TouchdownCalc rates every team from its results, plays out the rest of the season 50,000 times, and seeds every one of those seasons with the real NFL tiebreakers. A team's playoff chance is the share of those seasons in which it got in.
Where the numbers come from
The schedule, the scores and the official standings come from ESPN. We pull them again before every update.
Each update then checks itself. It adds up every team's record from the scores and compares it with the official standings. It also works out the playoff seeding with our own tiebreaker code and compares that with the official seeding. The result of that check is printed above and on the home page.
Step 1: rate every team
Each finished game gives one fact, the margin. We ask which set of team ratings explains all the margins best.
Expected margin = home rating − away rating + home field
Home field = 1.7 points (0 at a neutral site)
A blowout counts as at most 24 points
Every team is pulled toward average as if it had played 5 extra games that ended level
The pull toward average matters early in the season. Three games say little about a team. Seventeen say a lot.
A rating is in points per game against an average team. Today the highest rated team is the Jaguars, 7.0 points above average. The lowest is the Chargers, 6.7 points below average.
Step 2: turn two ratings into a win chance
Real games land around the expected margin, with a typical miss of 13.2 points. That spread turns an expected margin into a chance to win.
| Expected to win by | Margin ÷ spread | Wins |
|---|---|---|
| 0 points | 0 ÷ 13.2 = 0.00 | 50 in 100 |
| 3 points | 3 ÷ 13.2 = 0.23 | 59 in 100 |
| 7 points | 7 ÷ 13.2 = 0.53 | 70 in 100 |
| 10 points | 10 ÷ 13.2 = 0.76 | 78 in 100 |
| 14 points | 14 ÷ 13.2 = 1.06 | 86 in 100 |
The last column is the share of a bell curve below the number in the middle column.
Step 3: play out the season 50,000 times
In each run we first nudge every team's rating up or down a little, because nobody knows exactly how good a team is. Then every game left gets a random margin around its expected margin. A level game can end tied, as it can in real life.
At the end of a run every team has a final record, and we know the score of every game.
Step 4: seed each season with the real tiebreakers
Each conference sends seven teams: four division winners and three wild cards. When teams finish level, the NFL tiebreaking procedures decide the order. We follow them step by step.
Teams in the same division:
- Head to head record
- Record inside the division
- Record in common games
- Record inside the conference
- Strength of victory
- Strength of schedule
- Combined ranking in points scored and points allowed, among conference teams
- The same ranking among all teams
- Net points in common games
- Net points in all games
Teams in different divisions (wild cards and seeding): head to head, then conference record, then common games with at least four played, then the same steps from strength of victory down, with net points in conference games in place of common games.
With three or more teams level, we follow the league's rule: once a step removes a team, the teams left start again from the first step.
One step is left out. The eleventh division step, net touchdowns, needs data our feed does not carry. If a tie ever gets that far we flip a coin, which is the league's final step too.
Our guide to NFL tiebreakers shows these steps at work in real ties.
We tested this code on real seasons. We replayed the final standings of 2021 through 2025. All 160 conference places came out in the official order.
Step 5: play the playoffs
For the Super Bowl numbers we also play the bracket in every run. The top seed rests in round one. Seed 2 hosts seed 7, seed 3 hosts seed 6 and seed 4 hosts seed 5. After that the top seed always meets the lowest seed left. The better seed is at home until the Super Bowl, which is on neutral ground.
How the what if board works
When you tap a result, your own device replays the season 6,000 times with that result locked in. Nothing is sent to us.
Your pick is treated as news, not just as a score. A team that wins the game you marked is also a little more likely to be a strong team in that replay. That is why the board agrees with the "If they win" and "If they lose" numbers printed on each team page.
How we prove clinching and elimination
Clinched, eliminated and the magic numbers are not played out. They are proved. Playing a season 50,000 times can show that something is very likely. It can never show that something is certain. So for these answers we check every way the season can still finish.
The proof asks two questions about a team.
- Can it still miss? A team misses when it does not win its division and at least three teams that are not division winners finish ahead of it. If no set of results gets there, the team has clinched.
- Can it still make it? If no set of results puts it in the top seven, the team is eliminated.
Teams still play each other, so they cannot all win. We work out who can reach which record at the same time by sharing out the wins from those games exactly, with a standard method called network flow.
Tiebreakers. When the answer hangs on teams finishing level, we walk through the finishes where that happens and let the tiebreaker code from step 4 decide. Head to head, division record, common games and conference record depend only on the games of the tied teams, so they are known as soon as those games are fixed. Strength of victory needs every game of the season. A tie that goes that far is left open until six or fewer games remain, and the page says so.
Magic number. The fewest more wins that put a team in the playoffs, whatever any other team does. "Needs help" means even winning every game left is not enough on its own.
Elimination number. The fewest more losses that knock a team out, whatever else happens.
This week's scenarios. We try the results of this week's games one by one. A set of results is listed as soon as it settles the question whatever the other games do. Lists name wins and losses. A tied game is listed only when nothing else works.
When the proof cannot decide. If a search runs out of steps or meets a deep tiebreaker, the team is not called clinched or eliminated yet. The league's own flag from the official standings is shown as soon as it appears.
We tested the proof three ways.
- We replayed the 2021 to 2025 seasons week by week. The checker made 1,036 calls of clinched, eliminated or division won. Every call matched how the season ended.
- We built 550 made up late seasons with six to nine games left and counted every possible finish, ties included. That is more than 120,000 numbers and 4,600 scenarios. The checker agreed with the count every time.
- For 320 real clinches and eliminations in those five seasons we made the scenario list before the week was played. In 302 the real results were on the list. The other 18 were left open because the proof could not decide them yet. None was wrong.
How accurate is it?
We tested the method on the 2021 to 2025 seasons, and its numbers matched what happened closely. For each season we stopped the clock after every week from week 3 to week 17. The model saw only the games played by then. It gave every team a playoff chance, and we compared that chance with how the season really ended.
That gave 2,400 forecasts. The table groups them by what the model said.
| Model said (in 100) | Forecasts | Average it said | Teams that made it (in 100) |
|---|---|---|---|
| 0 to 5 | 541 | 1.0 | 0.0 |
| 5 to 15 | 237 | 9.5 | 9.7 |
| 15 to 25 | 193 | 20.2 | 20.2 |
| 25 to 35 | 167 | 29.9 | 34.1 |
| 35 to 45 | 157 | 39.9 | 37.6 |
| 45 to 55 | 146 | 49.8 | 54.1 |
| 55 to 65 | 154 | 60.1 | 59.1 |
| 65 to 75 | 169 | 69.8 | 71.6 |
| 75 to 85 | 142 | 80.0 | 78.9 |
| 85 to 95 | 182 | 90.7 | 88.5 |
| 95 to 100 | 312 | 98.9 | 98.7 |
A good forecast has the last two columns close together. A team given 70 in 100 should make it about 70 times in 100.
We also score the forecasts with the Brier score, a standard error measure where 0 is perfect. Ours is 0.117. Giving every team the same flat 7 in 16 chance scores 0.246.
What the calculator does not know
- Injuries, trades and quarterback changes. A rating only sees past scores.
- Teams that rest starters once their seed is locked.
- Weather and travel beyond the home field number.
- Early in the season the ratings are rough. That is why they are pulled toward average.
These numbers are estimates for fans. They are not betting advice.
How often it updates
The numbers are rebuilt after every game window. Every page shows the time of its last update at the top.
Changes to the method
- 14 November 2026: first version.
- 14 November 2026: clinched, eliminated and the magic numbers are now proved from every result still possible. Before, they came from the played out seasons.