Results 1 to 20 of 20

Thread: Statistical Analyses of Inter

  1. #1
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF

    Statistical Analyses of Inter

    Does anyone here program? At least capable of using Python?

    I'm wanting to do a statistical analysis of the last ~10 years of Inter or so, wondering if anyone here would be interested in helping out. I'm intending to design most of this program in python...

    What I will aim to do ultimately is put in all the information of the last 10 years of Serie A for every club, and looking to see what sort of indicators we can use to demonstrate form. look for what correlates best with, for example, managers being sacked. etc.


    If you dont know how to define a class in python, for example, this is probably not for you...

  2. #2
    Fapuccino's Avatar
    Join Date
    03 Apr 12
    Posts
    12,075
    Thanked
    9,265 times
    Fav. Player
    Schelotto

    Brazil

    10 years ago is useless. Start ever since Mancini took over, and just plot graphs which show the trends of collecting points/rankings.

  3. #3
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    I'm planning to do this in a bit more depth than just collecting points and rankings

    I'm planning to go back reasonably far, it just depends on how easily I can find the data in a ready format I can use for quick I/O in the program, and also making sure that we can make do it week-by-week analyses.

  4. #4
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    I'll take the silence as no interest then

    :foreveralone:
    Last edited by .h.; 04 Feb 13 at 12:21.

  5. #5
    Wobblz's Avatar
    Join Date
    05 Aug 12
    Posts
    5,448
    Thanked
    7,032 times
    Fav. Player
    Baggio

    Bulgaria

    33
    I could help you with the graphic design part. I'm even more than willing - I insist.

    I was starting to do some Processing programming a while ago but the sad part is I doubt we're going to find a ready-made DB for that and collecting data by hand would be tedious.

    Anyhow, I think football stat companies like Opta have this kind of databases but there's no way we make them give It to us, especially for free.

    -

    You need a proper thought through formula to begin with, depends on what you consider as 'performance'. The most conceptual one would be point based or more so points/ranking based. A slightly deeper thing would be to implement goal difference. The ability of a team to score and defend at the same time in the most straighforward perspective. Everything else depends on the style of play - passes and pass completion, crosses, etc. For example Rano has impressive stats every game but that's because our defense is under pressure all the time so such a comparison needs a great deal of afteranalysis and conclusion drawing.

    For a model to be as precise as possible you should pour an immense quantity of data and look in it in a myriad of ways. Even If you make this 5 year ranking by round analysis you would be able to only draw vague conclusions what would be the causes of form changes. Maybe marking important events such as new coach appointments, international breaks, bad losses and important victories might give some more in depth idea of what drives a team to perform well but I guess It won't be enough for It to make a point.

    -

    Anyhow, It would be way easier for you to compare last 3 years only. Thing is remembering or doing research about important moments that happened 10 years ago would take a lot of time and might be incorrect due to the lack of adequate data. For example. such an important event that could cause a dip in form is an injury crisis but if you want to be objective you should gather some data for this specific information on an entirely different note.

  6. #6
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    Well, the data I/O is the most important thing first of all, after that it gets to be very easy accessing it all and coming up with whatever metric you want.

    I've found a few websites which offer reasonably standardized week-by-week result reporting so I dont think it will be too difficult to build a program that will just parse that part and store it

    - - - Updated - - -

    though, interesting you mention Opta stats. I might apply for a job with them once I'm finished what I'm currently doing.

  7. #7
    Wobblz's Avatar
    Join Date
    05 Aug 12
    Posts
    5,448
    Thanked
    7,032 times
    Fav. Player
    Baggio

    Bulgaria

    33
    Quote Originally Posted by browha View Post
    Well, the data I/O is the most important thing first of all, after that it gets to be very easy accessing it all and coming up with whatever metric you want.

    I've found a few websites which offer reasonably standardized week-by-week result reporting so I dont think it will be too difficult to build a program that will just parse that part and store it
    As I said, depends on what you want to achieve - hence, first the formula.

    It's great that you've found some standardized result DBs but It's going to be exponentially harder if you wish to make the formula more complex.

