How the Sports Game Simulator Works

This page explains exactly what happens between picking two teams and reading a final score, so you can judge how much weight the numbers deserve.

Two inputs, one scoreline

Every result comes from two things: a strength rating for each team on a 0 to 100 scale, and a home-field advantage worth about 1.7 points. The rating is an editorial estimate — built from how a team is generally regarded, not from a live data feed — and it is deliberately coarse. A rating of 88 and a rating of 90 are treated as roughly the same team.

Home-field advantage is applied to the home side only. Flip the venue with the swap button and the same pairing produces a slightly different spread, which is the cleanest way to see what those 1.7 points are worth.

From ratings to possessions

A football game is a sequence of possessions, so the simulator works in possessions rather than in a single formula.

Each team gets 10 to 12 drives. For every drive, the model draws a random number and compares it against two thresholds:

• A touchdown threshold, which rises as the team's rating advantage grows. A two-point conversion is attempted on a small share of touchdowns.
• A field goal threshold, which sits just behind it.
• A small turnover band and a punt band fill out the rest.

That construction is why scores look like football scores. Points arrive in sevens and threes, so 27–20 and 24–17 show up while 26–19 does not. A drive that ends in a turnover also explains why a team can out-gain an opponent and still lose.

Each drive is then placed into one of the four quarters, weighted slightly toward the second and third. Tied games go to overtime, where a field goal decides it most of the time.

Why the win probability is not just the score

The line underneath the scoreboard replays the same matchup 1,000 times and counts how often each team finishes ahead. That is a different question from "who won this one run", and it is usually the more useful one.

• A 65–35 split with a three-point final score means the game was close but the underlying gap is real.
• A 51–49 split means the teams are effectively even, and the single score you are looking at is close to a coin flip.

This is also why pressing "Simulate again" is a legitimate way to use the tool. Ten runs of the same matchup tell you far more than one.

The box score

Total yards, passing yards, rushing yards, turnovers, sacks allowed, third-down conversions and time of possession are generated from the same simulated game, so the numbers agree with the score. Time of possession is split near 30 minutes for each side with a lean toward the winning team; passing yards land near 60% of total yards, which is the modern NFL norm.

These are plausible supporting numbers, not play-by-play data. There is no drive chart, no player-level stat line and no fourth-down decision model.

What the model does not know

No injuries, no weather, no short weeks, no travel distance, no coaching changes, no depth chart. It also does not adjust for a neutral site, so for a Super Bowl-style matchup the right move is to read the win probability and treat the home edge as noise.

If you want a sharper answer than this page can give, the honest path is a data source with real season results and a model fitted to it. That is a different build, and it is worth doing only after this version has an audience.

Fair use

Everything here is generated for entertainment. It is not betting advice, not a prediction of a real game and not affiliated with the NFL, MLB, the NCAA or any team.