T-RO
Well-Known Member
- Messages
- 14,832
- Reaction score
- 16,350
The usable data is limited. Using data from previous seasons is inaccurate, because the teams have changed. Using the Cowboys as an example, we have a different LG, Tyron was injured for large portions and the LT using past seasons is different, we will have 2 different TEs, we have several new WRs, different position coaches training the players, we will supposedly have scheme changes as well. That's a lot of change in variables and doesn't even mention the changes for our opponents or other factors. Before you say you can still have accurate data based on past seasons consider the difference between our stats from 2016 and 2017, 35% is large and illustrates why using past seasons will inherently be inaccurate. Using 3rd down data isn't accurate, because again, it's a different circumstance. The best usable data is from a current season, because then you are using the same variables, but results derived from them will not be accurate because of sample size.
Small sample sizes can change wildly year to year. Meanwhile, league data on this kind of thing changes very little. As with other big data sets it’s far less variable. So the past 10 years of data (10 years x 32 teams X 80 plays per game) is extremely informative.
And third down conversion rates provide a very good proxy for fourth down conversion rates. Dynamics are basically the same.
Last edited: