luckadjusted

Week 3 2020 · September 28, 2020

No FTN charting for this game (drops and interception-worthy throws are not modelled).

Chiefs at Ravens, Week 3 2020: who deserved to win?

The Chiefs beat the Ravens 34–20, and it was no fluke: replayed with the way both teams actually played, the Chiefs win 8 of 10. The Ravens had the bounces: kicks alone were worth 5.3 points to the Ravens.

An off day for the Ravens: their offense and defense both played far below their usual level. In a fresh game against the Chiefs they would win 50% of the time; with this day’s form, 17%. The numbers

Before kickoff the Ravens were 60% favorites. The biggest swing (Q2, 12:52 left): Return touchdown by the Ravens, worth 11 points of win probability to the Ravens.

Final

Kansas City Chiefs 34
Baltimore Ravens 20

September 28, 2020 · M&T Bank Stadium · Closing line: Ravens −3.5

Luck index +17 Chiefs about even · −17 Ravens

Each team’s own plays from this game, resampled into new drives 4,000 times: with the way both teams actually played, how often does this performance win? How the replay works

Ten replays. Chiefs win 8, Ravens win 2.

  1. 1 KC 39 BAL 9
  2. 2 KC 45 BAL 23
  3. 3 KC 52 BAL 34
  4. 4 KC 38 BAL 23
  5. 5 KC 21 BAL 10
  6. 6 KC 24 BAL 17
  7. 7 KC 40 BAL 36
  8. 8 KC 33 BAL 31
  9. 9 KC 35 BAL 38
  10. 10 KC 26 BAL 37

Three other ways to ask

Each answers a different question, so none of them replaces the replay above.

The closing betting line, turned into a probability. What everyone knew before the game.

The line plus what this game showed, weighted by how much one game can tell. It stays close to the line because one game says little.

Every kick, fumble bounce, drop, conversion and judgment call redrawn; everything else fixed. The narrowest view of luck.

An off day for the Ravens

Their offense and defense both played far below their usual level. In a fresh game against the Chiefs they would win 50% of the time; with this day’s form, 17%.

33 points apart for the Ravens: the gap is how differently this game went from what the two teams usually are.

UnitThis gameUsual levelBetter or worse than usual
Chiefs offenseEPA per play gained +0.53 +0.20 better by 0.33
Chiefs defenseEPA per play allowed −0.02 +0.11 better by 0.13
Ravens offenseEPA per play gained −0.02 +0.19 worse by 0.21
Ravens defenseEPA per play allowed +0.53 +0.07 worse by 0.46

EPA (expected points added) per play on competitive snaps, turnovers left out because they are valued as luck. Usual level: the team’s earlier games this season plus half of last season, pulled toward the league average. One team’s defense and the other team’s offense are the same snaps, so one game can’t fully say whose day it was; it shows how far this one was from both teams’ norms.

How 4,000 replays ended

Chiefs win 83% of them, Ravens 17%, counting the 1% still level after regulation as half each. The line is what actually happened: Chiefs by 14.

Show as a table
Final margin in the replaysShare of replays
Chiefs by 17+35.1%
Chiefs by 9–1622.9%
Chiefs by 1–824.4%
Level after regulation1.0%
Ravens by 1–810.7%
Ravens by 9–164.4%
Ravens by 17+1.5%

Where the luck was

Take the luck out

The final margin was Chiefs by 14. Remove the luck, one kind at a time, and it becomes Chiefs by 15.9.

Each row takes one kind of luck out of the margin. Color shows which team it helped. Luck shares are the fitted ones (how much of each component is luck).

The kicks

How likely each kicker was to make that kick (his own record, the distance and the weather), and what happened.

  • J. Tucker, 26 yds made · 99% expected
  • H. Butker, 42 yds missed · 85% expected
  • J. Tucker, 42 yds made · 85% expected
  • Extra point missed · 93% expected
Bar: expected chance of a make. Filled dot: made. Open dot: missed or blocked.

When it happened

Ravens’ chance of winning

As low as 0% and as high as 69%. Numbers mark the biggest luck moments, listed below.

nflfastR’s in-game win probability. It ignores the betting line, so it starts near even. Shaded above the middle while the Ravens were ahead in it, below while the Chiefs were.
#WhenWhat happenedWin probability forNext drive
1 Q1 13:33 The Ravens converted 3rd & 23rd & 4th downs · 62% expected +7% BAL BAL: Field goal
2 Q1 12:08 The Ravens converted 3rd & 53rd & 4th downs · 42% expected +5% BAL BAL: Field goal
3 Q1 10:06 The Ravens were stopped on 3rd & 33rd & 4th downs · 45% expected +6% KC BAL: Field goal
4 Q1 6:36 Tripping on the Ravens (Nick Boyle), 10 yards, wiping out the playJudgment calls · 7 flags in 1108 snaps before, position avg 1 per 276 +8% KC KC: Touchdown
5 Q1 2:26 The Chiefs converted 3rd & 53rd & 4th downs · 44% expected +7% KC KC: Touchdown
6 Q2 12:52 Return touchdown by the RavensSpecial teams +11% BAL BAL: Touchdown
7 Q2 1:54 The Chiefs converted 3rd & 143rd & 4th downs · 25% expected +7% KC KC: Touchdown
8 Q3 3:01 The Chiefs were stopped on 4th & 13rd & 4th downs · 77% expected +5% BAL KC: Touchdown

Missing players

Key players (usually on the field for half the snaps or more) who missed this game or part of it, valued by position and quality. The same players are missing in every replay, so this counts as luck in points only.

  • Andrew Wylie (KC T), out · 0.6 pts
  • Demone Harris (KC DE), out · 0.3 pts

Adjust the assumptions

Disagree with how much of a fumble bounce or a pass interference call is luck? Change it and the coin-flip replay reruns in your browser.

Only coin-flips re-rolled

Chiefs 100% · Ravens 0%

Final margin
Chiefs by 14
Luck in points
1.9 for the Ravens
Luck-adjusted margin
Chiefs by 15.9
Replay with same day's form
Chiefs 83% · Ravens 17% unchanged: why
How much of each is luck?

0% treats it as skill, 100% as pure luck. The tick is our fit.

What counts

We tested momentum on 19,000 lucky breaks and found none, so it's off by default. Why

Flip a play

Force the other outcome, in this game and in every replay: a made kick becomes a miss, a stop becomes a conversion, a flag goes the other way.

About these numbers

Play-by-play and win probability from nflfastR via nflverse; kicks valued against each kicker’s own record; drops and interception-worthy throws from FTN charting (not available for this game); injuries from snap counts and injury reports. No FTN charting for this game (drops and interception-worthy throws are not modelled).

Methodology and validation · All Week 3 2020 games · Chiefs · Ravens

More from Week 3 2020