SniffTheGlove Posted July 25, 2017 Share Posted July 25, 2017 Hi, it's been a while since I last posted here as life got in the way seriously. Anyway, I am looking to rebuild my systems as I still collect data every day from football-data.co.uk (since 2005) and I have just purchased a new programming language to convert my data collection software and also my prediction software to be faster etc.. One of the first things I am looking at is What makes a Football Team good?, ie how to discern this from football data statistics only! The obvious ones I can think off are... Goals Ratio: For/Against from last game, last 4 games and current season. Win Ratio: Number of matches won at home or away from last game, last 4 games and current season. Draw Ratio: Number of matches drawn at home or away from last game, last 4 games and current season. Lose Ratio: Number of matches lost at home or away from last game, last 4 games and current season. League Position Fulltime over Halftime results Distance? Do teams who travel a long way have a disadvantage? These are the main datasets one can work out from the football-data.co.uk database, though for distance data this is my own database table I have created for the location of every teams home stadium. I store the Long/Lat of the stadium and work out the distance using the Great Circle equation. How do you all workout what makes a team good/bad from data when working on your own predictions etc. Quote Link to comment Share on other sites More sharing options...
Matthew Posted August 7, 2017 Share Posted August 7, 2017 Hi, Define good.... Is this profitable? More likely to win a given fixture irrespective of odds on offer? Consistent? Always value? Good at upsetting the odds? Good away from home? Quote Link to comment Share on other sites More sharing options...
SniffTheGlove Posted August 18, 2017 Author Share Posted August 18, 2017 What I meant is what do other users use from data to indicate how good a team is. Unless you do stuff manually and use your own noggin to calculate how good 2 teams are using other data like injuries, most people will use just static data generated from W D L F A Dataset. Good is defined as good. The better team irrespective of external bookmakers data (odds) that will win. Quote Link to comment Share on other sites More sharing options...
Matthew Posted August 22, 2017 Share Posted August 22, 2017 OK, I'll post some information values and WOE's for my predictor variables when I get chance, that'll show you the best of what is at my, and others, disposal. Might put the mockers on my own ratings, time will tell! Quote Link to comment Share on other sites More sharing options...
Matthew Posted August 29, 2017 Share Posted August 29, 2017 (edited) As promised, Information Values for the predictor variables I use are available on this Google sheet:https://docs.google.com/spreadsheets/d/1EhkX6zJ8COYfaOR7K2-12Wr1K2fhOsPe0qxiZ_Z4Unk/edit?usp=sharing Fair to say none are that spectacular, ideally you'd want an information value > 0.3. Some good evidence that draws are very hard to predict! For all three outcomes my own ratings come out best of everything which goes to show that putting in the hard work helps! The best is my Season's Plus rating with an information value of 0.21 for HW, 0.01 for DR and 0.23 for AW. For anyone who is new to Information Value's and WOE:http://ucanalytics.com/blogs/information-value-and-weight-of-evidencebanking-case/ Edited August 29, 2017 by Matthew Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.