

[Zack]

I'm afraid you opened the wrong Microsoft Office program for inputting data and automatically sorting said data.

 

[Takhisis]

D'you mean to say, chox could've used PowerPoint all along?