OSN Data Sharing Initiative

[GreatGonzales]

(04-24-2014 07:03 AM)His_Daddy Wrote:  

(04-24-2014 06:27 AM)GreatGonzales Wrote:  Interesting, haven't heard of this. I'll have to file this away - did you teach yourself how to use R?

Nah, learnt it from class. I was stat major.
SAS, R, STATA, SPSS, Sudaan are the top packages in stat area and R is my personal favorite.
It's not that difficult to learn though, basically you can find most of the answers to your question on google.

Cool. I've used STATA in the past for stats. As an aside, I may have missed my calling as a statistician. I get excited about massive spreadsheets.

[GreatGonzales]

What do you all think is a more interesting statistic:

Win percentage of fluffy adorables players vs fluffy [insert team] players

or

Win percentage of fluffy adorables players vs ALL leagues [insert team] players

Basically the question is, do we need to only consider matches between two persons of the same league, so it's an apples to apples comparison. If we include matches from people of two different leagues, this will diminish the win percentage of fluffy, inflate the win percentage of ST (for example). But it might still be interesting?

[Necrocat219]

It may be interesting, but it's also less accurate. Kinda. There's pros and cons for both. The main reason I would prefer just fluffy vs fluffy only, clever vs clever only, is 1) it removes some of the replays when players are climbing the ladder from bottom to top, making it more realistic and 2) you could instead have the categories FvsF, FvsC, CvsC and so on for an even larger range (some may be small samples though). You may be able to even see how much of a gap in skill there is between high and lower rated players there are and which special lower rated players are best with. You can also see how many games where a ST is matched to a fluffy.

With the first option you would have duplicate matches crossing over different categories, making comparing league usage and w/l or races less accurate e.g 500 games are gifted vs master, this would be duplicated in both the gifted and master dataset.

[GreatGonzales]

Good point on matches being duplicated if we go with the second option. Anyway both are possible, but the second option is more difficult for me.

[.Memories.]

Here's a quick go at some stats...

Here are results for 1v1 CLAN vs CLAN league games from the (4-18-2014) data set.

Code:

results =
            Anon  DL    GG    IBBP  iPro  WP
     Anon   0     6     0     0     0     2
     DL     7     1     0     0     0     3
     GG     0     2     0     0     0     1
     IBBP   0     0     0     0     0     0
     iPro   0     0     0     0     0     0
     WP     0     4     0     0     0     2

Rows = winner, columns = loser. So, for example, Anonymous beat DL 6 times and WP 2 times.

I'm planning on doing more Clan stats based on the data - this was just something quick after I got my program working.

[Mag!cGuy]

IBBP still exists?

[CodePenguin]

I just dumped the rest of the data. Let me know if anything looks odd or missing.

[.Memories.]

(04-26-2014 11:40 PM)Mag!cGuy Wrote:  IBBP still exists?

On OSN they do.

---

(04-27-2014 12:27 AM)CodePenguin Wrote:  I just dumped the rest of the data. Let me know if anything looks odd or missing.

Thanks!

[lawtai]

Wow that's a lot of data.

[GreatGonzales]

(04-27-2014 12:27 AM)CodePenguin Wrote:  I just dumped the rest of the data. Let me know if anything looks odd or missing.

Eeeeexcellent.