clock menu more-arrow no yes mobile

Filed under:

Wild Card Watch Simulation: Week 13 – Vegas Edition

A lot can happen in six games

NFL: Washington Football Team at Buffalo Bills Rich Barnes-USA TODAY Sports

This article takes inspiration – some might say blatantly steals ideas – from Bill in Bangkok’s new series, tracking the Washington Football Team’s prospects of making the playoffs. Like Bill, each week until the season ends, or the Football Team is eliminated from contention, I am going to project the outcomes of the WFT’s remaining schedule, and the rest of the NFL for that matter.

While Bill’s weekly WFT season projections spring from an indefatigable optimism about the Team’s direction, mine comes from a different place. In this series I will attempt to embrace the broad range of outcomes which are possible in a sport where games are often decided by unpredictable events, like a kicker pulling his hamstring, the starting center breaking his leg, or the backup QB exceeding expectations after taking over for the injured starter.

Each week, I will set the stage by determining each team’s chance of winning its matchup, and then step back and let the games play out according to the will of the football gods, which will be represented by the random number generator in Microsoft Excel.

In contrast to the usual approach of calling games based on the matchup of the opposing team’s strengths and weaknesses, I will take a simpler approach based on the two teams’ recent win-loss records. The key point of departure is that my approach recognizes that, no matter how overmatched an underdog team like Taylor Heinicke’s Football Team might appear to be, there’s always a chance they’ll pull of an upset, even if it’s a tiny chance.

Just consider how we got to the point that it’s possible to talk about playoff scenarios, a mere three weeks after most of the fanbase, myself included, were ready to give up on the season. In Week 10, coming off the bye, the ragtag band of misfits in burgundy and gold, led by a quarterback less than a year removed from sleeping on his sister’s couch, took the field against the reigning Super Bowl champs, featuring an all-star roster headlined by the ageless wonder Tom Brady, and somehow managed to not only win, but beat the spread by 20 points.

What’s more, they weren’t alone. That same week featured two other major upsets with Miami beating Baltimore 22 to 10 and Carolina beating Arizona 34 to 10. Even winless Detroit managed to not lose to Pittsburgh, in a 16-point tie.

You aren’t going to call outcomes like that by just picking the favorites. While my approach isn’t likely to be particularly accurate at predicting individual game or team-season outcomes, I am hoping it illustrates the broad range of possibilities in a sport where unpredictable events frequently influence game outcomes.

In addition to making the projections probabilistic, I will also attempt to simulate another important dynamic which seems to take hold as the playoffs approach, and which we are all hoping might help propel the Football Team into the playoffs. Inspired by the post-bye-week winning streak, I will use a simple approach to attempt to simulate teams going on hot-streaks and slumps.

And who knows, by introducing an element of unpredictability into the remaining season projections, I just might come up with a scenario where the WFT wins the division, or not.

Just for fun, I’ll keep track of how well my extremely simplistic model predicts weekly game outcomes and the number of upsets per week, as well as how it compares to predictions simply based on W-L records.


Portable Device Photo by Jaap Arriens/NurPhoto via Getty Images

How It Works

Each week, until the WFT is out of it, or into the playoffs, I will project the outcome of every NFL game, using a simple method based on the two opposing teams’ W-L record. My method embodies two simple principles:

You are what your record says you are…

To save myself the effort of watching film, breaking down match ups, looking up RAS scores, and reading what the experts say to try to work out how the different teams match up from week to week, I decided to make a simplifying assumption.

When two teams meet head-to-head, the team with the better record is more likely to win. The following section contains some formulas. If you don’t like to see math formulas in football articles, skip ahead. All I am doing here is reducing that simple concept into numbers.

My process has two steps.

First, I created a simple measure of how good or bad each team is based on their win-loss record. The proportion of wins over a given number of games provides a rough estimate of the probability that a team will beat an average opponent. I’ll call that the Win Proportion, or Pwin.

For each team, each week, I calculated the Win Proportion (Pwin) over their last n games:

Team 1 Pwin = # Wins/n games

Team 2 Pwin = # Wins/n games

Second Step. The probability of Team 1 beating Team 2, p(win), can now be calculated as the ratio of Team 1’s Pwin to the combined Pwin of both teams:

p(win) = Team 1 Pwin/(Team 1 Pwin + Team 2 Pwin)

I had to make one tweak to make the model work. If a team loses five games in a row, then Pwin drops to zero, and it becomes impossible for them to ever win another game. To fix that, I set a floor on Pwin at 0.1. This means that when a team loses five in a row, Pwin is set to 0.1 instead of 0.

To see how that works, consider two teams with identical 5-5 records. When they play each other p(win) for team 1 is equal to 0.5/(0.5 + 0.5) = 0.5. The two teams are evenly matched, and the chance of one team winning is 50%. Now, imagine a 9-1 team plays a 1-9 team. p(win) for the better team is equal to 0.9/(0.9 + 0.1) = 0.9. Team 1 is much better and has a 90% chance of winning.

