Chapter 28 Exercises — Analytics and Offensive Decision-Making
Difficulty legend: ⭐ basic recall and definitions · ⭐⭐ apply the idea to a described situation · ⭐⭐⭐ judgment, trade-offs, and defending a choice · ⭐⭐⭐⭐ extension work with real film or data.
All Outlaws, Sentinels, Stallions, Pioneers, Forge, and Ospreys people, games, and numbers are fictional. Items marked (hypothetical) are practice situations, not snaps from the season described in the chapter. Where a problem uses expected points, use the illustrative ladder from 28.2 unless the problem gives you other values: 1st-and-10 at own 5 ≈ −0.3; own 20 ≈ +0.4; own 30 ≈ +1.0; midfield ≈ +2.0; opp 35 ≈ +2.8; opp 20 ≈ +3.8; opp 10 ≈ +4.6; opp 2 ≈ +5.6. Interpolate between them. Selected answers are in Appendix J.
Part A — Conceptual ⭐
A1. Define expected points (EP) in one plain-English sentence, then define EPA in one more. Why can't you calculate EPA without first having EP?
A2. State the common success-rate rule for first, second, and third/fourth down. Then name the alternative definition some analysts use, and explain why you should check which definition a number uses before comparing two teams.
A3. This book defines an explosive play as a run of 12+ yards or a pass of 20+. Explain why a dashboard needs both success rate and explosive rate — what does each one miss that the other catches?
A4. What is pass rate over expected (PROE)? Describe what a strongly negative PROE on early downs would tell you about a team's play caller.
A5. Explain the selection trap in your own words, using the Outlaws' play-action numbers (about 8.6 vs. 6.2 yards per dropback overall; about 8.6 vs. 7.0 on first and second down only).
A6. What is a break-even conversion rate on fourth down? Why is it the most useful single number in a fourth-down decision?
A7. Marisol's first honesty rule is "write 9 of 14, never 64 percent." Give two reasons the count is more honest than the percentage.
A8. Name the five assumptions a typical fourth-down model makes about the team using it.
A9. What does Wade Harmon's question — "What's the call if we go?" — have to do with the model's first assumption?
A10. Why do "perfect streaks get a question mark" on the Outlaws' scouting page?
Part B — Film, Chart & Diagram Recognition ⭐⭐
B1. Grade each snap by success rate (40/60/100) and say whether it's an explosive: (a) 1st-and-10, Ruffin inside zone +4; (b) 2nd-and-9, Powers quick out +5; (c) 3rd-and-6, Gaines dig +14; (d) 1st-and-10, Sutton back-shoulder +23; (e) 2nd-and-2, sneak +1; (f) 1st-and-goal at the 6, duo +2.
B2. Here is a dot chart in the chapter's style. Describe what it says and what you'd want to know before trusting it.
OPPONENT'S THIRD-AND-7+ vs. 2x2 -- last four games
(each B = pressure, five or more rushers; each . = four or fewer)
vs. rookie QBs BBBBBBB.
vs. veteran QBs BB.......
📊 Diagram (described): Two rows of symbols. Against rookie quarterbacks the defense brought five or more rushers on 7 of 8 long third downs. Against veteran quarterbacks it brought pressure on 2 of 9.
B3. A broadcast graphic shows "Fourth-down model: GO (+3.1% win probability)" on 4th-and-1 at the offense's own 45. The offense lines up in a spread formation with its regular personnel, and the defense shows six down linemen with both A gaps filled and both linebackers walked up. The offense throws a quick out that falls incomplete. Without arguing with the model, name two things on film that the model's recommendation didn't know.
B4. Look at the Outlaws' Inside family ledger from 28.4 (plain Inside 26 of 58 successful, about −0.02 EPA/play; Slant/Glance 18 of 31, +0.19; PA 22 calls, 12 successful, +0.41). A reporter writes, "The Outlaws' inside zone is a bad play; they should stop calling it." Using the ledger and its caveat, write a two-sentence correction.
B5. Draw the fourth-down decision tree (in the style of 28.6) for this situation: 4th-and-1 at your own 35, early second quarter, tied. A punt nets 42 yards. If you convert, assume you'll be at your own 37. If you fail, the opponent gets the ball at your 35. Use the ladder, then find the break-even.
