2. Analyzing the NFL
• NFL play data for 2002 – 2012
• Every snap of every regular season and playoff game
• Around 450,000 total plays
3. Data Fields
• Game ID
• Quarter
• Game Minutes
• Game Seconds
• Offense Team
• Defense Team
• Down
• To Go
• Yardline (1-99)
• Description
• Offense Score
• Defense Score
• Season
4. (:23) (Shotgun) T.Brady pass short left to D.Woodhead for 14 yards TOUCHDOWN.
(2:52) (Shotgun) M.Stafford pass incomplete short right to C.Johnson.
(14:51) M.Ingram up the middle to NO 23 for 3 yards (E.Sims).
G.Hartley extra point is GOOD Center-J.Drescher Holder-C.Daniel.
Chopping Up the Descriptions
6. Managing the Data
• 2012 Season
• AFC, NFC Subsets
• Redzone Subsets
• Scrimmage Plays Subset
– No Field Goals
– No Spikes, Kneels,
Penalties, or Fumbles
– Sack = Pass
• Touchdown Score?
– Binary
7. Digging Deeper
• Logistic Regression
– Factor conversion and yard line
• Target variable: TD (binary)
• Predictor variables:
– Yard Line (1-20, categorical)
– QB Rating (Linear)
• Touchdown probability at each yard line
• QB rating influence on success rate
8. NFL Success Rate
Compare expected points of a TD try
to expected points of a field goal try:
P(FG Made)*3
12. Conclusions
• Even with an average QB, probability of a TD is over
20% all the way out to the 6 yard line.
• With a top 5 QB (QBR>100), expected touchdown
probability is 0.4 (extra point situation)
• Gamestate (score of game, time left, play strategy) is
not represented in our analysis, but it heavily
influences coaching decisions.