Milestone goals
Spring break goals
Our spring break goals consist of finishing the modeling of the project and getting ready for the podcasts. Our schedules unfortunately do not line up all too well but the work can be done mostly separately, with our modeling work on our computers. We will also begin work on collecting data for the upcoming draft class so that we can use our model to project each player's draft position in our podcast. We are closing down on the end of our process here, which is very exciting to see all of our work coming together.
WInter break goals
Our winter break goals consist of getting a completed prototype of our model. Because we only have one draft class worth of data so far, our model will obviously remain semi-inaccurate. Even still, it remains very important to build a strong base so that we can add more and more stats towards the end of January and our model will come together in full strength. This model will be built with the guidance of Dr. Groves and it will hopefully be fully functional. While it is still a bit away, we will also be doing preliminary preparations for our podcasts, as we hope to start them by mid-February. Over break, this will mostly consist of simply locating objects like microphones and recording equipment.
Fall Goals
Throughout the fall we expect to devote much of our time to importing the entirety of our data into SPSS. Having spent much of the summer collecting data in excel, we have quite a large amount of data to export to SPSS. As we enter the section of the project that will involve a great deal of data analysis we will begin to use Dr. Groves as an even larger resource. With Dr. Groves' knowledge and experience with data analysis, she will be a huge resource for us going forward. Once we are able to implement all of our data for the 2012 Draft Class in SPSS, we plan on running a multiple stepwise regression to begin working towards our final goal of an equation that will allow us to predict a player's NFL Draft value. A multiple stepwise regression has a single dependent variable (the success of an NFL player) that is predicted by using multiple independent variables (college and NFL combine stats). Additionally, we hope to work through our current roadblocks. The largest of these roadblocks is the varying collegiate competition levels and how to value players who play in different conferences or divisions when comparing them.
Summer Goals
Over the course of the summer, we will complete the "learn" stage of the project first. By becoming completely confident in our programming abilities prior to the remainder of the project, we will minimize our mistakes throughout the process. In the “learn” stage of the project we will test different programs in order to decide on one to use for the remainder of the project. We have tested and studied Stata, SPSS, SAS/STAT, and Minitab. These are all modeling programs that have similar layouts to Microsoft Excel, but are able to run much deeper calculations. Essentially, we need these programs to be able to perform statistical calculations and experiments that will reveal which attributes make a "good" NFL player. Based on our understanding, SPSS is the most intuitive and easiest to pick up and learn in a short period of time, especially since there are plenty of YouTube tutorials explaining the program. While it isn't too complex, SPSS still satisfies our needs, the biggest of which is running a multiple stepwise regression. This type of regression is where a single dependent variable (the success of an NFL player) is predicted by using more than one independent variables (college and NFL combine stats). Hypothetically, these types of calculations are possible by hand, but would take an unrealistic amount of time and mathematical knowledge to complete. Therefore, these programs are essentially our super-calculators for this project.
With a full understanding of the capabilities of the SPSS program, we will begin with a bare-bones model to get us started as we head into the school year. Our goal is to have a model that can predict the success of the running back position with relative accuracy. Since we are new to statistics and modeling, we believe that our best course of action is to make consistent, small steps to our model instead of trying to make our final model all in one try. We aim to finish the first model by the end of summer and then continue to add new components to it until it is finalized. This model will be made by running the multiple stepwise regression with a few dozen independent statistics that apply to running backs. These statistics will all be scaled based on percentiles from 0 to 1 in order to make the statistics weighted equally. This regression will result in a prediction of the player's success. The player's success will be dictated by grades given to each NFL player from an NFL statistic site called Pro Football Focus (PFF). Statisticians there grade each play to come up with a number that effectively grades every NFL player based on how successful they are compared to the average player. While it will never be a perfect system, this PFF grade is the most streamline and accurate conveyor of a player's success.
With a full understanding of the capabilities of the SPSS program, we will begin with a bare-bones model to get us started as we head into the school year. Our goal is to have a model that can predict the success of the running back position with relative accuracy. Since we are new to statistics and modeling, we believe that our best course of action is to make consistent, small steps to our model instead of trying to make our final model all in one try. We aim to finish the first model by the end of summer and then continue to add new components to it until it is finalized. This model will be made by running the multiple stepwise regression with a few dozen independent statistics that apply to running backs. These statistics will all be scaled based on percentiles from 0 to 1 in order to make the statistics weighted equally. This regression will result in a prediction of the player's success. The player's success will be dictated by grades given to each NFL player from an NFL statistic site called Pro Football Focus (PFF). Statisticians there grade each play to come up with a number that effectively grades every NFL player based on how successful they are compared to the average player. While it will never be a perfect system, this PFF grade is the most streamline and accurate conveyor of a player's success.
presentations and documents
Provisional fellows presentation |
|
official fellows presentation |
|
Fellows application |
|