Losing 4th Down

The Ravens went 0/3 on 4th down while the Browns converted both of their attempts.  In the simplest terms, those 5 plays were the difference in a game the Ravens otherwise slightly outplayed Cleveland.  Let’s review with 4 metrics from the Ben Baldwin Bot (BBB, @ben_bot_baldwin on twitter):

WP gain: Expected win probability gain from the decision made as opposed to the alternative

Estimated chance of success: Percentage chance to succeed on the conversion per the BBB model

WP if success: Probability of winning the game had the conversion succeeded

WP if fail: Probability of winning the game if the conversion fails

The plays:

  • (Q1, 6:58):  on 4th/1 at the Browns 7, Ricard was stacked up at the LoS, had his forward progress ruled stopped for no gain, and the Ravens turned the ball over on downs.  WP gain +2% (estimated change of success 61%), WP if success: 63%, WP if fail: 48%, WP FG: 56%
  • (Q3, 3:41):  On 4th/1 at the Ravens 9, Browns inserted G Forbes as FB who led 24 RM2 for conversion.  WP gain 1.3% (estimated success 64%), WP if success: 82%, WP if fail: 67%, WP FG: 75%
  • (Q4, 9:34):  On 4th/4, Huntley threw incomplete to Proche by the right sideline.  WP gain 2.5% (estimated success 45%), WP if success: 15%, WP if fail: 5%, WP punt: 7%
  • (Q4, 7:28):  On 4th/1 at the Baltimore 37, the Browns inserted QB Brissett who converted with a 3-yard sneak.  WP gain 1.1% (estimated success 67%), WP if success: 97%, WP if fail: 90%, WP punt: 94%   
  • (Q4, 3:10):  On 4th/13 from their own 33, Huntley threw incomplete deep right for Proche on the Ravens final offensive snap.  WP gain 0.2% (estimated success 18%), WP if success: 5%, WP if fail: 1%, WP if punt: 1%

Analyzing the decision making on these plays individually:

  • (Q1, 6:58): This is the most controversial of the decisions.  With no score, the Ravens went for a 2% expected increment to win probability with a range of outcomes that spanned 15%.  Reasons to go for it included the Browns inability to stop the run, Ricard, and the notion the Ravens were not clearly superior to the Browns.  When not clearly the superior team, scraping for every bit of WP is appropriate, including cases which have the game determined by fewer plays.  The major reason not to go is that the expected chance of success is not accurately reflected by an offense led by Huntley and the fact the chance of a TD following a conversion was probably overstated by the model.  I liked the call to run Ricard, but because that was not going to be a TD, the Ravens might have been faced with another difficult decision 3 plays later.
  • (Q3, 3:41): In a decision very similar to that made by the Ravens (above), the Browns went for it on 4th and 1 at the Ravens 9 and converted on Chubb’s 2-yar run.  Per the BBB, the Browns gained only 1.3% in expected win probability for a range of outcomes spanning 15%.  Stated otherwise, the Browns took a much bigger risk on this play relative to the payoff than did the Ravens.  That is something the superior team should avoid, since it makes the result of the game a result of fewer big wagers.  Nonetheless, the Browns succeeded and scored the game’s only TD 3 plays later, so the decision looks good.
  • (Q4, 9:34): I don’t consider the decision to go for it on 4th and 4 from their own 46 as controversial, since the Ravens were down 13-3 and any offensive (or defensive) success was likely to take more time than the model might estimate.  This was offset by the fact the BBB model probably overstates the expected probability of success with Huntley.  As modeled, the Ravens picked up 2.5% WP in a game they trailed by 10 making this an easy choice with a bad result.
  • (Q4, 7:28): The Browns gained an expected 1.1% of WP with the decision to go on a range of outcomes spanning 7%.  The decision was on par with the Ravens first attempt on 4th and 1 (Q1, 6:58, above) in terms of the gain relative to wager.  I would not have loved the decision as a Browns fan, but it helped the Browns burn 3 more minutes of clock before York’s missed FG.  That time burned essentially sealed the game.  OK decision, good result.
  • (Q4, 3:10): The Ravens gained a small increment to WP (.2%) for what the BBB models as a small risk.  It was the right choice, born of desperation, but failed.

One overarching theme is that the BBB model may be using results for the Lamar-Jackson-led Ravens in determining expected win probabilities, conversion rates, and subsequent TD rates.