    Quote Originally Posted by browha View Post
    though, interesting you mention Opta stats. I might apply for a job with them once I'm finished what I'm currently doing.
    I tried to make something similar in the beginning of last year when I was out of job and really wanted to mix football and design. Contacted a couple of those football stat companies, including Opta - even whoscored.com to see whether they're interested in hiring a freelancer for data visualisations - no one even replied.


    BTW If you're into this kind of thing check Processing.

  8. #8
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    Once you have the data in, though, it gets easier. I'd look at simple things like measuring mean deflection from the seasonal average points to get a measure of form, stuff like that.

  9. #9
    Wallace
    Join Date
    18 Jun 05
    Posts
    11,213
    Thanked
    10,241 times
    Fav. Player
    Wallace
    Old name
    Wallace

    Brazil

    88 Forum Supporter 10 years of FIF
    Is there a reason why you wanna use python?

    Quote Originally Posted by Wobblz View Post
    Contacted a couple of those football stat companies, including Opta - even whoscored.com to see whether they're interested in hiring a freelancer for data visualisations - no one even replied.
    Speaking of Opta and whoscored.com

    Do you guys know how they acquire their stats? Are they real-time during live? Whoscored.com mentioned nothing about their stats acquisition method nor their statistical accuracy. Not sure about Opta, will read its site in a bit more detail later.


    Sent from my GT-I9300 using Tapatalk 2
    Last edited by Wallace; 05 Feb 13 at 07:45.
    Wallace

  10. #10
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    familiarity, to be honest. No idea how they get their stats.

    Python would make it quite eeasy, parsing all the data and loading it up.

  11. #11
    Wallace
    Join Date
    18 Jun 05
    Posts
    11,213
    Thanked
    10,241 times
    Fav. Player
    Wallace
    Old name
    Wallace

    Brazil

    88 Forum Supporter 10 years of FIF
    The most important thing to discuss here is the source of data, without it you can't really start this whole thing.

    I've been looking around, and it seems like there isn't any freely available statistics feed. ESPN's API supports GET's in JSON format and seems to have quite a full set of database, but after navigating a bit, their stats only seem to be provided for those premium members who are willing to pay.

    And then there is xmlsoccer.com, but it clearly says it requires monthly subscription.

    Which begs the question, what data source do you have? Are you looking for only teams' data, or are you looking for players' data too?
    Wallace

  12. #12
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    Currently I was going to start off only working on a result-by-result basis - for which I've found a few websites (at home, can link later) which will do the full week-by-week results for Serie A for quite a long time.

    I've been talking to Vito about it a bit, though, and he would like to expand it to things like number of passes, shots, etc as well - which of course, as you say, is much harder.

    If its available on a premium pay website, we can pretty much write something that will run through it all for all 10+ years of Serie A and download the stats for every match, and then cancel the subscription, no?

    on the XMLSoccer, it says it provides one league for free though? I dont know if that means current or historical, but that might work for time being?

  13. #13
    Wobblz's Avatar
    Join Date
    05 Aug 12
    Posts
    5,448
    Thanked
    7,032 times
    Fav. Player
    Baggio

    Bulgaria

    33
    You can do something else - contact the likes of Opta and Whoscored for example, tell them you up to doing such a thing and you need data, and that you'll give them the rights to spread your material with you as an author and 'powered by them'. It's a win-win.

  14. #14
    Wallace
    Join Date
    18 Jun 05
    Posts
    11,213
    Thanked
    10,241 times
    Fav. Player
    Wallace
    Old name
    Wallace

    Brazil

    88 Forum Supporter 10 years of FIF
    Quote Originally Posted by browha View Post
    If its available on a premium pay website, we can pretty much write something that will run through it all for all 10+ years of Serie A and download the stats for every match, and then cancel the subscription, no?
    Totally, but you'll have to host a server for your application to respond to requests, or are you simply doing this offline?

    But then, how are you gonna keep up-to-date with your dataset? I don't think you'd wanna do that manually .