B6. Below is a pre-snap picture on a fourth-and-1. Your Check menu is "bear = sneak; otherwise duo; safety down and both backers inside the tight ends = PA to the Y." Which play should the quarterback pick?
PRE-SNAP: 4th-and-1 -- Heavy, Under Wing Right, Check (hypothetical)
F
C C
W M $
E T N T E
U O O O O O Y
Q B
R
📊 Diagram (described): The offense is under center with two tight ends attached and a wing on the right. The defense has a nose shaded on the center and two tackles lined up out over the offensive tackles rather than over the guards, so both guards are uncovered and free to double-team. Two linebackers are at normal depth, the strong safety is walked down on the right, and one safety is deep.
B7. Marisol's slide reads: "THE NUMBER: Screens vs. 5+ rushers, 8.4 yards per call. THE SAMPLE: full season. THIS DOESN'T KNOW: — . WHAT I'D DO: call more screens." What's wrong with the slide, by her own rules? Rewrite the third and fourth lines.
B8. Use the illustrative ladder to estimate the EPA of each snap. State the "before" and "after" values you used. (a) 1st-and-10 at your own 20, completion for 30 yards to your own 50. (b) 1st-and-10 at the opponent's 35, interception returned to the opponent's 35 (they get first-and-10 there). (c) 3rd-and-8 at midfield, checkdown for 5 yards, then a punt on fourth down (treat the punt as giving the opponent first-and-10 at its own 15, worth about +0.1 to them).
Short answer
(a) About +0.4 → +2.0: ≈ +1.6. (b) Before about +2.8; after, the opponent has first-and-10 at its own 35 — that's 65 yards from your goal, roughly +1.3 to them, so −1.3 to you: ≈ −4.1. Interceptions in scoring territory are the most expensive plays in football. (c) Third-and-8 at midfield is worth well under a first-and-10 there — call it about +1.0. After the punt, the opponent's +0.1 is −0.1 to you: ≈ −1.1. The five-yard checkdown on third-and-8 was a failed play by success rate, and the EPA says why.
B9. A Marisol-style self-scout line reads: "Twelve personnel, Weeks 11–14: 22 snaps — run 16, PA 5, dropback 1, screen 0." Apply "kill the zeros" and the ~70%-on-10+ tell threshold. Is this a tell? Which "zero" would a defense notice first, and what would you add in the next two games?
Short answer
16 of 22 runs is about 73% on more than ten snaps — over the line; it's a tell. The zero is screens (and dropbacks are nearly zero). A defense would load the box against Twelve. Add a Y Screen or Slip and at least two straight dropbacks from Twelve in the next two games — real calls you'd be happy to see work, not throwaways.
Part C — Make the Call / Draw It Up ⭐⭐–⭐⭐⭐
C1. Make the call ⭐⭐ (hypothetical) Week 16-type regular-season game, third quarter, you lead 17–14. It's 3rd-and-2 at the opponent's 44, and Wade has said "Four-down." The defense has played a single-high, eight-man box on third-and-short all day. Your menu: (a) Heavy duo; (b) play-action shot, Gaines post with seven-man protection; (c) quick slants; (d) QB sneak. Make the call and explain how "Four-down" changed it.
Short answer
Because fourth down is available, third-and-2 is one of two downs to gain two yards, and the likely failure of a play-action shot is an incompletion that still leaves a fourth-and-2 inside your go zone. An eight-man box with one deep safety is exactly the picture the shot is built for. (b) is the strongest answer if your protection is sound; (a) or (d) are defensible if you'd rather have two cracks at short yardage. The wrong answer is any play with sack risk and no payoff — a sack turns a go-for-it fourth-and-2 into a punt.
C2. Make the call ⭐⭐⭐ 4th-and-goal at the 2, second quarter, trailing 10–3. Your kicker is reliable. Using rough values — a touchdown and extra point is worth about 7; a made field goal about 3; a failure leaves the opponent pinned at its own 2 (worth roughly −0.5 to you) — find the break-even conversion rate for going. Then name the call you'd want before you'd tell your head coach to go.
