My little STATS corner (entire ranking - ranking projection - etc.)

0. Open access to data sheets

Data sheets 090209

For those who are interested.

1. Player's Descriptives I


a) Entire Ranking PPG WPG and GPD incl. (order by RANK)

b) Entire Ranking PPG WPG and GPD incl. (order by PPG)

c) Entire Ranking PPG WPG and GPD incl. (order by WPG)

d) Entire Ranking PPG WPG and GPD incl. (order by GPD)

Here you find an entire ranking table of all players that were active and had at least 10 games finished by 25.Jan.09 .
You'll find all Ranking Point Averages (PPG) , Winning Game Ratio (WPG), and some Games Per Day stats (GPD - top 150 only so far).

Any values as of 09.02.09 9 a.m.

Please note that some players are still in the table that had an active account on 25.01.09 but were deactivated any time since then. Most prominent examples are 'merremec' , 'Marcos' , and 'dekano'. In the hope and anticipation that they might come back I left them in the table to have an more accurate ranking - since they still hold there points.

2. Player's Descriptives II

TOP 50 Pointage scatter plot (20/01/2009)

As you all know there are several ways in order to gain a higher ranking position:

A ) increase your number of games ('Played Games') and maintain your points per game ratio
B ) increase your winning ratio [(WINS+TIES)/TOTAL] by playing games with less (and weaker) opponents
C ) increase your points per game ratio ['Average points'] by playing and winning higher 'pots' (more/better opponents)
D ) and of course any combination of above strategies

This scatter plot might help to understand in which ways players have achieved there ranking position.



3. TOP 100 Ranking Projection

Ranking projection in 30 / 60 /90 days from now (20/01/09)

I am curious if (and how) it is possible to predict changes in the ranking in the near future based on the available user data.
I started to look at the top 40 as of todays ranking. Using people's average points per game values and the average number of games they have played since they registered on this site I linearly extrapolated their future pointage and ranking.
Let's see how close I will have gotten when 30 / 60 / 90 days have passed.
Any thoughts and suggestions very welcome.

cheers Zerebbru2



Editado 11 vez/veces. Última edición el 11/02/2009 19:39 por Zerebru.

interesting question but so difficult to predict

i agree about the 9 first positions

I'll think in this

the main problem is that 90 days is maybe the shorter predictable time available. A 5 winning.-games streak will help you to climb up in the rankings ... and average points per games is not that representative for short-time players.

And you must take account of the "let's disturb the high ranked player" game style, that assures you a free fall on reakint top-20 and top-10. It's easy to reach that high... but it's difficult to stay there for half a year or so (like Azathoth or ender did, and Guanche is about to reach )

I think that with a minimum of games played (about 500) u dont make big moves but its very impredececible try it for people who is starting.
E.g. In 200 games I across from 54th until 4th.

-400 partidas 127-30-400 (*52*-(238.27)-0.32)
-600 partidas 215-36-600 (*4*-(506.87)-0.68)

Im agree u go until a point where then u got balanced not up too much and not down too either. The varianza not explained by variabilities is too high in ur table, really imho have 0 base cientifical. (but good work still so) Its good people do thing for community

Ranking projection with pointage etc.

Thank you all for your feedback. I agree with most what you have noted. However, I disagree with Guanche's notion that this venture is entirely unscientific - I think it might not be very elaborate but still a scientific approach. As you have pointed out the individual ranking seems to be highly unpredictable. Please see this as an experiment. Especially because this is a projection that is based on one single time point and the stats of each player regardless how they have obtained them. One idea is to readjust the projection after another 30 days and see if the prediction for time point 3 (90 days) gets better.
My thinking is, that I might not be able to predict the individual outcome for a given player but on average the prediction might be much better. I assume that on the long run - maybe more than 90 days - 'discontinuities' such as winning or losing streaks will even out. Let's see.
Have a good one!

Zerebbru2

p.s. I added the actual and projected ranking points, 'seniority', games per day, and average ranking points per game to the table.
p.p.s. If I only had access to the entire data base - I could incorporate more information (such as ranking point distribution, number of player per game and number of district per game stats and so on... - ACHILES: is there a way to get those?



Editado 5 vez/veces. Última edición el 21/01/2009 12:35 por Zerebru.

(I added some descriptive data to my first post.)

2 new tables look very interesting. Good job.
Sorry if I was too "tajante" when said it had not a cientifical base. Im agree with u that its a experiment but for what u can call it cientifical this must is under some conditions.Now we look ur way, u got averages of players; that can be enough to save the part of got variable and hipotesis and consider it cientifical

Thank you.
(Happy belated birthday by the way!). Don't worry Guanche. No te apurre. I am always happy about any critical response.
You are right there are many assumptions and little data - not to mention all the variance unaccounted for... But still I am intrigued... Let's see how it works out.

Zerebbru2

I like the 2 new files

Now, there are a lot of players with only a few games that are thinking to leave the game for a time
So, it's impossible that some players have your point's prediction (ranking prediction is good)

I wish I have the stadistics you asked to Achiles

why do you think all players of top 40 have more points than now in the next 30,60, 90 days?

I think i can anwser fran's question.
Due to the algorythm used to split the points after the games, high-ranked players get more points, as they have higher scoring (some would say they drain points to low-ranked players quicker). A new good-player that starts playing with 100 point will beat old good-players, but will earn more point on his victories than he would have earned in the past (his foes now have more points), and he will climb high quicker. So, due to large numbers effects, as time goes by, people climb higher and quicker, and you need more points to become #1.
Let me show you an example. Some 20 months ago, I enrolled a top-8 players game. Lower score of players was near 400 points (myself), higher score was some 800 (Azathoth's). Winner earned some 55 points, i think. Should a brand-new player have enrolled the game and win it, we would get directly to 155 points. Sounds nice, indeed . But... if that game was to be played now, there are 23 player that may qualify for the game having 400 points or more. And, in fact, the brand-new 100 points player winning the game (guanche is a fair example, he wasnt a frequent player at that time ), he would earn 78 points. In fact, you have to have 640 points now to join the game. In just 20 monts after the first match

I'm also wondering about the average points for ALL players right now. The point is: since some players left the site, did they leave while having less than 100 points? If so, we must also count all that extra points that are now added in some frequent-winner player. And that do acelerates the first process i described. Achiles, can you put us a light on this data, please? (just out of curiosity, of course )

just kidding on the subject at 1 a.m. . dont take me too seriously



Editado 2 vez/veces. Última edición el 22/01/2009 02:35 por mariano.

Thank you mariano for you clarifications. [Saves me some typing ]
Indeed it would be very interesting to have the whole picture (including the bottom of the curve players).
I will ask ACHILES about it right now!

Zerebbru2

Hi again. Now top 100 is up and running.

TOP 100 projection

Sorry, i almost do not had read the forum in two days, i'll try to make something tomorow to show all the ranking...

Achiles(X)

Hey everyone. I did some more stats lately. You'll find the links in my first posting. More stats and graphs will follow.
Any questions and request are welcome.
Cheers!

open access to data sheets added. (see first post)

or:

data sheets



Editado 1 vez/veces. Última edición el 11/02/2009 19:40 por Zerebru.

Thank you Zerebru2!