predictor variables for regression model

[Ace123]

Hi guys, im building a logistic regression model was wondering if you could recommend any variables that i should consider in my model. so far vars like % of last 6 games won are coming in v well. was wondering if there is anything else i should try out. Cheers

[Mr Intensity]

Re: predictor variables for regression model Think it's depends on what you're trying to predict:D

[Mr Intensity]

Re: predictor variables for regression model Just seen your others posts so ignore me:tongue2.

[asc]

Re: predictor variables for regression model At betfair labs, I came across victor the predictor (http://content.labs.betfair.com/victorthepredictor/). It's a tool for predicting UK soccer matches and makes use of neural network techniques. The following input parameters are used:

• Home Team ELO Ranking, continuously calculated since start of data.
• Away Team ELO Ranking, continuously calculated since start of data.
• Home Team ELO Ranking Medium Term
• Away Team ELO Ranking Medium Term
• Home Team ELO Ranking Short Term
• Away Team ELO Ranking Short Term
• Home Team Points won in last four weeks
• Home Team Points won in last ten Weeks
• Home Team Points won in last thirty weeks
• Away Team Points won in last four weeks
• Away Team Points won in last ten weeks
• Away Team Points won in last thirty weeks
• Home Team Ratio of Home wins to Away Wins
• Home Team Ratio of Draws to Total Matches
• Away Team Ratio of Home wins to Away Wins
• Away Team Ratio of Draws to Total Matches
• Home Team Average goals For in last 10 matches
• Home Team Average goals For in last 40 matches
• Home Team Average goals Against in last 10 matches
• Home Team Average goals Against in last 40 matches
• Away Team Average goals For in last 10 matches
• Away Team Average goals For in last 40 matches
• Away Team Average goals Against in last 10 matches
• Away Team Average goals Against in last 40 matches
• Home Team Percentage of Home games won in one year
• Away Team Percentage of Away games won in one year
• Home Team Home Draw Percentage
• Away Team Away Draw Percentage
• Home Team relative performance vs Away team, last two matches
• Week number in season

Hope this may help.

[Ace123]

Re: predictor variables for regression model From the variables mentioned above, which ones should be most likely to come in predictive. And what R sq are people getting for their regression models? It'll just give me an idea of what to expect when im building my model. Cheers

[bigs]

Re: predictor variables for regression model Just beware of one thing, Long term data may not make sense overall! I would use at tops, 2 months data. One good idea that might come handy is: collect data from last 2 months (10 games +-) and the starting 11's (soccer case) then you can make statistics based on players, i.e. when a player starts the game, what are the % of winnings, draw and losses, do that for all players, that will "weigh'em". Then use the predictable starting 11 for the game you want, for both teams, weight it against data such as winning ratio, goals for and against, wether it's a home or away game, and make it happen!

[Sgt.Sunshine]

Re: predictor variables for regression model Have you tried this bigs?

[bigs]

Re: predictor variables for regression model Not yet, but working on it! Didn't found the time to compile all data... and still to find a site with the starting 11's for all matches from Portuguese league! Note: Just found the site, probably will start compiling data today still, because I'm watching Setubal x Sporting and that killed my poker night ;) And I have a couple of hours to spare after the game!