I have 10 years worth of data from NCAAF and NFL. When I get Excel, how do I make my model? Completely new to this, just wondered what the next step was after I had my data.
-
-
find some music that helps you think! then think about capping and your data while you listen to this music!Comment -
okay I just wondered how to create the actual model in Excel. Lets say I decide TO margin and YPR are the most importantComment -
When you want to see how covariates (YPR, say) affect an outcome (final margin of victory, say) the technique is called regression. Linear regression, the most familiar approach, can be done in Excel, but you might want to import your data into a statistics program to do more sophisticated analysis.Comment -
I kind of understand that, one should select the two or more variables and outcome they want and use like a scatter plot or something right. I remember some of this from statistics but I don't have Excel in front of me so I cant remember the ways to compare variablesComment -
Powell you can use the correl function in excel to see how strong the data correlatesComment -
duh thinks pal, completely forgot that. Okay onto next NOOB question...
If I find that 3rd Down Conversion Percentage, Penalty Yards, and Time of Possesion correlate, what do I do next?Comment -
I just don't get how I turn all this data into a productive model that spits out numbersComment -
Comment -
I just don't get how to take the correlated variables I like and turn them into a model. What does regression matter if you don't know how many penalties, turnovers, etc that the teams will have in a future game.Comment -
Some say a good model doesn't actually predict results, a good model shows the factors that influence resultsComment -
TP: what factors did you use to achieve a 63% ATS as in your NCAA FB sig block? Certainly you had something in mind getting to that figure. Perhaps you start with those stats.Comment -
What's next? Try to figure out how many points a team is expected to score given their 3rd Down Conversion Percentage, Penalty Yards, and Time of Possesion, etc.Comment -
What to do... make it come up with predicted spreads/final scores. Then compare the predicted spread with the actual spread...and then back test it to see if it wins X% or > amount of the time. (Presumably you'd want 55% or more)Comment -
Simple example (for illustration only): you run a linear regression of points scored on yards per passing attempt and average punt return. The model fits the equation (points) = A*(yards per passing attempt) + B*(punt return) + error term. It tells you A=2.3 and B=0.03. Then say you have a team that averages 6 yards per passing attempt and 10 yards per punt return. The predicted value under your linear model is 6*A+10*B = 6*2.3+10*0.03=14.1. Your model thinks this team will average 14.1 points.
These numbers are completely made up and a linear regression would do horribly in this setting, but that is an example of how a linear model produces predictions.Comment -
How does 1 run a linear regression? I've been checking correlations and found a few things.Comment -
-
@xbalto: How were you able to do a linear regression for two different independent variables to produce ONE equation for your dependent variable? Usually, I would see an equation for points vs. pass yards per attempt and another equation for points vs. return yards per return. So I would see two different equations in the form of y = mx + b, but somehow you created one equation with all variables. Did you just combine the two equations? Sorry if this is an amateur question.Comment -
There should be a few online engines if you do a search that can help you do a regression to help figure out what variables are important.
AS for learning how to put it all togetherand figure out whats worth it, it sounds like you need to slow down and really figure out what you're doing. Definitely take a look at some of the statistic lessons posted above. Books like Mathletics or Conquering Risk can also give you the basic how's and why's of doing a linear regression related to sports. Spend a day away from your data reading those books and I think you might have a better idea on how to approach things.Comment -
This will sound harsh, but if you cannot figure out how to run a basic linear regression scheme either off the many items you'll find on the net or in excel, you are lost as you haven't enough math sense to make this work. Find another hobby that won't burn up your money.Comment -
who are you addressing wrektangle? because i know i have enough math sense to make this work. i was just wondering about a multiple regression because i have never had to do one throughout all of my undergrad engineering projects. i have studied multiple regression by hand but have never had to implement oneComment