…But you are really only as good as your last five games

Fans of Ron Rivera’s Football Team know as well as anyone that some teams get hot and others get cold as the season progresses. By Week 12, a team’s W-L record at the beginning of the season is usually not much use in predicting how they will do in their next game. Teams’ performance is often streaky, particularly in the run up to the playoffs toward the end of the season. In fact, the whole reason we are talking about the playoffs is that the WFT has started a winning streak at just the right point in the season.

In order to simulate this streakiness, I limited the calculation of teams’ Win Proportion (Pwin) to their last five games. Setting a short window for calculating Pwin allows one or two wins in succession to have a big impact on the probability that a team will win its next game. With a five-game window, a single win can improve a team’s chance of winning the next game by close to 20%. This is intended to allow momentum shifts to occur reasonably frequently, causing teams to go on winning and losing streaks. Using a longer window, the season W-L record for example, lessens the impact of single wins and losses, and thereby has a stabilizing effect.

We don’t want stability. We want to ride the hot streak into the playoffs, while it lasts. So I went with a short anlysis window.

Final Nuts and Bolts

To pull that all together, to do my weekly projections, ahead of each game I’ll calculate the probability of one team beating the other (p(win)) as detailed above, then use the random number function (RAND) in Excel to roll the dice. A random number less than or equal to p(win) means that Team 1 beat Team 2.

Then I will use the next week’s projected win-loss results to update each team’s record and repeat the process to project game outcomes for the following week and each week thereafter to the end of the season.


NFL: Las Vegas Raiders at Dallas Cowboys Tim Heitman-USA TODAY Sports

Projected Week 13 Game Outcomes

The Football Team head into their Week 13 game having won three of their last five games for a win proportion of 0.6. They faced the Raiders who have won two of their last five games for a win proportion of 0.4. In this simulation the WFT’s probability of winning is therefore 0.6/(0.6 + 0.4) = 0.6. The random number generator rolled 0.036 giving the WFT an easy win and bringing their record up to 6-6.

The full Week 13 results are shown in the following table:

Asterisks denote upsets, defined as a team with a worse season W-L record beating a team with a better record. The model projected 5 upsets, with the most remarkable being Detroit picking up its first win of the season against Kirk Cousins’ Vikings.

**Author’s Note: this was written before the Thursday game, so it is lucky the projection held up.

Week 13 NFC Playoff Picture

After the projected Week 13 games, the NFC division leaders are:

West – Arizona (10-2)

North - Green Bay (9-3)

South – Tampa (8-4)

East – Dallas (8-4)

The NFC Wild Card chase is shaping up as follows:

8-4 LAR

7-5 SF

6-6 WFT, ATL

5-7 NYG, MIN, NO

5-8 PHI

The WFT’s win over the Raiders keeps them in the lead for the third wild card spot. Atlanta’s surprise upset of Tampa pulls them even with the WFT at 6-6, but the WFT wins the head-to-head competition tie breaker.

Season Projection

I’m not going to go into detail on the full league wide outcomes for weeks 14 through 17, because it would take up a ton of space and there’s really no point. The NFC East results for the remainder of the season, which is what we really care about, are shown in the following table.

The model projected the WFT to extend the winning streak from the Week 10 upset of Tampa with an improbable run of four straight wins through Week 13, including a sweep of Dallas. The winning streak finally came to an end with a loss to the Giants in Week 18, which is actually believable.

This is about the best case scenario for WFT fans. The key to the WFT’s improbable winning streak was the Week 13 upset of the Raiders, which pushed their Win Proportion up to 0.8. In this model, once teams get up to that level, momentum takes over and they can become hard to stop.

If you want to relate this back to reality, it simulates a team getting hot in a late season run to the playoffs, sparked by an unlikely upset win. This kind of thing does happen just about every NFL season. Although, it does seem pretty unlikely for a team playing Cinderella at QB, starting its fourth kicker of the season this week, and with its pro scouts scrambling to find available street free agents to play center in case Ismael goes down like the three previous starters before him.

Dallas managed to put a check on their losing streak, thanks to their Week 13 win over a weak New Orleans team, to coast to a 10-7 finish. However, they lose the tie breaker for the division title to the WFT emphatically due to the two head-to-head losses.

The Giants managed to post a few more wins than might be expected based on the state of their franchise, including the lucky upset of the surging WFT in Week 18, but not enough to end with a winning record. And the sad Philadelphia team managed just one more win by beating the Giants in a meaningless Week 16 contest.

Aside from the WFT overtaking Dallas, the NFC Division Champions remain unchanged since week 13:

North – Green Bay (14-3)

West – Arizona (13-4)

South – Tampa (12-5)

East – Football Team (10-7)

How the Wild Card Hunt Played Out

Flipping back to Week 13, Dallas was the 4th seeded division champ, and the wild card competition was as follows:

5th Seed Wild Card – LA Rams (7-4)

6th Seed Wild Card – San Francisco (6-5)

7th Seed Wild Card – Football Team (5-6)

Close on Washington’s tail were three teams with identical records (Minnesota, Atlanta, New Orleans) and just behind them were the 5-7 Philadelphia team and the Christian McCaffery-less Carolina Panthers.

