Risky Business
EdjSports examines critical decisions and their impact on GWC (Game-Winning Chance).
Deshaun Watson
Photo: USA Today Sports Images

In Week 7 we addressed an unusual situation where the Detroit Lions’ defense intentionally allowed the Atlanta Falcons’ Todd Gurley to score near the goal line late in the game. Gurley was aware of the ploy but couldn’t hit the brakes quickly enough. Moments later, the Lions capitalized with a game-winning drive. This is a very specialized tactic that doesn’t arise often and is typically used as a desperate measure.

One of those instances occurred yesterday in Houston as the Texans were threatening to take the lead from the Colts late in the game. While the game’s ending will be remembered for the Deshaun Watson fumble, there were some fascinating decisions that preceded that conclusion.

Before we go deep on whether the Colts should have considered letting the Texans score prior to the fumble, let’s look at what transpired in the 4th quarter.

Anthony (“AJ”) Jones of EdjSports prepares an NFL analysis and lesson each week for high school coaches that subscribe to the EdjVarsity analytics tool. Here is his breakdown of an exciting fourth quarter between the Colts and the Texans.

Indianapolis Fourth Down Decision

With 11:07 remaining in the fourth quarter and leading by 4 points (both teams had all their timeouts), Indianapolis faced a first-and-10 from their own 32-yard line. And, after a 10-play drive that covered 63 yards, the Colts faced fourth-and-1 on the Houston 5-yard line.

If you were Coach Reich, would you attempt the field goal (in hopes of going up by 7 points) or would you go for it (in hopes of extending the drive, scoring a touchdown, and making it a two-score game)? And, how big of a decision is this?

Coach Reich made the correct decision by going for it, increasing the Colts (pre-snap) win probability by +4.0%. Based on how these teams match up and the game state, the Colts' decision to go for it comes with an expectation to go on to win the game 89.7% of the time, while a decision to attempt the field goal comes with and expectation to go on to win the game 85.7% of the time.

Our model has the ability to evaluate all possible outcomes associated with field goal attempts as well as going for it -- which includes taking into consideration the opponent’s resulting field position in each case. And, all recommendations are based on what gives a team the best chance to go on to win the game!

The Colts went for it and failed to convert on this fourth-and-1, but left the Texans backed up deep in their own territory. And, on third-and-11 from the Houston 4-yard line, Texans QB Deshaun Watson was sacked in the end zone for a safety. While the game announcers (and probably most viewers) were highly critical of the Coach Reich’s decision (based on the outcome), few people will consider how it indirectly contributed to the Texans safety.

Obviously, Coach Reich wasn’t expecting his defense to come up with a safety if his offense failed to convert, but rest assured he understands that analytics take that (unlikely) possibility into consideration, so he doesn’t have to.

Run or Pass?

After receiving the post-safety kick, the Colts put together a drive that ran just over three minutes off the clock and forced the Texans to burn all three of their timeouts, in an effort to preserve time. Ultimately, on fourth-and-12 from Indianapolis 49-yard line, the Colts' punt resulted in a touchback. The Texans took possession at their own 20-yard line but quickly drove the length of the field. With 1:28 remaining in regulation and trailing by 6 points, the Texans faced a second-and-goal from the Indianapolis 2-yard line.

If you were calling plays for the Texans offense, would you call a run or pass play in this situation, and does it make a difference?

Based on how these teams match up and the game state, on this play the best decision is to call a run play. If the Texans call a run play they are expected to win the game 53.2% of the time and if they call a pass play they are expected to win the game 46.2% of the time. The 7.0% difference between these two options can largely be attributed to the value of time. On this play, if the Texans score the go-ahead touchdown, the Colts comeback efforts will be pressed for time. If the Texans don’t score on this play (barring a turnover), a run play keeps the clock running and further reduces the Colts chances to mount a comeback, if the Texans are able to score.

Let them Score?

On the flip side, if you were calling plays for the Colts defense would you let the Texans score on this play, in hopes of giving your offense as much time as possible? If the Texans score a touchdown, the Colts can regain the lead with a field goal (or touchdown).

Based on how these teams match up and the game state, it’s a close call but the best decision is to let the Texans score the touchdown. Let’s assume the Texans score (running 8 seconds off the clock), convert the extra point, and their kickoff results in a touchback. Taking possession with 1:20 remaining, no timeouts, and trailing by 1 point, the Colts would be expected to win the game 47.8% of the time. Conversely, the Colts will have a 46.8% GWC if they play defense on second down. So, allowing the Texans to score would actually increase the Colts win probability by +1.0%.

As AJ describes, the final moments of this contest were a case study in game theory.

  • Should the Texans run or pass?
  • Should the Colts accept the high probability that the Texans score a touchdown and immediately let them score to preserve clock for a final drive and game-winning field goal attempt?
  • If the Colts let the Texans score, should they accept the offer or take a knee just before the goal line, run the clock down further, and ensure they have enough time for two attempts at the endzone without giving the Colts any time to respond if they do score?

How often do the Texans score a touchdown?

Assuming they have three opportunities, this is basically the equivalent of a two-point conversion attempt. Based upon this matchup we estimate the Texans have a 45.5% chance of converting from the 2-yard line. Therefore, their chance of scoring on the final possession is approximately 1-(.545³) = 83.8%. With many iterations in the simulation considered, the Texans are expected to win the game 53.2% of the time.

What are the Colts’ chances if trailing by 26-27 with no timeouts and 1:20 remaining after a touchback?

As indicated above, the simulation assesses there is enough time for Phillip Rivers to execute a game-winning drive 47.8% of the time.

What is the Texans’ game-winning chance (GWC) if the Colts try to let them score but they stop short of the goal line, run the clock down to 40 seconds, and have two attempts at a touchdown inside the 1-yard line?

Because of the better field position and the reduction in game clock, the Texans’ GWC would jump to nearly 60%.

Conclusion: It appears the Colts would have slightly improved their GWC by allowing the Texans to score right away BUT ONLY IF the Texans take the bait. If the Texans can manage to exploit the ploy by stopping just short then they will actually improve their winning chances considerably, and the strategy would backfire on Frank Reich. It is hard to say how the Texans would have responded, but we know from what happened to Todd Gurley against the Lions in Week 7 that it is not very easy to execute this type of restraint near the goal line.


2 comments, Last at 08 Dec 2020, 4:54am

1 Moving forward, but not scoring

"With 1:28 remaining in regulation and trailing by 6 points, the Texans faced a second-and-goal from the Indianapolis 2-yard line."

Comparing w/the Lions vs. Falcons game, Gurley actually had ~5 yards to think about going down or actually being accidentally tackled. When a runner is on the 2 yard line, by the time he realizes that the defense isn't trying to tackle him, he has already broken the plane. I mean, if the Texans actually wanted to gain just one yard to run clock without scoring, the best play would probably be a QB sneak. There is no risk of a fumbled hand-off, almost no risk of a loss of yardage, and almost no risk of a fumbled snap. Shotgun snaps are slightly riskier, and the risk of loss of yardage is higher. 

BUT, this is not a situation where two teams are tied, or the trailing team can just take a couple of kneel-downs, run clock, and then kick an almost automatic GW FG. The Texans needed a TD just to tie, and the XP to go-ahead. Scoring the TD after burning another 60-80 seconds would be best, but there is no guarantee. A simple false-start or holding penalty would extremely complicate the situation. On one hand, any gain on a passing play will almost assuredly result in the TD needed. On the other hand, a straight-ahead-type run carries the least risk. 

So, in my book, the best result analytically would be to run a QB sneak on at least 2nd and 3rd down. I am not saying that the Texans should have done this; but it would not have been the worst idea on 2nd down. Come out with split WR's, 1 yard off the line (so that a slant is a threat, but so is a fade--also gives space against the DB); 2 TE's, one to either side; and a RB in the I. Gives Watson the opportunity to have defenders spread out instead of concentrated in the middle, and gives him a chance to audible if he thinks a pass to either receiver is a better choice based on the defensive alignment. The O-line can still run block, even if it's a quick pass. 

Still, the biggest thing to remember is that the TD is not a given, and not scoring or advancing the ball on 2nd down decreases the opportunity to score that TD.

2 Error bars and unlikely events

I would be confident in asserting that margin for error on the GWC calculations is at least 3x the 1% gain the calculations reveal by letting the Titans score.

I also don't think the scenarios included a lot of "snap the ball off the QB's knee"; I'm sure they included fumbles, but the Titans don't fumble that much anyway.