General NFL Stats Questions

**suicidekings** · 10-17-11, 11:15 PM

Originally posted by jbrent95

Gentlemen,

I am trying to figure out the best approach to analyse a set of NFL stats. How do I determine the best way to estimate the expected Non-offensive Scoring during the next game? It's now game 5 or 6 of the season, and I want to know how to determine the best approach. There has to be a way to analyze this. Some of my initial approaches are:

1. This seasons stats--- Miami has not had or allowed a non-offensive TD this season according to teamrankings.com. The Jets have allowed 0.6 non-offensive TDs per game and scored 0.6 non-offensive TDs per game.

2. Average this season and last season results for Non-offensive TDs per game. This doesn't seem right.

3. 20-game moving average

Any thoughts or discussion?

You might be better off starting with some research into the league wide averages for defensive scores per game over the course of several seasons. That will give you a reference point to work from. I think you'd probably be better served by breaking it down a little more and looking at the stats that lead to defensive touchdowns like fumbles/interceptions and put that into perspective with the league-wide average.

ie:

1) A team has a QB that throws __ interceptions per game and the team loses possession via fumble __ times per game (plus whatever other stats you think are relevant)
2) On average over the last 5 years, league-wide, teams score __ points per game on defense, fumble __ times per game and throw __ interceptions per game (+wtv).

Even then, this will only give you a notion of teams' tendencies, not a hard number you can apply when capping as +/- points.

**doublec** · 10-18-11, 10:14 PM

Does this include special teams scores or just defensive?

**buby74** · 10-19-11, 04:51 PM

My first thought would be to assume all teams are equally good at scoring defensive touchdowns and then see if the distribution of Tds per team-season matches a poisson distribution I would so the same for special teams although I would guess they would be a poorer fit. If the data fits a poisson distribution them you don't have to worry about 20 game averages

**jbrent95** · 10-20-11, 09:06 AM

Doublec-- This includes all non-offensive TDs (punt returns, kickoff returns, block field goals returned for a TD, and defensive scores....faked punts and field goals are not included).

**jbrent95** · 10-20-11, 09:10 AM

I guess that my questions boils down to finding a methodology that helps me determine which set of past occurrences is the most reliable indicator of future occurrences. The answer to my question is probably a regression analysis. To do that, I will likely have to build a data base.