    Quote Originally Posted by browha View Post
    on the XMLSoccer, it says it provides one league for free though? I dont know if that means current or historical, but that might work for time being?
    Not sure where you read this, but I only see that they are providing the free demo access to their database for the Scottish league, which is kinda useless...

    Quote Originally Posted by Wobblz View Post
    You can do something else - contact the likes of Opta and Whoscored for example, tell them you up to doing such a thing and you need data, and that you'll give them the rights to spread your material with you as an author and 'powered by them'. It's a win-win.
    It's worth a try, but I can't see that happening at this stage.

    In order for them to provide their data for you to access, you have to show that your product is actually advantageous for them to provide data for you, i.e. such that your product would provide marketability for their database.

    I can't see them providing data for anyone who has nothing on hand and simply go and ask to get accessed to their data, for free.

    - - - Updated - - -

    This is definitely free (OpenFooty): http://www.footytube.com/openfooty/

    Not really sure what data they have, but I see player.getStats, league.getTable, team.getResults, team.getStats, etc...in their API, sounds like they might have enough data to work with.

    Just registered for a key, have to wait for a couple of days for them to issue it to me.

    Perhaps you should take a look too.
    Wallace

  15. #15
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    I was just going to do it offline. It would be easy enough, I guess (though I have no experience with APIs), to have a python script that you manually run that will periodically update for more information?

  16. #16

    Join Date
    17 Nov 12
    Posts
    54
    Thanked
    18 times
    Fav. Player
    couuuuuuuuuu

    Malaysia

    Quote Originally Posted by browha View Post
    I was just going to do it offline. It would be easy enough, I guess (though I have no experience with APIs), to have a python script that you manually run that will periodically update for more information?
    why not spss ? and what kind of research design d u want ?

  17. #17
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    again its just a question of familiarity

    im sure there are tools better suited for the job, but i'm used to python and i've got plenty of experience of doing statistical/workload heavy applications in python to generate plots of things like this

  18. #18
    Fapuccino's Avatar
    Join Date
    03 Apr 12
    Posts
    12,075
    Thanked
    9,265 times
    Fav. Player
    Schelotto

    Brazil

    guys im sure whatever you're planning on doing you can do on excel

  19. #19
    Wallace
    Join Date
    18 Jun 05
    Posts
    11,213
    Thanked
    10,241 times
    Fav. Player
    Wallace
    Old name
    Wallace

    Brazil

    88 Forum Supporter 10 years of FIF
    Quote Originally Posted by browha View Post
    I was just going to do it offline. It would be easy enough, I guess (though I have no experience with APIs), to have a python script that you manually run that will periodically update for more information?
    In general, you just have to make a http GET request inside your script. You can make that perform automatically on a fixed date, it'd download the updated information in a format (i.e. XML or JSON), and you'd parse that into your application and store it in your database.

    Depending on the provider of the dataset, it'd probably have something in the form of http://www.****.com/leagueID/teamID/playerID/stats, or http://www.****.com/leagueID/results/timestamp

    So yes, they usually make it fairly easy for you to download the relevant data.
    Wallace

  20. #20
    .h.'s Avatar
    Join Date
    08 Jun 05
    Posts
    19,405
    Thanked
    19,122 times
    Old name
    browha

    Europe

    25 Forum Supporter 10 years of FIF
    Quote Originally Posted by Wallace View Post
    In general, you just have to make a http GET request inside your script. You can make that perform automatically on a fixed date, it'd download the updated information in a format (i.e. XML or JSON), and you'd parse that into your application and store it in your database.

    Depending on the provider of the dataset, it'd probably have something in the form of http://www.****.com/leagueID/teamID/playerID/stats, or http://www.****.com/leagueID/results/timestamp

    So yes, they usually make it fairly easy for you to download the relevant data.
    Cool. Good to know. I dont have much experience with API stuff, but what I guess I should do is do a 'proof of principle' first, just using this season's data for Inter, and once we have that working we can expand?

    Quote Originally Posted by ninuk View Post
    guys im sure whatever you're planning on doing you can do on excel
    anything we do is faster/easier in python.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •