Dan
11-19-2005, 11:32 PM
Football Sabermetrics, The Future of Football Statistical Analysis
By Dan
Ok, so my sig says I’m PSH’s sabermetrician, but you may not know what that means or why I claim to be it, but I will explain both here. Sabermetrics is the scientific study of sports statistics. It has been revolutionizing baseball statistics, but hasn’t really been applied to football to the same extent, and I hope that it will be in the future. There obviously have been sabermetric applications in football such as the BCS and the quarterback rating system, but there is more that can be done. I will explore two different applications of sabermetrics in this article, analyzing teams and players, and using statistics to make play calling decisions. I will try to stay away from too much math jargon for those of you who do not have degrees in math.
The first application is analysis of players and teams. This is actually where I started in sabermetrics was trying to find a formula to analyze NFL teams. One problem in analyzing football teams is that there are so many fewer games, so there is less opportunity to the best teams to separate themselves from the rest of the league, which is why NFL tie-breaking procedures have become so complex. This is what inspired me to begin my sabermetric work, which began based on the tie-breaking procedure.
I start out by putting the teams in order of record, using the tie-breakers to break all ties. This can get somewhat complicated when teams in different conferences are tied or there are large numbers of teams tied. I have tried to get as close as possible to the NFL tie-breaking procedure when breaking these ties. This gives me a ranking by record 1-32 of all NFL teams, which is the slightly more important of the two parts of the system.
I am still working on developing the other half of the system, but I will discuss what I am working on. The idea is that I am trying to factor in the different elements of the tie-breaking procedure into a points system. I’m not sure exactly what gave me the idea for that, but I think it was because I was trying to analyze team with different records to see who is really better and I thought about tie-breakers. The factors that I use are record, conference record, division record, opponents beaten, net points, net division points, net conference points, strength of schedule, and net touchdowns. I started out trying to just add these factors together, but what I’m now working on is making all of the record factors together being equally weighted with all of the net point factors together and the strength of schedule factor. I’m hoping to come up with a system that balances the different factors and accurately analyzes the teams. Then I rank the teams 1-32 based on ratings. The final step is to put the rankings together and get a final ranking.
I think this system has potential as method for analyzing NFL teams, but it does need some work. An interesting fact about this is that I actually was able to improve my system when I started using a computer rather than pencil and paper because it was easier to see patterns in the numbers.
The second application of sabermetrics in football is situational and risk analysis. This was actually what I wrote my college thesis on, using probability to analyze sports statistics. I talked about both football and baseball in my thesis, but here I will only talk about football, since this is a football site. If anyone is interested in learning about my baseball sabermetric work, they are free to contact me.
The first thing I looked at was two-point conversions versus conventional extra points, and I expected that conventional extra points would be successful about twice as often as two point conversions. This would mean that mathematically, a team could expect the same number of points from extra points, regardless of what type, and over the long term, it wouldn’t matter. In reality, what I found was that conventional extra points are successful less than twice as often as two-point conversions. This means that a team could expect to get more points out of two point conversions than conventional extra points over the long term.
It may be just because I’m a math geek and a huge sports fan, but I’m always excited by any types of sabermetrics. I hope to bring sabermetric analysis to PSH, and continue to be PSH’s sabermetrician.
By Dan
Ok, so my sig says I’m PSH’s sabermetrician, but you may not know what that means or why I claim to be it, but I will explain both here. Sabermetrics is the scientific study of sports statistics. It has been revolutionizing baseball statistics, but hasn’t really been applied to football to the same extent, and I hope that it will be in the future. There obviously have been sabermetric applications in football such as the BCS and the quarterback rating system, but there is more that can be done. I will explore two different applications of sabermetrics in this article, analyzing teams and players, and using statistics to make play calling decisions. I will try to stay away from too much math jargon for those of you who do not have degrees in math.
The first application is analysis of players and teams. This is actually where I started in sabermetrics was trying to find a formula to analyze NFL teams. One problem in analyzing football teams is that there are so many fewer games, so there is less opportunity to the best teams to separate themselves from the rest of the league, which is why NFL tie-breaking procedures have become so complex. This is what inspired me to begin my sabermetric work, which began based on the tie-breaking procedure.
I start out by putting the teams in order of record, using the tie-breakers to break all ties. This can get somewhat complicated when teams in different conferences are tied or there are large numbers of teams tied. I have tried to get as close as possible to the NFL tie-breaking procedure when breaking these ties. This gives me a ranking by record 1-32 of all NFL teams, which is the slightly more important of the two parts of the system.
I am still working on developing the other half of the system, but I will discuss what I am working on. The idea is that I am trying to factor in the different elements of the tie-breaking procedure into a points system. I’m not sure exactly what gave me the idea for that, but I think it was because I was trying to analyze team with different records to see who is really better and I thought about tie-breakers. The factors that I use are record, conference record, division record, opponents beaten, net points, net division points, net conference points, strength of schedule, and net touchdowns. I started out trying to just add these factors together, but what I’m now working on is making all of the record factors together being equally weighted with all of the net point factors together and the strength of schedule factor. I’m hoping to come up with a system that balances the different factors and accurately analyzes the teams. Then I rank the teams 1-32 based on ratings. The final step is to put the rankings together and get a final ranking.
I think this system has potential as method for analyzing NFL teams, but it does need some work. An interesting fact about this is that I actually was able to improve my system when I started using a computer rather than pencil and paper because it was easier to see patterns in the numbers.
The second application of sabermetrics in football is situational and risk analysis. This was actually what I wrote my college thesis on, using probability to analyze sports statistics. I talked about both football and baseball in my thesis, but here I will only talk about football, since this is a football site. If anyone is interested in learning about my baseball sabermetric work, they are free to contact me.
The first thing I looked at was two-point conversions versus conventional extra points, and I expected that conventional extra points would be successful about twice as often as two point conversions. This would mean that mathematically, a team could expect the same number of points from extra points, regardless of what type, and over the long term, it wouldn’t matter. In reality, what I found was that conventional extra points are successful less than twice as often as two-point conversions. This means that a team could expect to get more points out of two point conversions than conventional extra points over the long term.
It may be just because I’m a math geek and a huge sports fan, but I’m always excited by any types of sabermetrics. I hope to bring sabermetric analysis to PSH, and continue to be PSH’s sabermetrician.