Short answer
Going: p × 7 + (1 − p) × (−0.5) = 7.5p − 0.5. Kicking: about 3 (ignore the small miss chance and the kickoff for this rough version). Setting them equal: 7.5p = 3.5, so p ≈ 0.47. Go if you believe you convert about half the time or better. League-wide two-yard tries convert roughly half the time, so this is close — which is exactly when your goal-line package, your opponent's front, and your saved calls decide it. Name the call: e.g., Jumbo Power Left vs. a six-down front, or Sprint Right, Snag vs. Cover 0.
C3. Draw it up ⭐⭐ Draw (pre-snap and post-snap) a quarterback sneak against a bear front in the book's notation. Label who blocks the nose, how each guard fits his 3-technique, and which A gap the quarterback takes. Add the 📊 described line.
C4. Make the call ⭐⭐⭐ (hypothetical) Fourth quarter, down 24–17, 4:40 left, one timeout. 4th-and-3 at the opponent's 33. A field goal would be 50 yards; your kicker's range line is the opponent's 35. Your 4th-and-3 calls: Snag from Trips, Mesh, and a Slip screen. The defense has played man on 11 of its 14 third-and-medium snaps today. Do you go, and what's the call? Explain why the field goal is worth less than it looks in this score.
Short answer
Go. Down 7, a field goal leaves you needing another score with under four minutes and one timeout; it does not change the number of possessions you need. The kick is also at the edge of range. Against man coverage, Mesh (or Snag with a stack/bunch release) is the man-beater; the Slip screen is a poor fourth-and-3 call against man because a defender is assigned to the back.
C5. Make the call ⭐⭐⭐ You score a touchdown with 6:10 left to cut the deficit from 14 to 8 (before the try). Kick or go for two? Explain both the case for going and the case for kicking, and say which you'd choose if your two-point play has converted 5 of your last 8 tries in practice-quality situations.
C6. Draw it up ⭐⭐–⭐⭐⭐ Build a one-page fourth-down chart (field zones × distances, GO / LG / TU / K) for a high-school team whose kicker is reliable only inside 25 yards and whose punter averages 31 yards. Explain the two biggest differences from the sample NFL chart in 28's Project Checkpoint.
C7. Make the call ⭐⭐ (hypothetical) It's second-and-10 at your own 34 after a first-down incompletion, first quarter, 0–0. Your old habit is the reset run. The defense shows two high safeties and a six-man box (an apex nickel counted as half). Your Kill call is "Inside, Kill, Smash." Which play should the quarterback keep, and what box count would flip it?
Short answer
Six in the box on second-and-long meets the tightened rule ("six or fewer, keep the run"), so Inside is live — a two-high, six-man box is the run's best picture. Seven or more flips it to Smash, and against two-high you'd expect the smash read to go to the hitch or checkdown unless the corner squats. The point of the Kill call is that the defense decides which play you get.
C8. Make the call ⭐⭐⭐ (hypothetical) You score with 2:50 left to go up 1 (before the try). The opponent has two timeouts and a kicker with a strong leg. Your two-point play has been good in practice. Kick or go? Then consider: what if you score to go up 1 with 11:00 left instead?
Short answer
With 2:50 left, go: up 3 means the opponent's field goal only ties, while up 2 lets a field goal win. Failing leaves you up 1 — the same field goal still beats you, exactly as it would at up 2, so the cost of failure is small. With 11:00 left, the game has several possessions remaining, and the two-point logic mostly washes out; kick unless your two-point play is clearly better than a coin flip.
C9. Draw it up ⭐⭐⭐ Using the book's notation, draw a one-page "Green panel" excerpt with two rows: fourth-and-short, failure cheap and fourth-and-short, failure expensive. Put two plays in each row (call grammar: personnel, formation, protection or run, concept). Under each play, write its most likely failure (incompletion, no gain, loss, sack, turnover).
Part D — Judgment & Trade-offs ⭐⭐⭐
D1. Your head coach, a former defensive coordinator, says: "Analytics doesn't know my defense is playing great today. When we're playing good defense, I punt." Using "name it, then price it," write the response you'd give him — respectfully, in no more than five sentences — and say whether his objection passes the "name it" test.
D2. Your analyst shows that your second-and-long reset run succeeds 7 of 25 times. Your offensive line coach argues that the run "keeps the edge rushers honest" for third down. Design a way to test the line coach's claim with data you can collect in four games. What would you count?