Note: all snap totals exclude penalties resulting in no play, kneels, spikes, and specials team plays resulting in a run or pass.  As such, they will be lower than other published totals.

The Browns ran 61 such snaps.

Super Jumbo (1): For the second consecutive week, the Ravens lined up with all 5 active DL on a snap (I called it Jumbo last week but want to differentiate here).  The package also included JPP, Queen, Smith, Humphrey, Stone, and Clark for 4th and 1 at the Ravens 9 (Q3, 3:41).  Chubb converted and the Ravens allowed the only TD of the game 3 plays later.  1 play, 2 yards.  

Jumbo (4): The Ravens lined up with 4 DL, 3 ILB, 1 OLB, 3 DB for the Browns final 4 meaningful offensive snaps on their final 2 drives.  Urban’s drive-ending S-9 (Q4, 5:15) was among them.  4 plays, 9 yards, 2.3 YPP, 1 sack. 

Base (19): The Ravens used their base 3-4 defense for 19 snaps including 12 runs for 40 yards (3.3 YPP) and 7 passes for 13 (1.9 YPP).  19 plays, 53 yards, 2.8 YPP.  

Big Nickel (28): The Ravens used the big nickel (3 S, 2 CB) for 28 snaps.  Kyle Hamilton acted as the SCB for each of these snaps and was on the field for every nickel play.  These included 11 runs for 55 yards (5.0 YPP) and 17 passes for 114 yards (6.7 YPP). 28 plays, 169 yards, 6.0 YPP.

Jumbo Nickel (3): This included 2 versions of the nickel with 3 down linemen, 2 OLB, 1 ILB (or 1 OLB, 2 ILB), 2 CB, and 3 S.  3 plays, 19 yards, 6.3 YPP.

Rush Nickel (5): This version of the nickel included 1 down linemen, 3 OLB, and 2 ILB. All of these were big nickel and included Hamilton.  All of them came on 3rd and 6+ and 4 of these plays ended drives.  5 plays, 24 yards, 4.8 YPP.

Racecar Nickel (1): These were the first 2 packages of the season Macdonald has schemed with 4 OLBs.  Each included Ar”Darius Washington as the SCB instead of Hamilton and both ended drives.  Given the success (RL-1, S-5), we should see more.  2 plays, -6 yards, -3.0 YPP.

Pass Rush

Macdonald used numbers but very limited deception in his pass-rush scheme. 

For the game, Watson had ATS on 12 of 31 drop backs (39%).  On those plays, the Browns gained 49 yards (4.1 YPP, 1 sack).  Watson also delivered the ball before pressure could develop (BOQ) 12 times (39%, 85 yards, 7.1 YPP).  The Ravens generated a pressure event on only 7 plays (23%) which gained 6 net yards (0.9 YPP) including 2 sacks.

Summarizing by number of pass rushers:

3: 1 play, 10 yards

4: 19 plays, 74 yards, 3.9 YPP

5: 8 plays, 44 yards, 5.5 YPP, 3 sacks

6+: 3 plays, 12 yards, 4.0 YPP

Total:  31 plays, 140 yards, 4.5 YPP, 3 sacks

Macdonald dialed up 7 individual blitzes (.23 per pass play) from off the LoS, all of which were singles.  Of the 7, 3 each were from Hamilton and Queen.  On the 7 plays where they blitzed, the Ravens allowed 45 yards (6.4 YPP) with 2 sacks.   

The Ravens did not employ a single stunt, choosing instead to maintain disciplined rush lanes against Watson.

The Ravens showed simulated pressure on only 3 occasions where 2+ dropped from the LoS.  Those plays went for a total of 12 yards (4.0 YPP).

I define a deceptive pass rush as incorporating 2 or more of the above elements (off ball blitzes, stunts, and simulated pressures).  The Ravens ran only 1 deceptive rush (3%) on which the Browns gained 10 yards.

Individually, Oweh was the only Raven with 2 pressure events in less than 3 seconds (1 full, 1/2 sack).  Urban had a pressure to go along with a late S-9 on an ATS play. 

