HOW A PREDICTION GETS MADE
We don't start by guessing. Before we simulate a single game, we start from the actual point spread and point total that real sportsbooks publish before kickoff. That number already reflects an enormous amount of information — team news, injuries, public opinion, expert analysis, all of it, priced in by people risking real money to get it right.
From there, we look at decades of real NFL games to measure exactly how far final scores have historically landed from that opening number. Sometimes the favorite covers easily. Sometimes the underdog keeps it close. Sometimes the total blows right past expectations. That spread of outcomes is measured history, not a guess.
Then, instead of handing you one single prediction, we play the game out 100,000 different ways — each time nudging the score up or down using that same real-world variability — and count what actually happens across all of them. A team winning 68% of the time isn't a vibe. It's a tally.
WHY START FROM THE BETTING LINE AT ALL?
Some sites act like the betting line doesn't exist, as if a purely "independent" prediction is somehow more impressive. We think that's backwards. The market line already contains more real-time, real-money-tested information than almost anything else available before a game. Ignoring it on purpose would make our numbers worse, not better. Our job isn't to replace that number — it's to test whether we can responsibly build on top of it.
WHAT WE TESTED — AND CUT
Not every idea that sounds smart actually helps. Here's what we tested against real outcomes, and were honest about, when it didn't move the needle:
Game-Day Weather
Wind, temperature, precipitation — tested directly against real results. Once the betting line was already accounted for, weather added nothing measurable. Left out.
Indoors vs. Outdoors
Whether a game was played in a dome made no reliable difference either. A real, testable idea — it just didn't hold up. Left out.
Rest Days & Travel
Extra rest, cross-country travel, short weeks — a genuine hypothesis. It didn't add anything beyond what the betting line already captures. Left out.
None of these were dead ends — they're proof the process works. A model that never says "no" to itself isn't rigorous. It's just decorated.
WHY 100,000 GAMES, NOT ONE GUESS
Think about flipping a coin. Flip it 10 times and you might get 7 heads — that doesn't mean the coin is unfair, it just means small samples wobble. Flip it 100,000 times and the result settles in close to the true odds. We do the same thing with every matchup: instead of one guess dressed up as a percentage, we run the game forward 100,000 times and report exactly how often each outcome actually happened.
We will never round a number up to make it sound better. If the honest math says 54%, we show 54% — not a confident-sounding 65%. Sports has real randomness in it that no model erases, and we'd rather earn your trust with a number you can rely on than sell you one that just sounds good.
Most sites just hand you a percentage and ask you to trust it.
We named this site Matrix because we'd rather show you what's underneath.