D3. Your quarterback's designed-run budget is three carries a game. It's the fourth quarter of a game you lead by three, he's already carried three times, and it's fourth-and-1 at midfield against a bear front where the sneak is clearly the best call. Do you sneak? Weigh player safety, the playoff picture, and the conversion chance — and propose a rule so the decision isn't made for the first time on the field.
D4. Marisol's family ledger supported Lyle's argument, but she said: "You win the call sheet. You don't win the argument." Explain the difference. Why is it healthy for a staff when an analyst says this?
D5. A columnist writes after a failed fourth-and-1 that cost a game: "That's what happens when you let a spreadsheet coach." Write a short reply that (a) separates the decision from the result and (b) identifies the one question about the call that the columnist should have asked instead.
D6. Your analyst's family ledger says your featured run family is efficient, but your new opponent's conflict defender (a strong safety) is coached not to react to the run at all — he plays pass first on every snap. What happens to the value of the plain run and the play-action pass against this opponent? Which member of the family should you call more, and what evidence on Monday would tell you whether you were right?
D7. Some coaches argue that a high-variance play (a deep shot) is more valuable for an underdog and a low-variance play (a run, a quick throw) is more valuable for a favorite. Explain the reasoning in terms of the distribution of outcomes, and connect it to the two calls in Case Study 2.
Part M — Mixed / Interleaved ⭐⭐–⭐⭐⭐
M1. (Ch 28 + Chapter 25) Chapter 25's threshold idea is that a play call is a bet on a distribution, not a promise of an outcome. Apply it to the Portland fourth-and-1 (stopped for no gain, Austin up 17–10, late third quarter). Write the Monday review in three lines: decision, call, execution.
M2. (Ch 28 + Chapter 7 + Chapter 9) The box-count rule "seven or fewer, run it; eight, take the throw" was born from counting. For second-and-long, the Outlaws tightened it by a man. Explain, using success rate, why the box count should matter more on second-and-10 than on first-and-10.
M3. (Ch 28 + Chapter 19) Chapter 19 says the run game is the first half of every play-action pass. Chapter 28 says some public analyses find play-action doesn't depend much on run volume. Reconcile them in one paragraph, using Teo's Holloway card (15 of 21 triggers) as your evidence.
M4. (Ch 28 + Chapter 23) The Sentinels stayed two-high on third-and-7+ against 2x2 for 29 straight snaps before Game 3. Using Chapter 23's scouting rules and this chapter's "opponent reads the data too," explain what you'd have written next to that row and what you'd have added to the call sheet.
M5. (Ch 28 + Chapter 14) Chapter 14 builds a short-yardage and goal-line package. Explain how a trusted fourth-and-1 package changes the fourth-down chart itself — not just the play — and why a model's recommendation is only as good as that package.
M6. (Ch 28 + Chapter 22) Marisol bucketed the Outlaws' 31 sacks by cause: lost the rep 12, held the ball 7, count 6, coverage 4, unblockable 2. Which bucket would a fourth-down or third-down decision change most, and which ones would only a protection or practice fix change? Explain.
Part E — Research & Film Study ⭐⭐⭐⭐
E1. Chart a season of fourth downs. Pick one NFL team and, using publicly available play-by-play (the nflverse community's data is free) or broadcast film, list every fourth down of a season: field position, distance, score, time, the decision, and the result. Then mark each one with a public model's recommendation where you can find it. Count how often the coach agreed. For every disagreement, write one "name it" line — the specific thing the coach could have seen that the model couldn't — or write "none found." What share of disagreements had a nameable reason?
E2. Build a family ledger. Choose a team known for play-action off one featured run. Over four games, chart every snap from its base play-action picture: the run, the RPOs, and the play-action passes. Record success (40/60/100) for each. Then, for one conflict defender (often a safety or the Will linebacker), chart his first step on every snap. Do the play-action results improve when he steps downhill? Report counts, not percentages, and write a "This doesn't know…" line.
E3. The two-point audit. Over one season of any league, list every two-point try in the fourth quarter with the margin before the try. Compare each decision with the chart in 28.8. Find the rows where coaches most often disagree with the chart and explain, in a paragraph, why those rows are the ones still debated.