Star Treatment

Roquan Smith

  • (Q1, 13:37):  On 3rd/6, he rushed after the pocket was broken and chased Watson to the sideline to force throw away and punt
  • (Q1, 5:40):  He and Urban diagnosed the screen left for RB Chubb which was thrown incomplete at the LoS
  • (Q1, 4:58):  Watson threw to WR Cooper PR10 (3 + 7) [4] in front of Queen and Smith
  • (Q1, 4:15):  He assisted Oweh on tackle of Chubb RM1
  • (Q2, 11:03):  He worked off penetration from B. Washington and Urban’s hold of left edge to tackle Chubb RL-1 with assist from Urban
  • (Q2, 10:24):  On 3rd/11, he missed the tackle at -2 on RB Hunt’s PR3 (-5 + 8) [4], but Houston cleaned up to force punt
  • (Q2, 4:37):  He was sealed by LT Wills on Chubb RL18
  • (Q2, 2:05):  He delivered a bone-jarring assist of Bowser on Chubb RL0
  • (Q3, 7:09):  He was blocked by C Froholdt in L2 to lead Chubb RM11
  • (Q4, 11:37):  On 3rd/11, he was driven back by RG Teller on RB Hunt RM9
  • (Q4, 9:30):  He had initial contact at 1 but was dragged to tackle Chubb RR6
  • (Q4, 8:09):  He and Queen diagnosed Watson’s option read to tackle RM0
  • (Q4, 5:19):  He was knocked down by Chubb as he missed the tackle at 0, but retracked the RB into B. Washington’s tackle RM1
  • (Q4, 2:23):  He was blocked by TE Bryant to lead Chubb RM11 on the Browns last conversion

Roquan had more good than bad on defense, including some big hits around the LoS and containment of Watson.

Kyle Hamilton

  • (Q1, 6:13):  He dragged down RB Chubb after 4 YACo on RM7
  • (Q1, 3:36):  He missed the tackle on TE Njoku at 4 on PL7 (1 + 6) [1]
  • (Q1, 2:59):  On 3rd/2 with ATS, WR Cooper beat him to the edge on flat route PR4 (4 + 0) [5] to convert
  • (Q1, 1:02):  He blitzed off the slot left and beat Chubb for a fast S-7
  • (Q2, 15:00):  On 3rd/15, he took down WR Peoples-Jones for PL8 (7 + 1) [1]
  • (Q2, 5:31):  He delivered a fast tackle on Peoples-Jones for PL6 (5 + 1) [1]
  • (Q2, 2:11):  On 1st/Goal from the 4 with ATS, he closed quickly from behind to knock down a pass intended for TE Njoku 2 yards [4]
  • (Q3, 6:33):  With ATS, Watson threw to Cooper PL16 (16 + 0) [1] who beat Hamilton to the outside
  • (Q3, 5:56):  He slipped rushing the passer unblocked off the ORS and Watson threw BOQ to Cooper PM28 (12 + 16) [3]
  • (Q3, 4:16):  On 2nd/5, the Ravens rushed 6 including Hamilton.  Watson’s pass to WR Woods PL4 (3 + 1) [2] went between Kyle’s raised arms
  • (Q3, 1:37):  He drew offensive holding on WR Peoples-Jones to negate RL3 conversion on 3rd/2
  • (Q4, 12:00):  On 2nd/6, he tracked down Watson’s ill-advised pitch to Woods for immediate tackle on RL-5

Kyle had a mixed game during which the Browns looked for matchup advantages against him, but he made contributions as a downhill tackler, as a pass rusher, and in coverage.

Brent Urban

  • (Q1, 5:40):  He and Smith diagnosed screen left to RB Chubb which fell incomplete at the LoS
  • (Q1, 1:48):  He bulled then shed TE Njoku for pressure as Watson threw to the releasing Njoku PR8 (-1 + 9) [4]
  • (Q2, 11:03):  He held the left edge vs. LT Wills and assisted Smith to tackle Chubb RL-1
  • (Q4, 5:15):  On 3rd/6, he beat LT Wills inside for a late S-9 after ATS

Brent played 20 snaps and the Browns averaged just 2.8 YPP with him in the game and 5.6 YPP without him.

David Ojabo

  • (Q2, 15:00):  On 3rd/15, he entered for his 1st NFL snap.  His opponent, RT Conklin, appeared to false start by 2-3 clicks (0.07 – 0.10 seconds) but was not flagged (Michael Oher got away with a number of these in his career).  He bulled Conklin but the ball was out quickly for PL8 that ended the drive.

The game situation did not allow for much play from Ojabo, but if Campbell misses time, the Ravens may return to more 4-OLB (racecar) packages which feature him.

I do not choose defensive MVPs after a loss.