The Rams entered the simulation at Week 13, fighting to turn around a three-game losing streak while facing a tough schedule. They managed just two more wins against Jacksonville and Baltimore, while losing to Arizona, Seattle, Minnesota, and San Francisco to finish 9-8.

Meanwhile San Francisco entered Week 13 on a three game winning streak and the momentum carried them to wins over Seattle, Cincinnati, Atlanta, Houston and the Rams to finish 11-6 and clinch the 5th seed wild card.

Dallas, as we already saw, closed the season with three wins and three losses, to finish 10-7, good enough for the 6th seed wild card. Think about it. If Dallas and the WFT meet in the playoffs, there is a possibility that the WFT could beat Dallas three times in one season. That could be the perfect WFT season.

Minnesota, New Orleans and Philadelphia all lost any momentum they might have had from Week 13, to finish 8-9, 6-11 and 6-11, respectively, and out of the playoffs.

Atlanta seemed like one of the less likely challengers to the Football Team in Week 13, holding an identical record, but a tie-break disadvantage due to a head-to-head loss, and facing a schedule including Tampa, San Francisco and Buffalo. But, like the WFT in Week 10, the Falcons upset the Bucs in Week 13, and added wins over Carolina, Buffalo and New Orleans (but losing to Detroit!) to finish 9-8, pulling even with the Rams in contention for the 7th seed wild card.

Atlanta easily wins the tiebreak for the 7th seed wild card by virtue of a far superior division record, 4-2, vs the Rams at 1-5. As a result, the wild card race ended as follows:

5th Seed Wild Card – San Francisco (11-6)

6th Seed Wild Card – Dallas (10-7)

7th Seed Wild Card – Atlanta (9-8)


NFL Tampa Bay Buccaneers at Washington Football Team Photo by Jonathan Newton/The Washington Post via Getty Images

Closing Thoughts

The point of this exercise was to create a simple simulation of the remaining NFL season, based on some realistic assumptions, which captured the unpredictability of the sport and the momentum shifts which occur in the run up to the playoffs. It worked better than I expected.

The simulation produced a few improbable outcomes, the most notable being the Football Team closing the season with a 5-1 run. Probably next to that was Detroit winning three games to finish 3-13-1.

On the other hand, aside from the WFT taking the NFC East division championship from Dallas, the NFC division champions ended pretty much as expected. LA crashing out of the playoffs and Atlanta sneaking in were the other surprises. Atlanta does seem a longshot to make the playoffs. But LA has lost three straight games to quality opponents, and still has to face Arizona, Minnesota, Baltimore and San Franciso, making their wild card berth far from guaranteed.

While I didn’t focus on the AFC playoff picture, the outcomes on that side of the bracket were also mostly within the more likely scenarios. The season ended with New England and Tennessee clinching their divisions at 11-6, and Baltimore and Cincinnati vying for the AFC North title at 10-7. The biggest surprise in the AFC playoffs was that Denver moved ahead of the Chiefs to take the West division championship at 11-6, with the Chiefs settling for a wild card seed at 10-7. While I might not bet against the Chiefs winning the West, Denver is playing well, and a division win is not out of the question. The final AFC wild card seed was 9-8 Buffalo, who won the tiebreak with Miami by virtue of two head-to-head wins.

It will be interesting to see what happens in the coming week in the actual NFL. If the WFT can pull off the road win in Las Vegas, will it keep the momentum building, or a will a loss bring them back down to earth? I’ll reset the simulation with the Week 13 game results and see what trends the new data produces. I’ll also start reporting on how this week’s model predictions compared to the actual game results.

Acknowledgement: Thanks to James Dorsett for his usual fine editorial assistance


Poll

Which game on the Football Team’s remaining schedule will be the most critical to their chance of making the playoffs?

This poll is closed

  • 21%
    This Sunday in Vegas
    (49 votes)
  • 71%
    Either Dallas game - have to win at least one
    (167 votes)
  • 1%
    Season final against the Giants
    (3 votes)
  • 6%
    One of the Phildelphia games
    (14 votes)
233 votes total Vote Now

Poll

Which player will be most critical to the WFT’s playoff run?

This poll is closed

  • 78%
    Taylor Heinicke
    (179 votes)
  • 0%
    Jonathan Allen
    (1 vote)
  • 2%
    Terry McLaurin
    (5 votes)
  • 14%
    Antonio Gibson
    (32 votes)
  • 0%
    Kam Curl
    (0 votes)
  • 0%
    Whoever starts at center against the Giants
    (2 votes)
  • 3%
    Whoever is kicking two weeks from now
    (7 votes)
  • 0%
    Alex Armah
    (2 votes)
  • 0%
    Someone else
    (0 votes)
228 votes total Vote Now