Collecting player statistics

[shehzadem]

Hi there!

I'm currently collecting and organizing Castle Fight player statistics in order to deduce some statistically significant corellations (e.g. how hours played affect W/L ratio), but since I barely started, the only method I know of that lets me get enough info is to use !stats on every player in lobby and ingame to get hours played and percentage of stay and then to check the replay for other info like W/L ratio, ELO etc.. Obviously it's a slow method and I can only get info on 5 players at a time.

My question is whether the aforementioned info is already stored somewhere that I haven't yet found. Any info is greatly appreciated.

[Stealer]

Yes.

It can be found under the section of the website that can be accessed by clicking on "stats".
https://entgaming.net/customstats/castl ... e/7725108/


All of the stats are stored in a MySQL (afaik) database so if you know a specific query you might be able to ask @art)y but it's probably be faster to write a python script with beautifulsoup to parse the webpages.

[shehzadem]

Yeah, I've already checked the stats page. Unfortunately, it doesn't have %stay and total hours played statistics (which is why I have to go ingame for it). And it's also unfortunate that I don't have any skills in Python since it really would be useful and I have to do it manually lol. Anyway, I've already asked arty and he said he'll think about it.

So far, I've got about 50 player stats, which is way too few to actually do anything useful with it (i.e. may not be accurate). But I'll post a few stats just for the fun of it.



NOTE: this may not be accurate due to small sample; N=53
Here you can see a few variables, of which only games won, games lost and rank are statistically significant (i.e. relevant) that affect ELO. Looking at the coeficients - for every game won, ELO is increased by 3.56, for every game lost, ELO is decreased by 2.81 and for every increase in rank by 1 ELO is decreased by -.13. Other coeficients are invalid due to the reason given above.

And a few images.

[aRt)Y]

As providing bulk data-sets is rather unpractical, you should stick to python public parsing. @Neco might be able to help you as he does the same for ID.

Sorry.